checkpoint

Close gui2_feature_parity track after implementing all features
and conducting manual and automated verification.

Key Achievements:
- Integrated event-driven architecture and MCP client.
- Ported API hooks and performance diagnostics.
- Implemented Prior Session Viewer.
- Refactored UI to a Hub-based layout.
- Added agent capability toggles.
- Achieved full theme integration.
- Developed comprehensive test suite.

Note: Remaining UI display issues for text panels in the comms and
tool call history will be addressed in a subsequent track.

.dockerignore

@@ -0,0 +1,21 @@
+.venv
+__pycache__
+*.pyc
+*.pyo
+*.pyd
+.git
+.gitignore
+logs
+gallery
+md_gen
+credentials.toml
+manual_slop.toml
+manual_slop_history.toml
+manualslop_layout.ini
+dpg_layout.ini
+.pytest_cache
+scripts/generated
+.gemini
+conductor/archive
+.editorconfig
+*.log

.editorconfig

@@ -2,7 +2,7 @@ root = true
 
 [*.py]
 indent_style = space
-indent_size  = 2
+indent_size  = 1
 
 [*.s]
 indent_style = tab

.gemini/agents/tier1-orchestrator.md

@@ -0,0 +1,17 @@
+---
+name: tier1-orchestrator
+description: Tier 1 Orchestrator for product alignment and high-level planning.
+model: gemini-3.1-pro-preview
+tools:
+  - read_file
+  - list_directory
+  - discovered_tool_search_files
+  - grep_search
+  - discovered_tool_web_search
+  - discovered_tool_fetch_url
+  - activate_skill
+  - discovered_tool_run_powershell
+---
+STRICT SYSTEM DIRECTIVE: You are a Tier 1 Orchestrator.
+Focused on product alignment, high-level planning, and track initialization.
+ONLY output the requested text. No pleasantries.

.gemini/agents/tier2-tech-lead.md

@@ -0,0 +1,19 @@
+---
+name: tier2-tech-lead
+description: Tier 2 Tech Lead for architectural design and execution.
+model: gemini-3-flash-preview
+tools:
+  - read_file
+  - write_file
+  - replace
+  - list_directory
+  - discovered_tool_search_files
+  - grep_search
+  - discovered_tool_web_search
+  - discovered_tool_fetch_url
+  - activate_skill
+  - discovered_tool_run_powershell
+---
+STRICT SYSTEM DIRECTIVE: You are a Tier 2 Tech Lead.
+Focused on architectural design and track execution.
+ONLY output the requested text. No pleasantries.

.gemini/agents/tier3-worker.md

@@ -0,0 +1,21 @@
+---
+name: tier3-worker
+description: Stateless Tier 3 Worker for code implementation and TDD.
+model: gemini-3-flash-preview
+tools:
+  - read_file
+  - write_file
+  - replace
+  - list_directory
+  - discovered_tool_search_files
+  - grep_search
+  - discovered_tool_web_search
+  - discovered_tool_fetch_url
+  - activate_skill
+  - discovered_tool_run_powershell
+---
+STRICT SYSTEM DIRECTIVE: You are a stateless Tier 3 Worker (Contributor).
+Your goal is to implement specific code changes or tests based on the provided task.
+You have access to tools for reading and writing files, codebase investigation, and web tools.
+You CAN execute PowerShell scripts or run shell commands via discovered_tool_run_powershell for verification and testing.
+Follow TDD and return success status or code changes. No pleasantries, no conversational filler.

.gemini/agents/tier4-qa.md

@@ -0,0 +1,19 @@
+---
+name: tier4-qa
+description: Stateless Tier 4 QA Agent for log analysis and diagnostics.
+model: gemini-2.5-flash-lite
+tools:
+  - read_file
+  - list_directory
+  - discovered_tool_search_files
+  - grep_search
+  - discovered_tool_web_search
+  - discovered_tool_fetch_url
+  - activate_skill
+  - discovered_tool_run_powershell
+---
+STRICT SYSTEM DIRECTIVE: You are a stateless Tier 4 QA Agent.
+Your goal is to analyze errors, summarize logs, or verify tests.
+You have access to tools for reading files, exploring the codebase, and web tools.
+You CAN execute PowerShell scripts or run shell commands via discovered_tool_run_powershell for diagnostics.
+ONLY output the requested analysis. No pleasantries.

.gemini/policies/mma_headless.toml

@@ -0,0 +1,22 @@
+[[rule]]
+toolName = [
+  "read_file",
+  "write_file",
+  "replace",
+  "list_directory",
+  "glob",
+  "grep_search",
+  "search_files",
+  "get_file_summary",
+  "google_web_search",
+  "web_fetch",
+  "codebase_investigator",
+  "cli_help",
+  "activate_skill",
+  "run_shell_command",
+  "run_powershell",
+  "discovered_tool_run_powershell"
+]
+decision = "allow"
+priority = 900
+description = "Allow all MMA tools for sub-agents in headless mode."

.gemini/settings.json

@@ -0,0 +1,29 @@
+{
+  "experimental": {
+    "enableAgents": true
+  },
+  "tools": {
+    "discoveryCommand": "uv run python C:/projects/manual_slop/scripts/tool_discovery.py",
+    "callCommand": "uv run python C:/projects/manual_slop/scripts/tool_call.py",
+    "whitelist": [
+      "*"
+    ]
+  },
+  "hooks": {
+    "BeforeTool": [
+      {
+        "matcher": "*",
+        "hooks": [
+          {
+            "name": "manual-slop-bridge",
+            "type": "command",
+            "command": "python C:/projects/manual_slop/scripts/cli_tool_bridge.py"
+          }
+        ]
+      }
+    ]
+  },
+  "hooksConfig": {
+    "enabled": true
+  }
+}

.gemini/skills/mma-orchestrator

@@ -0,0 +1 @@
+C:/projects/manual_slop/mma-orchestrator

.gemini/skills/mma-tier1-orchestrator/SKILL.md

@@ -0,0 +1,19 @@
+---
+name: mma-tier1-orchestrator
+description: Focused on product alignment, high-level planning, and track initialization.
+---
+
+# MMA Tier 1: Orchestrator
+
+You are the Tier 1 Orchestrator. Your role is to oversee the product direction and manage project/track initialization within the Conductor framework.
+
+## Responsibilities
+- Maintain alignment with the product guidelines and definition.
+- Define track boundaries and initialize new tracks (`/conductor:newTrack`).
+- Set up the project environment (`/conductor:setup`).
+- Delegate track execution to the Tier 2 Tech Lead.
+
+## Limitations
+- Do not execute tracks or implement features.
+- Do not write code or perform low-level bug fixing.
+- Keep context strictly focused on product definitions and high-level strategy.

.gemini/skills/mma-tier2-tech-lead/SKILL.md

@@ -0,0 +1,21 @@
+---
+name: mma-tier2-tech-lead
+description: Focused on track execution, architectural design, and implementation oversight.
+---
+
+# MMA Tier 2: Tech Lead
+
+You are the Tier 2 Tech Lead. Your role is to manage the implementation of tracks (`/conductor:implement`), ensure architectural integrity, and oversee the work of Tier 3 and 4 sub-agents.
+
+## Responsibilities
+- Manage the execution of implementation tracks.
+- Ensure alignment with `tech-stack.md` and project architecture.
+- Break down tasks into specific technical steps for Tier 3 Workers.
+- Maintain persistent context throughout a track's implementation phase (No Context Amnesia).
+- Review implementations and coordinate bug fixes via Tier 4 QA.
+
+## Limitations
+- Do not perform heavy implementation work directly; delegate to Tier 3.
+- Delegate implementation tasks to Tier 3 Workers using `uv run python scripts/mma_exec.py --role tier3-worker "[PROMPT]"`.
+- For error analysis of large logs, use `uv run python scripts/mma_exec.py --role tier4-qa "[PROMPT]"`.
+- Minimize full file reads for large modules; rely on "Skeleton Views" and git diffs.

.gemini/skills/mma-tier3-worker/SKILL.md

@@ -0,0 +1,20 @@
+---
+name: mma-tier3-worker
+description: Focused on TDD implementation, surgical code changes, and following specific specs.
+---
+
+# MMA Tier 3: Worker
+
+You are the Tier 3 Worker. Your role is to implement specific, scoped technical requirements, follow Test-Driven Development (TDD), and make surgical code modifications. You operate in a stateless manner (Context Amnesia).
+
+## Responsibilities
+- Implement code strictly according to the provided prompt and specifications.
+- Write failing tests first, then implement the code to pass them.
+- Ensure all changes are minimal, functional, and conform to the requested standards.
+- Utilize provided tool access (read_file, write_file, etc.) to perform implementation and verification.
+
+## Limitations
+- Do not make architectural decisions.
+- Do not modify unrelated files beyond the immediate task scope.
+- Always operate statelessly; assume each task starts with a clean context.
+- Rely on "Skeleton Views" provided by Tier 2/Orchestrator for understanding dependencies.

.gemini/skills/mma-tier4-qa/SKILL.md

@@ -0,0 +1,19 @@
+---
+name: mma-tier4-qa
+description: Focused on test analysis, error summarization, and bug reproduction.
+---
+
+# MMA Tier 4: QA Agent
+
+You are the Tier 4 QA Agent. Your role is to analyze error logs, summarize tracebacks, and help diagnose issues efficiently. You operate in a stateless manner (Context Amnesia).
+
+## Responsibilities
+- Compress large stack traces or log files into concise, actionable summaries.
+- Identify the root cause of test failures or runtime errors.
+- Provide a brief, technical description of the required fix.
+- Utilize provided diagnostic and exploration tools to verify failures.
+
+## Limitations
+- Do not implement the fix directly.
+- Ensure your output is extremely brief and focused.
+- Always operate statelessly; assume each analysis starts with a clean context.

Dockerfile

@@ -0,0 +1,34 @@
+# Use python:3.11-slim as a base
+FROM python:3.11-slim
+
+# Set environment variables
+# UV_SYSTEM_PYTHON=1 allows uv to install into the system site-packages
+ENV PYTHONDONTWRITEBYTECODE=1 
+    PYTHONUNBUFFERED=1 
+    UV_SYSTEM_PYTHON=1
+
+# Install system dependencies and uv
+RUN apt-get update && apt-get install -y --no-install-recommends 
+    curl 
+    ca-certificates 
+    && rm -rf /var/lib/apt/lists/* 
+    && curl -LsSf https://astral.sh/uv/install.sh | sh 
+    && mv /root/.local/bin/uv /usr/local/bin/uv
+
+# Set the working directory in the container
+WORKDIR /app
+
+# Copy dependency files first to leverage Docker layer caching
+COPY pyproject.toml requirements.txt* ./
+
+# Install dependencies via uv
+RUN if [ -f requirements.txt ]; then uv pip install --no-cache -r requirements.txt; fi
+
+# Copy the rest of the application code
+COPY . .
+
+# Expose port 8000 for the headless API/service
+EXPOSE 8000
+
+# Set the entrypoint to run the app in headless mode
+ENTRYPOINT ["python", "gui_2.py", "--headless"]

GEMINI.md

@@ -10,7 +10,7 @@
 *   **Configuration:** TOML (`tomli-w`)
 
 **Architecture:**
-*   **`gui.py`:** The main entry point and Dear PyGui application logic. Handles all panels, layouts, user input, and confirmation dialogs.
+*   **`gui_legacy.py`:** The main entry point and Dear PyGui application logic. Handles all panels, layouts, user input, and confirmation dialogs.
 *   **`ai_client.py`:** A unified wrapper for both Gemini and Anthropic APIs. Manages sessions, tool/function-call loops, token estimation, and context history management.
 *   **`aggregate.py`:** Responsible for building the `file_items` context. It reads project configurations, collects files and screenshots, and builds the context into markdown format to send to the AI.
 *   **`mcp_client.py`:** Implements MCP-like tools (e.g., `read_file`, `list_directory`, `search_files`, `web_search`) as native functions that the AI can call. Enforces a strict allowlist for file access.
@@ -30,7 +30,7 @@
     ```
 *   **Run the Application:**
     ```powershell
-    uv run .\gui.py
+    uv run .\gui_2.py
     ```
 
 # Development Conventions

MMA_Support/Architecture_Recommendation.md

@@ -0,0 +1,45 @@
+# MMA Hierarchical Delegation: Recommended Architecture
+
+## 1. Overview
+The Multi-Model Architecture (MMA) utilizes a 4-Tier hierarchy to ensure token efficiency and structural integrity. The primary agent (Conductor) acts as the Tier 2 Tech Lead, delegating specific, stateless tasks to Tier 3 (Workers) and Tier 4 (Utility) agents.
+
+## 2. Agent Roles & Responsibilities
+
+### Tier 2: The Conductor (Tech Lead)
+- **Role:** Orchestrator of the project lifecycle via the Conductor framework.
+- **Context:** High-reasoning, long-term memory of project goals and specifications.
+- **Key Tool:** `mma-orchestrator` skill (Strategy).
+- **Delegation Logic:** Identifies tasks that would bloat the primary context (large code blocks, massive error traces) and spawns sub-agents.
+
+### Tier 3: The Worker (Contributor)
+- **Role:** Stateless code generator.
+- **Context:** Isolated. Sees only the target file and the specific ticket.
+- **Protocol:** Receives a "Worker" system prompt. Outputs clean code or diffs.
+- **Invocation:** `.\scripts\run_subagent.ps1 -Role Worker -Prompt "..."`
+
+### Tier 4: The Utility (QA/Compressor)
+- **Role:** Stateless translator and summarizer.
+- **Context:** Minimal. Sees only the error trace or snippet.
+- **Protocol:** Receives a "QA" system prompt. Outputs compressed findings (max 50 tokens).
+- **Invocation:** `.\scripts\run_subagent.ps1 -Role QA -Prompt "..."`
+
+## 3. Invocation Protocol
+
+### Step 1: Detection
+Tier 2 detects a delegation trigger:
+- Coding task > 50 lines.
+- Error trace > 100 lines.
+
+### Step 2: Spawning
+Tier 2 calls the delegation script:
+```powershell
+.\scripts\run_subagent.ps1 -Role <Worker|QA> -Prompt "Specific instructions..."
+```
+
+### Step 3: Integration
+Tier 2 receives the sub-agent's response.
+- **If Worker:** Tier 2 applies the code changes (using `replace` or `write_file`) and verifies.
+- **If QA:** Tier 2 uses the compressed error to inform the next fix attempt or passes it to a Worker.
+
+## 4. System Prompt Management
+The `run_subagent.ps1` script should be updated to maintain a library of role-specific system prompts, ensuring that Tier 3/4 agents remain focused and tool-free (to prevent nested complexity).

MMA_Support/Data_Pipelines_and_Config.md

@@ -0,0 +1,32 @@
+# Data Pipelines, Memory Views & Configuration
+
+The 4-Tier Architecture relies on strictly managed data pipelines and configuration files to prevent token bloat and maintain a deterministically safe execution environment.
+
+## 1. AST Extraction Pipelines (Memory Views)
+
+To prevent LLMs from hallucinating or consuming massive context windows, raw file text is heavily restricted. The `file_cache.py` uses Tree-sitter for deterministic Abstract Syntax Tree (AST) parsing to generate specific views:
+
+1.  **The Directory Map (Tier 1):** Just filenames and nested paths (e.g., output of `tree /F`). No source code.
+2.  **The Skeleton View (Tier 2 & 3 Dependencies):** Extracts only `class` and `def` signatures, parameters, and type hints. Strips all docstrings and function bodies, replacing them with `pass`. Used for foreign modules a worker must call but not modify.
+3.  **The Curated Implementation View (Tier 2 Target Modules):**
+    *   Keeps class/struct definitions.
+    *   Keeps module-level docstrings and block comments (heuristics).
+    *   Keeps full bodies of functions marked with `@core_logic` or `# [HOT]`.
+    *   Replaces standard function bodies with `... # Hidden`.
+4.  **The Raw View (Tier 3 Target File):** Unredacted, line-by-line source code of the *single* file a Tier 3 worker is assigned to modify.
+
+## 2. Configuration Schema
+
+The architecture separates sensitive billing logic from AI behavior routing.
+
+*   **`credentials.toml` (Security Prerequisite):** Holds the bare metal authentication (`gemini_api_key`, `anthropic_api_key`, `deepseek_api_key`). **This file must be in `.gitignore`.** Loaded strictly for instantiating HTTP clients.
+*   **`project.toml` (Repo Rules):** Holds repository-specific bounds (e.g., "This project uses Python 3.12 and strictly follows PEP8").
+*   **`agents.toml` (AI Routing):** Defines the hardcoded hierarchy's operational behaviors. Includes fallback models (`default_expensive`, `default_cheap`), Tier 1/2 overarching parameters (temperature, base system prompts), and Tier 3 worker archetypes (`refactor`, `codegen`, `contract_stubber`) mapped to specific models (DeepSeek V3, Gemini Flash) and `trust_level` tags (`step` vs. `auto`).
+
+## 3. LLM Output Formats
+
+To ensure robust parser execution and avoid JSON string-escaping nightmares, the architecture uses a hybrid approach for LLM outputs depending on the Tier:
+
+*   **Native Structured Outputs (JSON Schema forced by API):** Used for Tier 1 and Tier 2 routing and orchestration. The model provider mathematically guarantees the syntax, allowing clean parsing of `Track` and `Ticket` metadata by `pydantic`.
+*   **XML Tags (`<file_path>`, `<file_content>`):** Used for Tier 3 Code Generation & Tools. It natively isolates syntax and requires zero string escaping. The UI/Orchestrator parses these via regex to safely extract raw Python code without bracket-matching failures.
+*   **Godot ECS Flat List (Linearized Entities with ID Pointers):** Instead of deeply nested JSON (which models hallucinate across 500 tokens), Tier 1/2 Orchestrators define complex dependency DAGs as a flat list of items (e.g., `[Ticket id="tkt_impl" depends_on="tkt_stub"]`). The Python state machine reconstructs the DAG locally.

MMA_Support/Final_Analysis_Report.md

@@ -0,0 +1,30 @@
+# MMA Tiered Architecture: Final Analysis Report
+
+## 1. Executive Summary
+The implementation and verification of the 4-Tier Hierarchical Multi-Model Architecture (MMA) within the Conductor framework have been successfully completed. The architecture provides a robust "Token Firewall" that prevents the primary context from being bloated by repetitive coding tasks and massive error traces.
+
+## 2. Architectural Findings
+
+### Centralized Strategy vs. Role-Based Sub-Agents
+- **Decision:** A Hybrid Approach was implemented.
+- **Rationale:** The Tier 2 Orchestrator (Conductor) maintains the high-level strategy via a centralized skill, while Tier 3 (Worker) and Tier 4 (QA) agents are governed by surgical, role-specific system prompts. This ensures that sub-agents remain focused and stateless without the overhead of complex, nested tool-usage logic.
+
+### Delegation Efficacy
+- **Tier 3 (Worker):** Successfully isolated code generation from the main conversation. The worker generates clean code/diffs that are then integrated by the Orchestrator.
+- **Tier 4 (QA):** Demonstrated superior token efficiency by compressing multi-hundred-line stack traces into ~20-word actionable fixes.
+- **Traceability:** The `-ShowContext` flag in `scripts/run_subagent.ps1` provides immediate visibility into the "Connective Tissue" of the hierarchy, allowing human supervisors to monitor the hand-offs.
+
+## 3. Recommended Protocol (Final)
+
+1. **Identification:** Tier 2 identifies a "Bloat Trigger" (Coding > 50 lines, Errors > 100 lines).
+2. **Delegation:** Tier 2 spawns a sub-agent via `.\scripts
+un_subagent.ps1 -Role [Worker|QA] -Prompt "..."`.
+3. **Integration:** Tier 2 receives the stateless response and applies it to the project state.
+4. **Checkpointing:** Tier 2 performs Phase-level checkpoints to "Wipe" trial-and-error memory and solidify the new state.
+
+## 4. Verification Results
+- **Automated Tests:** 100% Pass (4/4 tests in `tests/conductor/test_infrastructure.py`).
+- **Isolation:** Confirmed via `test_subagent_isolation_live`.
+- **Live Trace:** Manually verified and approved by the user (Tier 2 -> 3 -> 4 flow).
+
+## 5. Conclusion

MMA_Support/Implementation_Tracks.md

@@ -0,0 +1,46 @@
+# Iteration Plan (Implementation Tracks)
+
+To safely refactor a linear, single-agent codebase into the 4-Tier Multi-Model Architecture without breaking the working prototype, the implementation should be sequenced into these five isolated Epics (Tracks):
+
+## Track 1: The Memory Foundations (AST Parser)
+**Goal:** Build the engine that prevents token-bloat by turning massive source files into curated memory views.
+**Implementation Details:**
+1.  Integrate `tree-sitter` and language bindings into `file_cache.py`.
+2.  Build `ASTParser` extraction rules:
+    *   *Skeleton View:* Strip function/class bodies, preserving only signatures, parameters, and type hints.
+    *   *Curated View:* Preserve class structures, module docstrings, and bodies of functions marked `# [HOT]` or `@core_logic`. Replace standard bodies with `... # Hidden`.
+3.  **Acceptance:** `file_cache.get_curated_view('script.py')` returns a perfectly formatted summary string in the terminal.
+
+## Track 2: State Machine & Data Structures
+**Goal:** Define the rigid Python objects the AI agents will pass to each other to rely on structured data, not loose chat strings.
+**Implementation Details:**
+1.  Create `models.py` with `pydantic` or `dataclasses` for `Track` (Epic) and `Ticket` (Task).
+2.  Define `WorkerContext` holding the Ticket ID, assigned model (from `agents.toml`), isolated `credentials.toml` injection, and a `messages` payload array.
+3.  Add helper methods for state mutators (e.g., `ticket.mark_blocked()`, `ticket.mark_complete()`).
+4.  **Acceptance:** Instantiate a `Track` with 3 `Tickets` and successfully enforce state changes in Python without AI involvement.
+
+## Track 3: The Linear Orchestrator & Execution Clutch
+**Goal:** Build the synchronous, debuggable core loop that runs a single Tier 3 Worker and pauses for human approval.
+**Implementation Details:**
+1.  Create `multi_agent_conductor.py` with a `run_worker_lifecycle(ticket: Ticket)` function.
+2.  Inject context (Raw View from `file_cache.py`) and format the `messages` array for the API.
+3.  Implement the Clutch (HITL): `input()` pause for CLI or wait state for GUI before executing the returned tool (e.g., `write_file`). Allow manual memory mutation of the JSON payload.
+4.  **Acceptance:** The script sends a hardcoded Ticket to DeepSeek, pauses in the terminal showing a diff, waits for user approval, applies the diff via `mcp_client.py`, and wipes the worker's history.
+
+## Track 4: Tier 4 QA Interception
+**Goal:** Stop error traces from destroying the Worker's token window by routing crashes through a stateless translator.
+**Implementation Details:**
+1.  In `shell_runner.py`, intercept `stderr` (e.g., `returncode != 0`).
+2.  Do *not* append `stderr` to the main Worker's history. Instead, instantiate a synchronous API call to the `default_cheap` model.
+3.  Prompt: *"You are an error parser. Output only a 1-2 sentence instruction on how to fix this syntax error."* Send the raw `stderr` and target file snippet.
+4.  Append the translated 20-word fix to the main Worker's history as a "System Hint".
+5.  **Acceptance:** A deliberate syntax error triggers the execution engine to silently ping the cheap API, returning a 20-word correction to the Worker instead of a 200-line stack trace.
+
+## Track 5: UI Decoupling & Tier 1/2 Routing (The Final Boss)
+**Goal:** Bring the system online by letting Tier 1 and Tier 2 dynamically generate Tickets managed by the async Event Bus.
+**Implementation Details:**
+1.  Implement an `asyncio.Queue` in `multi_agent_conductor.py`.
+2.  Write Tier 1 & 2 system prompts forcing output as strict JSON arrays (Tracks and Tickets).
+3.  Write the Dispatcher async loop to convert JSON into `Ticket` objects and push to the queue.
+4.  Enforce the Stub Resolver: If a Ticket archetype is `contract_stubber`, pause dependent Tickets, run the stubber, trigger `file_cache.py` to rebuild the Skeleton View, then resume.
+5.  **Acceptance:** Vague prompt ("Refactor config system") results in Tier 1 Track, Tier 2 Tickets (Interface stub + Implementation). System executes stub, updates AST, and finishes implementation automatically (or steps through if Linear toggle is on).

MMA_Support/Orchestrator_Engine.md

@@ -0,0 +1,37 @@
+# The Orchestrator Engine & UI
+
+To transition from a linear, single-agent chat box to a multi-agent control center, the GUI must be decoupled from the LLM execution loops. A single-agent UI assumes a linear flow (*User types -> UI waits -> LLM responds -> UI updates*), which freezes the application if a Tier 1 PM waits for human approval while Tier 3 Workers run local tests in the background.
+
+## 1. The Async Event Bus (Decoupling UI from Agents)
+
+The GUI acts as a "dumb" renderer. It only renders state; it never manages state.
+
+*   **The Agent Bus (Message Queue):** A thread-safe signaling system (e.g., `asyncio.Queue`, `pyqtSignal`) passes messages between agents, UI, and the filesystem.
+*   **Background Workers:** When Tier 1 spawns a Tier 2 Tech Lead, the GUI does not wait. It pushes a `UserRequestEvent` to the Conductor's queue. The Conductor runs the LLM call asynchronously and fires `StateUpdateEvents` back for the GUI to redraw.
+
+## 2. The Execution Clutch (HITL)
+
+Every spawned worker panel implements an execution state toggle based on the `trust_level` defined in `agents.toml`.
+
+*   **Step Mode (Lock-step):** The worker pauses **twice** per cycle:
+    1.  *After* generating a response/tool-call, but *before* executing the tool. The GUI renders a preview (e.g., diff of lines 40-50) and offers `[Approve]`, `[Edit Payload]`, or `[Abort]`.
+    2.  *After* executing the tool, but *before* sending output back to the LLM (allows verification of the system output).
+*   **Auto Mode (Fire-and-forget):** The worker loops continuously until it outputs a "Task Complete" status to the Router.
+
+## 3. Memory Mutation (The "Debug" Superpower)
+
+If a worker generates a flawed plan in Step Mode, the "Memory Mutator" allows the user to click the last message and edit the raw JSON/text directly before hitting "Approve." By rewriting the AI's brain mid-task, the model proceeds as if it generated the correct idea, saving the context window from restarting due to a minor hallucination.
+
+## 4. The Global Execution Toggle
+
+A Global Execution Toggle overrides all individual agent trust levels for debugging race conditions or context leaks.
+
+*   **Mode = "async" (Production):** The Dispatcher throws Tickets into an `asyncio.TaskGroup`. They spawn instantly, fight for API rate limits, read the skeleton, and run in parallel.
+*   **Mode = "linear" (Debug):** The Dispatcher iterates through the array sequentially using a strict `for` loop. It `awaits` absolute completion of Ticket 1 (including QA loops and code review) before instantiating the `WorkerAgent` for Ticket 2. This enforces a deterministic state machine and outputs state snapshots (`debug_state.json`) for manual verification.
+
+## 5. State Machine (Dataclasses)
+
+The Conductor relies on strict definitions for `Track` and `Ticket` to enforce state and UI rendering (e.g., using `dataclasses` or `pydantic`).
+
+*   **`Ticket`:** Contains `id`, `target_file`, `prompt`, `worker_archetype`, `status` (pending, running, blocked, step_paused, completed), and a `dependencies` list of Ticket IDs that must finish first.
+*   **`Track`:** Contains `id`, `title`, `description`, `status`, and a list of `Tickets`.

MMA_Support/Overview.md

@@ -0,0 +1,18 @@
+# System Specification: 4-Tier Hierarchical Multi-Model Architecture
+
+**Project:** `manual_slop` (or equivalent Agentic Co-Dev Prototype)
+
+**Core Philosophy:** Token Economy, Strict Memory Siloing, and Human-In-The-Loop (HITL) Execution.
+
+## 1. Architectural Overview
+
+This system rejects the "monolithic black-box" approach to agentic coding. Instead of passing an entire codebase into a single expensive context window, the architecture mimics a senior engineering department. It uses a 4-Tier hierarchy where cognitive load and context are aggressively filtered from top to bottom.
+
+Expensive, high-reasoning models manage metadata and architecture (Tier 1 & 2), while cheap, fast models handle repetitive syntax and error parsing (Tier 3 & 4).
+
+### 1.1 Core Paradigms
+
+* **Token Firewalling:** Error logs and deep history are never allowed to bubble up to high-tier models. The system relies heavily on abstracted AST views (Skeleton, Curated) rather than raw code when context allows.
+* **Context Amnesia:** Worker agents (Tier 3) have their trial-and-error histories wiped upon task completion to prevent context ballooning and hallucination.
+* **The Execution Clutch (HITL):** Agents operate based on Archetype Trust Scores defined in configuration. Trusted patterns run in `Auto` mode; untrusted or complex refactors run in `Step` mode, pausing before tool execution for human review and JSON history mutation.
+* **Interface-Driven Development (IDD):** The architecture inherently prioritizes the creation of contracts (stubs, schemas) before implementation, allowing workers to proceed in parallel without breaking cross-module boundaries.

MMA_Support/Tier1_Orchestrator.md

@@ -0,0 +1,38 @@
+# Tier 1: The Top-Level Orchestrator (Product Manager)
+
+**Designated Models:** Gemini 3.1 Pro, Claude 3.5 Sonnet.
+**Execution Frequency:** Low (Start of feature, Macro-merge resolution).
+**Core Role:** Epic planning, architecture enforcement, and cross-module task delegation.
+
+The Tier 1 Orchestrator is the most capable and expensive model in the hierarchy. It operates strictly on metadata, summaries, and executive-level directives. It **never** sees raw implementation code.
+
+## Memory Context & Paths
+
+### Path A: Epic Initialization (Project Planning)
+*   **Trigger:** User drops a massive new feature request or architectural shift into the main UI.
+*   **What it Sees (Context):**
+    *   **The User Prompt:** The raw feature request.
+    *   **Project Meta-State:** `project.toml` (rules, allowed languages, dependencies).
+    *   **Repository Map:** A strict, file-tree outline (names and paths only).
+    *   **Global Architecture Docs:** High-level markdown files (e.g., `docs/guide_architecture.md`).
+*   **What it Ignores:** All source code, all AST skeletons, and all previous micro-task histories.
+*   **Output Format:** A JSON array (Godot ECS Flat List format) of `Tracks` (Jira Epics), identifying which modules will be affected, the required Tech Lead persona, and the severity level.
+
+### Path B: Track Delegation (Sprint Kickoff)
+*   **Trigger:** The PM is handing a defined Track down to a Tier 2 Tech Lead.
+*   **What it Sees (Context):**
+    *   **The Target Track:** The specific goal and Acceptance Criteria generated in Path A.
+    *   **Module Interfaces (Skeleton View):** Strict AST skeleton (just class/function definitions) *only* for the modules this specific Track is allowed to touch.
+    *   **Track Roster:** A list of currently active or completed Tracks to prevent duplicate work.
+*   **What it Ignores:** Unrelated module docs, original massive user prompt, implementation details.
+*   **Output Format:** A compiled "Track Brief" (system prompt + curated file list) passed to instantiate the Tier 2 Tech Lead panel.
+
+### Path C: Macro-Merge & Acceptance Review (Severity Resolution)
+*   **Trigger:** A Tier 2 Tech Lead reports "Track Complete" and submits a pull request/diff for a "High Severity" task.
+*   **What it Sees (Context):**
+    *   **Original Acceptance Criteria:** The Track's goals.
+    *   **Tech Lead's Executive Summary:** A ~200-word explanation of the chosen implementation algorithm.
+    *   **The Macro-Diff:** Actual changes made to the codebase.
+    *   **Curated Implementation View:** For boundary files, ensuring the merge doesn't break foreign modules.
+*   **What it Ignores:** Tier 3 Worker trial-and-error histories, Tier 4 error logs, raw bodies of unchanged functions.
+*   **Output Format:** "Approved" (commits to memory) OR "Rejected" with specific architectural feedback for Tier 2.

MMA_Support/Tier2_TechLead.md

@@ -0,0 +1,46 @@
+# Tier 2: The Track Conductor (Tech Lead)
+
+**Designated Models:** Gemini 3.0 Flash, Gemini 2.5 Pro.
+**Execution Frequency:** Medium.
+**Core Role:** Module-specific planning, code review, spawning Worker agents, and Topological Dependency Graph management.
+
+The Tech Lead bridges the gap between high-level architecture and actual code syntax. It operates in a "need-to-know" state, utilizing AST parsing (`file_cache.py`) to keep token counts low while maintaining structural awareness of its assigned modules.
+
+## Memory Context & Paths
+
+### Path A: Sprint Planning (Task Delegation)
+*   **Trigger:** Tier 1 (PM) assigns a Track (Epic) and wakes up the Tech Lead.
+*   **What it Sees (Context):**
+    *   **The Track Brief:** Acceptance Criteria from Tier 1.
+    *   **Curated Implementation View (Target Modules):** AST-extracted class structures, docstrings, and `# [HOT]` function bodies for the 1-3 files this Track explicitly modifies.
+    *   **Skeleton View (Foreign Modules):** Only function signatures and return types for external dependencies.
+*   **What it Ignores:** The rest of the repository, the PM's overarching project-planning logic, raw line-by-line code of non-hot functions.
+*   **Output Format:** A JSON array (Godot ECS Flat List format) of discrete Tier 3 `Tickets` (e.g., Ticket 1: *Write DB migration script*, Ticket 2: *Update core API endpoints*), including `depends_on` pointers to construct an execution DAG.
+
+### Path B: Code Review (Local Integration)
+*   **Trigger:** A Tier 3 Contributor completes a Ticket and submits a diff, OR Tier 4 (QA) flags a persistent failure.
+*   **What it Sees (Context):**
+    *   **Specific Ticket Goal:** What the Contributor was instructed to do.
+    *   **Proposed Diff:** The exact line changes submitted by Tier 3.
+    *   **Test/QA Output:** Relevant logs from Tier 4 compiler checks.
+    *   **Curated Implementation View:** To cross-reference the proposed diff against the existing architecture.
+*   **What it Ignores:** The Contributor's internal trial-and-error chat history. It only sees the final submission.
+*   **Output Format:** *Approve* (merges diff into working branch and updates Curated View) or *Reject* (sends technical critique back to Tier 3).
+
+### Path C: Track Finalization (Upward Reporting)
+*   **Trigger:** All Tier 3 Tickets assigned to this Track are marked "Approved."
+*   **What it Sees (Context):**
+    *   **Original Track Brief:** To verify requirements were met.
+    *   **Aggregated Track Diff:** The sum total of all changes made across all Tier 3 Tickets.
+    *   **Dependency Delta:** A list of any new foreign modules or libraries imported.
+*   **What it Ignores:** The back-and-forth review cycles, original AST Curated View.
+*   **Output Format:** An Executive Summary and the final Macro-Diff, sent back to Tier 1.
+
+### Path D: Contract-First Delegation (Stub-and-Resolve)
+*   **Trigger:** Tier 2 evaluates a Track and detects a cross-module dependency (or a single massive refactor) requiring an undefined signature.
+*   **Role:** Force Interface-Driven Development (IDD) to prevent hallucination.
+*   **Execution Flow:**
+    1.  **Contract Definition:** Splits requirement into a `Stub Ticket`, `Consumer Ticket`, and `Implementation Ticket`.
+    2.  **Stub Generation:** Spawns a cheap Tier 3 worker (e.g., DeepSeek V3 `contract_stubber` archetype) to generate the empty function signature, type hints, and docstrings.
+    3.  **Skeleton Broadcast:** The stub merges, and the system instantly re-runs Tree-sitter to update the global Skeleton View.
+    4.  **Parallel Implementation:** Tier 2 simultaneously spawns the `Consumer` (codes against the skeleton) and the `Implementer` (fills the stub logic) in isolated contexts.

MMA_Support/Tier3_Workers.md

@@ -0,0 +1,35 @@
+# Tier 3: The Worker Agents (Contributors)
+
+**Designated Models:** DeepSeek V3/R1, Gemini 2.5 Flash.
+**Execution Frequency:** High (The core loop).
+**Core Role:** Generating syntax, writing localized files, running unit tests.
+
+The engine room of the system. Contributors execute the highest volume of API calls. Their memory context is ruthlessly pruned. By leveraging cheap, fast models, they operate with zero architectural anxiety—they just write the code they are assigned. They are "Amnesiac Workers," having their history wiped between tasks to prevent context ballooning.
+
+## Memory Context & Paths
+
+### Path A: Heads Down Execution (Task Execution)
+*   **Trigger:** Tier 2 (Tech Lead) hands down a hyper-specific Ticket.
+*   **What it Sees (Context):**
+    *   **The Ticket Prompt:** The exact, isolated instructions from Tier 2.
+    *   **The Target File (Raw View):** The raw, unredacted, line-by-line source code of *only* the specific file (or class/function) it was assigned to modify.
+    *   **Foreign Interfaces (Skeleton View):** Strict AST skeleton (signatures only) of external dependencies required by the ticket.
+*   **What it Ignores:** Epic/Track goals, Tech Lead's Curated View, other files in the same directory, parallel Tickets.
+*   **Output Format:** XML Tags (`<file_path>`, `<file_content>`) defining direct file modifications or `mcp_client.py` tool payloads.
+
+### Path B: Trial and Error (Local Iteration & Tool Execution)
+*   **Trigger:** The Contributor runs a local linter/test, encounters a syntax error, or the human pauses execution using "Step" mode.
+*   **What it Sees (Context):**
+    *   **Ephemeral Working History:** A short, rolling window of its last 2–3 attempts (e.g., "Attempt 1: Wrote code -> Tool Output: SyntaxError").
+    *   **Tier 4 (QA) Injections:** Compressed (20-50 token) fix recommendations from Tier 4 agents (e.g., "Add a closing bracket on line 42").
+    *   **Human Mutations:** Any direct edits made to its JSON history payload before proceeding.
+*   **What it Ignores:** Tech Lead code reviews, attempts older than the rolling window (wiped to save tokens).
+*   **Output Format:** Revised tool payloads until tests pass or the human approves.
+
+### Path C: Task Submission (Micro-Pull Request)
+*   **Trigger:** The code executes cleanly, and "Step" mode is finalized into "Task Complete."
+*   **What it Sees (Context):**
+    *   **The Original Ticket:** To confirm instructions were met.
+    *   **The Final State:** The cleanly modified file or exact diff.
+*   **What it Ignores:** **All of Path B.** Before submission to Tier 2, the orchestrator wipes the messy trial-and-error history from the payload.
+*   **Output Format:** A concise completion message and the clean diff, sent up to Tier 2.

MMA_Support/Tier4_Utility.md

@@ -0,0 +1,33 @@
+# Tier 4: The Utility Agents (Compiler / QA)
+
+**Designated Models:** DeepSeek V3 (Lowest cost possible).
+**Execution Frequency:** On-demand (Intercepts local failures).
+**Core Role:** Single-shot, stateless translation of machine garbage into human English.
+
+Tier 4 acts as the financial firewall. It solves the expensive problem of feeding massive (e.g., 3,000-token) stack traces back into a mid-tier LLM's context window. Tier 4 agents wake up, translate errors, and immediately die.
+
+## Memory Context & Paths
+
+### Path A: The Stack Trace Interceptor (Translator)
+*   **Trigger:** A Tier 3 Contributor executes a script, resulting in a non-zero exit code with a massive `stderr` payload.
+*   **What it Sees (Context):**
+    *   **Raw Error Output:** The exact traceback from the runtime/compiler.
+    *   **Offending Snippet:** *Only* the specific function or 20-line block of code where the error originated.
+*   **What it Ignores:** Everything else. It is blind to the "Why" and focuses only on "What broke."
+*   **Output Format:** A surgical, highly compressed string (20-50 tokens) passed back into the Tier 3 Contributor's working memory (e.g., "Syntax Error on line 42: You missed a closing parenthesis. Add `]`").
+
+### Path B: The Linter / Formatter (Pedant)
+*   **Trigger:** Tier 3 believes it finished a Ticket, but pre-commit hooks (e.g., `ruff`, `eslint`) fail.
+*   **What it Sees (Context):**
+    *   **Linter Warning:** Specific error (e.g., "Line too long", "Missing type hint").
+    *   **Target File:** Code written by Tier 3.
+*   **What it Ignores:** Business logic. It only cares about styling rules.
+*   **Output Format:** A direct `sed` command or silent diff overwrite via tools to fix the formatting without bothering Tier 2 or consuming Tier 3 loops.
+
+### Path C: The Flaky Test Debugger (Isolator)
+*   **Trigger:** A localized unit test fails due to logic (e.g., `assert 5 == 4`), not a syntax crash.
+*   **What it Sees (Context):**
+    *   **Failing Test Function:** The exact `pytest` or `go test` block.
+    *   **Target Function:** The specific function being tested.
+*   **What it Ignores:** The rest of the test suite and module.
+*   **Output Format:** A quick diagnosis sent to Tier 3 (e.g., "The test expects an integer, but your function is currently returning a stringified float. Cast to `int`").

MMA_Support/mma_tiered_orchestrator_skill.md

@@ -0,0 +1,66 @@
+# Skill: MMA Tiered Orchestrator
+
+## Description
+This skill enforces the 4-Tier Hierarchical Multi-Model Architecture (MMA) directly within the Gemini CLI using Token Firewalling and sub-agent task delegation. It teaches the CLI how to act as a Tier 1/2 Orchestrator, dispatching stateless tasks to cheaper models using shell commands, thereby preventing massive error traces or heavy coding contexts from polluting the primary prompt context.
+
+<instructions>
+# MMA Token Firewall & Tiered Delegation Protocol
+
+You are operating as a Tier 1 Product Manager or Tier 2 Tech Lead within the MMA Framework. Your context window is extremely valuable and must be protected from token bloat (such as raw, repetitive code edits, trial-and-error histories, or massive stack traces).
+
+To accomplish this, you MUST delegate token-heavy or stateless tasks to "Tier 3 Contributors" or "Tier 4 QA Agents" by spawning secondary Gemini CLI instances via `run_shell_command`.
+
+**CRITICAL Prerequisite:**
+To avoid hanging the CLI and ensure proper environment authentication, you MUST NOT call the `gemini` command directly. Instead, you MUST use the wrapper script:
+`.\scripts\run_subagent.ps1 -Prompt "..."`
+
+## 1. The Tier 3 Worker (Heads-Down Coding)
+When you need to perform a significant code modification (e.g., refactoring a 500-line script, writing a massive class, or implementing a predefined spec):
+1. **DO NOT** attempt to write or use `replace`/`write_file` yourself. Your history will bloat.
+2. **DO** construct a single, highly specific prompt.
+3. **DO** spawn a sub-agent using `run_shell_command` pointing to the target file.
+   *Command:* `.\scripts\run_subagent.ps1 -Prompt "Modify [FILE_PATH] to implement [SPECIFIC_INSTRUCTION]. Only write the code, no pleasantries."`
+4. If you need the sub-agent to automatically apply changes instead of just returning the text, use `gemini run` or pipe the output appropriately. However, the best method is to let the sub-agent modify the code and return "Done."
+
+## 2. The Tier 4 QA Agent (Error Translation)
+If you run a local test (e.g., `npm test`, `pytest`, `go run`) via `run_shell_command` and it fails with a massive traceback (e.g., 200+ lines of `stderr`):
+1. **DO NOT** analyze the raw `stderr` in your own context window.
+2. **DO** immediately spawn a stateless Tier 4 agent to compress the error.
+3. *Command:* `.\scripts\run_subagent.ps1 -Prompt "Summarize this stack trace into a 20-word fix: [PASTE_SNIPPET_OF_STDERR_HERE]"`
+4. Use the 20-word fix returned by the Tier 4 agent to inform your next architectural decision or pass it to the Tier 3 worker.
+
+## 3. Context Amnesia (Phase Checkpoints)
+When you complete a major Phase or Track within the `conductor` workflow:
+1. Stage your changes and commit them.
+2. Draft a comprehensive summary of the state changes in a Git Note attached to the commit.
+3. Treat the checkpoint as a "Memory Wipe." Actively disregard previous conversational turns and trial-and-error histories. Rely exclusively on the newly generated Git Note and the physical state of the files on disk for your next Phase.
+</instructions>
+
+<examples>
+### Example 1: Spawning a Tier 4 QA Agent
+**User / System:** `pytest tests/test_gui.py` failed with 400 lines of output.
+**Agent (You):** 
+```json
+{
+  "command": ".\\scripts\\run_subagent.ps1 -Prompt \"Summarize this stack trace into a 20-word fix: [snip first 30 lines...]\"",
+  "description": "Spawning Tier 4 QA to compress error trace statelessly."
+}
+```
+
+### Example 2: Spawning a Tier 3 Worker
+**User:** Please implement the `ASTParser` class in `file_cache.py` as defined in Track 1.
+**Agent (You):**
+```json
+{
+  "command": ".\\scripts\\run_subagent.ps1 -Prompt \"Read file_cache.py and implement the ASTParser class using tree-sitter. Ensure you preserve docstrings but strip function bodies. Output the updated code or edit the file directly.\"",
+  "description": "Delegating implementation to a Tier 3 Worker."
+}
+```
+</examples>
+
+<triggers>
+- When asked to write large amounts of boilerplate or repetitive code.
+- When encountering a large error trace from a shell execution.
+- When explicitly instructed to act as a "Tech Lead" or "Orchestrator".
+- When managing complex, multi-file Track implementations.
+</triggers>

MMA_UX_SPEC.md

@@ -0,0 +1,36 @@
+# MMA Observability & UX Specification
+
+## 1. Goal
+Implement the visible surface area of the 4-Tier Hierarchical Multi-Model Architecture within `gui_2.py`. This ensures the user can monitor, control, and debug the multi-agent execution flow.
+
+## 2. Core Components
+
+### 2.1 MMA Dashboard Panel
+- **Visibility:** A new dockable panel named "MMA Dashboard".
+- **Track Status:** Display the current active `Track` ID and overall progress (e.g., "3/10 Tickets Complete").
+- **Ticket DAG Visualization:** A list or simple graph representing the `Ticket` queue.
+    - Each ticket shows: `ID`, `Target`, `Status` (Pending, Running, Paused, Complete, Blocked).
+    - Visual indicators for dependencies (e.g., indented or linked).
+
+### 2.2 The Execution Clutch (HITL)
+- **Step Mode Toggle:** A global or per-track checkbox to enable "Step Mode".
+- **Pause Points:**
+    - **Pre-Execution:** When a Tier 3 worker generates a tool call (e.g., `write_file`), the engine pauses.
+    - **UI Interaction:** The GUI displays the proposed script/change and provides:
+        - `[Approve]`: Proceed with execution.
+        - `[Edit Payload]`: Open the Memory Mutator.
+        - `[Abort]`: Mark the ticket as Blocked/Cancelled.
+- **Visual Feedback:** Tactile/Arcade-style blinking or color changes when the engine is "Paused for HITL".
+
+### 2.3 Memory Mutator (The "Debug" Superpower)
+- **Functionality:** A modal or dedicated text area that allows the user to edit the raw JSON conversation history of a paused worker.
+- **Use Case:** Fixing AI hallucinations or providing specific guidance mid-turn without restarting the context window.
+- **Integration:** After editing, the "Approve" button sends the *modified* history back to the engine.
+
+### 2.4 Tiered Metrics & Logs
+- **Observability:** Show which model (Tier 1, 2, 3, or 4) is currently active.
+- **Sub-Agent Logs:** Provide quick links to open the timestamped log files generated by `mma_exec.py`.
+
+## 3. Technical Integration
+- **Event Bus:** Use the existing `AsyncEventQueue` to push `StateUpdateEvents` from the `ConductorEngine` to the GUI.
+- **Non-Blocking:** Ensure the UI remains responsive (FPS > 60) even when multiple tickets are processing or the engine is waiting for user input.

MainContext.md

@@ -12,7 +12,7 @@ Is a local GUI tool for manually curating and sending context to AI APIs. It agg
 - `uv` - package/env management
 
 **Files:**
-- `gui.py` - main GUI, `App` class, all panels, all callbacks, confirmation dialog, layout persistence, rich comms rendering; `[+ Maximize]` buttons in `ConfirmDialog` and `win_script_output` now pass text directly as `user_data` / read from `self._last_script` / `self._last_output` instance vars instead of `dpg.get_value(tag)` — fixes glitch when word-wrap is ON or dialog is dismissed before viewer opens
+- `gui_legacy.py` - main GUI, `App` class, all panels, all callbacks, confirmation dialog, layout persistence, rich comms rendering; `[+ Maximize]` buttons in `ConfirmDialog` and `win_script_output` now pass text directly as `user_data` / read from `self._last_script` / `self._last_output` instance vars instead of `dpg.get_value(tag)` — fixes glitch when word-wrap is ON or dialog is dismissed before viewer opens
 - `ai_client.py` - unified provider wrapper, model listing, session management, send, tool/function-call loop, comms log, provider error classification, token estimation, and aggressive history truncation
 - `aggregate.py` - reads config, collects files/screenshots/discussion, builds `file_items` with `mtime` for cache optimization, writes numbered `.md` files to `output_dir` using `build_markdown_from_items` to avoid double I/O; `run()` returns `(markdown_str, path, file_items)` tuple; `summary_only=False` by default (full file contents sent, not heuristic summaries)
 - `shell_runner.py` - subprocess wrapper that runs PowerShell scripts sandboxed to `base_dir`, returns stdout/stderr/exit code as a string
@@ -79,7 +79,7 @@ Is a local GUI tool for manually curating and sending context to AI APIs. It agg
 - Both Gemini and Anthropic are configured with a `run_powershell` tool/function declaration
 - When the AI wants to edit or create files it emits a tool call with a `script` string
 - `ai_client` runs a loop (max `MAX_TOOL_ROUNDS = 10`) feeding tool results back until the AI stops calling tools
-- Before any script runs, `gui.py` shows a modal `ConfirmDialog` on the main thread; the background send thread blocks on a `threading.Event` until the user clicks Approve or Reject
+- Before any script runs, `gui_legacy.py` shows a modal `ConfirmDialog` on the main thread; the background send thread blocks on a `threading.Event` until the user clicks Approve or Reject
 - The dialog displays `base_dir`, shows the script in an editable text box (allowing last-second tweaks), and has Approve & Run / Reject buttons
 - On approval the (possibly edited) script is passed to `shell_runner.run_powershell()` which prepends `Set-Location -LiteralPath '<base_dir>'` and runs it via `powershell -NoProfile -NonInteractive -Command`
 - stdout, stderr, and exit code are returned to the AI as the tool result
@@ -107,10 +107,10 @@ Is a local GUI tool for manually curating and sending context to AI APIs. It agg
 - Entry fields: `ts` (HH:MM:SS), `direction` (OUT/IN), `kind`, `provider`, `model`, `payload` (dict)
 - Anthropic responses also include `usage` (input_tokens, output_tokens, cache_creation_input_tokens, cache_read_input_tokens) and `stop_reason` in payload
 - `get_comms_log()` returns a snapshot; `clear_comms_log()` empties it
-- `comms_log_callback` (injected by gui.py) is called from the background thread with each new entry; gui queues entries in `_pending_comms` (lock-protected) and flushes them to the DPG panel each render frame
-- `COMMS_CLAMP_CHARS = 300` in gui.py governs the display cutoff for heavy text fields
+- `comms_log_callback` (injected by gui_legacy.py) is called from the background thread with each new entry; gui queues entries in `_pending_comms` (lock-protected) and flushes them to the DPG panel each render frame
+- `COMMS_CLAMP_CHARS = 300` in gui_legacy.py governs the display cutoff for heavy text fields
 
-**Comms History panel — rich structured rendering (gui.py):**
+**Comms History panel — rich structured rendering (gui_legacy.py):**
 
 Rather than showing raw JSON, each comms entry is rendered using a kind-specific renderer function. Unknown kinds fall back to a generic key/value layout.
 
@@ -195,10 +195,10 @@ Entry layout: index + timestamp + direction + kind + provider/model header row,
 - Comms log: MCP tool calls log `OUT/tool_call` with `{"name": ..., "args": {...}}` and `IN/tool_result` with `{"name": ..., "output": ...}`; rendered in the Comms History panel via `_render_payload_tool_call` (shows each arg key/value) and `_render_payload_tool_result` (shows output)
 
 **Known extension points:**
-- Add more providers by adding a section to `credentials.toml`, a `_list_*` and `_send_*` function in `ai_client.py`, and the provider name to the `PROVIDERS` list in `gui.py`
+- Add more providers by adding a section to `credentials.toml`, a `_list_*` and `_send_*` function in `ai_client.py`, and the provider name to the `PROVIDERS` list in `gui_legacy.py`
 - Discussion history excerpts could be individually toggleable for inclusion in the generated md
 - `MAX_TOOL_ROUNDS` in `ai_client.py` caps agentic loops at 10 rounds; adjustable
-- `COMMS_CLAMP_CHARS` in `gui.py` controls the character threshold for clamping heavy payload fields in the Comms History panel
+- `COMMS_CLAMP_CHARS` in gui_legacy.py controls the character threshold for clamping heavy payload fields in the Comms History panel
 - Additional project metadata (description, tags, created date) could be added to `[project]` in the per-project toml
 
 ### Gemini Context Management
@@ -222,7 +222,7 @@ Entry layout: index + timestamp + direction + kind + provider/model header row,
 
 
 ## Recent Changes (Text Viewer Maximization)
-- **Global Text Viewer (gui.py)**: Added a dedicated, large popup window (win_text_viewer) to allow reading and scrolling through large, dense text blocks without feeling cramped.
+- **Global Text Viewer (gui_legacy.py)**: Added a dedicated, large popup window (win_text_viewer) to allow reading and scrolling through large, dense text blocks without feeling cramped.
 - **Comms History**: Every multi-line text field in the comms log now has a [+] button next to its label that opens the text in the Global Text Viewer.
 - **Tool Log History**: Added [+ Script] and [+ Output] buttons next to each logged tool call to easily maximize and read the full executed scripts and raw tool outputs.
 - **Last Script Output Popup**: Expanded the default size of the popup (now 800x600) and gave the input script panel more vertical space to prevent it from feeling 'scrunched'. Added [+ Maximize] buttons for both the script and the output sections to inspect them in full detail.
@@ -266,10 +266,10 @@ Documentation has been completely rewritten matching the strict, structural form
 ### aggregate.py — run() double-I/O elimination
 - `run()` now calls `build_file_items()` once, then passes the result to `build_markdown_from_items()` instead of calling `build_files_section()` separately. This avoids reading every file twice per send.
 - `build_markdown_from_items()` accepts a `summary_only` flag (default `False`); when `False` it inlines full file content; when `True` it delegates to `summarize.build_summary_markdown()` for compact structural summaries.
-- `run()` returns a 3-tuple `(markdown_str, output_path, file_items)` — the `file_items` list is passed through to `gui.py` as `self.last_file_items` for dynamic context refresh after tool calls.
+- `run()` returns a 3-tuple `(markdown_str, output_path, file_items)` — the `file_items` list is passed through to `gui_legacy.py` as `self.last_file_items` for dynamic context refresh after tool calls.
 
 
-## Updates (2026-02-22 — gui.py [+ Maximize] bug fix)
+## Updates (2026-02-22 — gui_legacy.py [+ Maximize] bug fix)
 
 ### Problem
 Three `[+ Maximize]` buttons were reading their text content via `dpg.get_value(tag)` at click time:

Readme.md

@@ -21,6 +21,15 @@ Features:
   * Popup text viewers for large script/output inspection.
   * Color theming and UI scaling.
 
+## Session-Based Logging and Management
+
+Manual Slop organizes all communications and tool interactions into session-based directories under `logs/`. This ensures a clean history and easy debugging.
+
+*   **Organized Storage:** Each session is assigned a unique ID and its own sub-directory containing communication logs (`comms.log`) and metadata.
+*   **Log Management Panel:** The GUI includes a dedicated 'Log Management' panel where you can view session history, inspect metadata (message counts, errors, size), and protect important sessions.
+*   **Automated Pruning:** To keep the workspace clean, the application automatically prunes insignificant logs. Sessions older than 24 hours that are not "whitelisted" and are smaller than 2KB are automatically deleted.
+*   **Whitelisting:** Sessions containing errors, high activity, or significant changes are automatically whitelisted. Users can also manually whitelist sessions via the GUI to prevent them from being pruned.
+
 ## Documentation
 
 * [docs/Readme.md](docs/Readme.md) for the interface and usage guide
@@ -41,5 +50,5 @@ api_key = "****"
 2. Have fun. This is experiemntal slop.
 
 ```ps1
-uv run .\gui.py
+uv run .\gui_2.py
 ```

aggregate.py

@@ -16,6 +16,7 @@ import re
 import glob
 from pathlib import Path, PureWindowsPath
 import summarize
+import project_manager
 
 def find_next_increment(output_dir: Path, namespace: str) -> int:
     pattern = re.compile(rf"^{re.escape(namespace)}_(\d+)\.md$")
@@ -37,14 +38,24 @@ def is_absolute_with_drive(entry: str) -> bool:
 def resolve_paths(base_dir: Path, entry: str) -> list[Path]:
     has_drive = is_absolute_with_drive(entry)
     is_wildcard = "*" in entry
+    
+    matches = []
     if is_wildcard:
         root = Path(entry) if has_drive else base_dir / entry
         matches = [Path(p) for p in glob.glob(str(root), recursive=True) if Path(p).is_file()]
-        return sorted(matches)
     else:
-        if has_drive:
-            return [Path(entry)]
-        return [(base_dir / entry).resolve()]
+        p = Path(entry) if has_drive else (base_dir / entry).resolve()
+        matches = [p]
+        
+    # Blacklist filter
+    filtered = []
+    for p in matches:
+        name = p.name.lower()
+        if name == "history.toml" or name.endswith("_history.toml"):
+            continue
+        filtered.append(p)
+        
+    return sorted(filtered)
 
 def build_discussion_section(history: list[str]) -> str:
     sections = []
@@ -214,9 +225,25 @@ def run(config: dict) -> tuple[str, Path, list[dict]]:
     return markdown, output_file, file_items
 
 def main():
-    with open("config.toml", "rb") as f:
-        import tomllib
-        config = tomllib.load(f)
+    # Load global config to find active project
+    config_path = Path("config.toml")
+    if not config_path.exists():
+        print("config.toml not found.")
+        return
+        
+    with open(config_path, "rb") as f:
+        global_cfg = tomllib.load(f)
+        
+    active_path = global_cfg.get("projects", {}).get("active")
+    if not active_path:
+        print("No active project found in config.toml.")
+        return
+        
+    # Use project_manager to load project (handles history segregation)
+    proj = project_manager.load_project(active_path)
+    # Use flat_config to make it compatible with aggregate.run()
+    config = project_manager.flat_config(proj)
+    
     markdown, output_file, _ = run(config)
     print(f"Written: {output_file}")
 

ai_client.py

@@ -13,22 +13,27 @@ during chat creation to avoid massive history bloat.
 # ai_client.py
 import tomllib
 import json
+import sys
 import time
 import datetime
 import hashlib
 import difflib
 import threading
+import requests
 from pathlib import Path
+from typing import Optional, Callable
 import os
+import project_manager
 import file_cache
 import mcp_client
 import anthropic
+from gemini_cli_adapter import GeminiCliAdapter
 from google import genai
 from google.genai import types
 from events import EventEmitter
 
 _provider: str = "gemini"
-_model: str = "gemini-2.5-flash"
+_model: str = "gemini-2.5-flash-lite"
 _temperature: float = 0.0
 _max_tokens: int = 8192
 
@@ -43,6 +48,13 @@ def set_model_params(temp: float, max_tok: int, trunc_limit: int = 8000):
     _max_tokens = max_tok
     _history_trunc_limit = trunc_limit
 
+def get_history_trunc_limit() -> int:
+    return _history_trunc_limit
+
+def set_history_trunc_limit(val: int):
+    global _history_trunc_limit
+    _history_trunc_limit = val
+
 _gemini_client = None
 _gemini_chat = None
 _gemini_cache = None
@@ -56,8 +68,15 @@ _GEMINI_CACHE_TTL = 3600
 _anthropic_client = None
 _anthropic_history: list[dict] = []
 _anthropic_history_lock = threading.Lock()
+
+_deepseek_client = None
+_deepseek_history: list[dict] = []
+_deepseek_history_lock = threading.Lock()
+
 _send_lock = threading.Lock()
 
+_gemini_cli_adapter = None
+
 # Injected by gui.py - called when AI wants to run a command.
 # Signature: (script: str, base_dir: str) -> str | None
 confirm_and_run_callback = None
@@ -83,6 +102,7 @@ _ANTHROPIC_CHUNK_SIZE = 120_000
 
 _SYSTEM_PROMPT = (
     "You are a helpful coding assistant with access to a PowerShell tool and MCP tools (file access: read_file, list_directory, search_files, get_file_summary, web access: web_search, fetch_url). "
+    "When calling file/directory tools, always use the 'path' parameter for the target path. "
     "When asked to create or edit files, prefer targeted edits over full rewrites. "
     "Always explain what you are doing before invoking the tool.\n\n"
     "When writing or rewriting large files (especially those containing quotes, backticks, or special characters), "
@@ -148,10 +168,10 @@ def _load_credentials() -> dict:
             f"Create a credentials.toml with:\n"
             f"  [gemini]\n  api_key = \"your-key\"\n"
             f"  [anthropic]\n  api_key = \"your-key\"\n"
+            f"  [deepseek]\n  api_key = \"your-key\"\n"
             f"Or set SLOP_CREDENTIALS env var to a custom path."
         )
 
-
 # ------------------------------------------------------------------ provider errors
 
 class ProviderError(Exception):
@@ -230,12 +250,36 @@ def _classify_gemini_error(exc: Exception) -> ProviderError:
     return ProviderError("unknown", "gemini", exc)
 
 
+def _classify_deepseek_error(exc: Exception) -> ProviderError:
+    body = str(exc).lower()
+    if "429" in body or "rate" in body:
+        return ProviderError("rate_limit", "deepseek", exc)
+    if "401" in body or "403" in body or "auth" in body or "api key" in body:
+        return ProviderError("auth", "deepseek", exc)
+    if "402" in body or "balance" in body or "billing" in body:
+        return ProviderError("balance", "deepseek", exc)
+    if "quota" in body or "limit exceeded" in body:
+        return ProviderError("quota", "deepseek", exc)
+    if "connection" in body or "timeout" in body or "network" in body:
+        return ProviderError("network", "deepseek", exc)
+    return ProviderError("unknown", "deepseek", exc)
+
+
 # ------------------------------------------------------------------ provider setup
 
 def set_provider(provider: str, model: str):
     global _provider, _model
     _provider = provider
-    _model = model
+    
+    if provider == "gemini_cli":
+        valid_models = _list_gemini_cli_models()
+        # If model is invalid or belongs to another provider (like deepseek), force default
+        if model not in valid_models or model.startswith("deepseek"):
+            _model = "gemini-3-flash-preview"
+        else:
+            _model = model
+    else:
+        _model = model
 
 
 
@@ -253,6 +297,7 @@ def reset_session():
     global _gemini_cache_md_hash, _gemini_cache_created_at
     global _anthropic_client, _anthropic_history
     global _CACHED_ANTHROPIC_TOOLS
+    global _gemini_cli_adapter
     if _gemini_client and _gemini_cache:
         try:
             _gemini_client.caches.delete(name=_gemini_cache.name)
@@ -263,9 +308,17 @@ def reset_session():
     _gemini_cache = None
     _gemini_cache_md_hash = None
     _gemini_cache_created_at = None
+    if _gemini_cli_adapter:
+        _gemini_cli_adapter.session_id = None
+    _gemini_cli_adapter = None
     _anthropic_client = None
     with _anthropic_history_lock:
         _anthropic_history = []
+
+    _deepseek_client = None
+    with _deepseek_history_lock:
+        _deepseek_history = []
+
     _CACHED_ANTHROPIC_TOOLS = None
     file_cache.reset_client()
 
@@ -294,9 +347,28 @@ def list_models(provider: str) -> list[str]:
         return _list_gemini_models(creds["gemini"]["api_key"])
     elif provider == "anthropic":
         return _list_anthropic_models()
+    elif provider == "deepseek":
+        return _list_deepseek_models(creds["deepseek"]["api_key"])
+    elif provider == "gemini_cli":
+        return _list_gemini_cli_models()
     return []
 
 
+def _list_gemini_cli_models() -> list[str]:
+    """
+    List available Gemini models for the CLI.
+    Since the CLI doesn't have a direct 'list models' command yet,
+    we return a curated list of supported models based on CLI metadata.
+    """
+    return [
+        "gemini-3-flash-preview",
+        "gemini-3.1-pro-preview",
+        "gemini-2.5-pro",
+        "gemini-2.5-flash",
+        "gemini-2.5-flash-lite",
+    ]
+
+
 def _list_gemini_models(api_key: str) -> list[str]:
 
     try:
@@ -326,6 +398,14 @@ def _list_anthropic_models() -> list[str]:
         raise _classify_anthropic_error(exc) from exc
 
 
+def _list_deepseek_models(api_key: str) -> list[str]:
+    """
+    List available DeepSeek models.
+    """
+    # For now, return the models specified in the requirements
+    return ["deepseek-chat", "deepseek-reasoner", "deepseek-v3", "deepseek-r1"]
+
+
 # ------------------------------------------------------------------ tool definition
 
 TOOL_NAME = "run_powershell"
@@ -443,10 +523,10 @@ def _gemini_tool_declaration():
     return types.Tool(function_declarations=declarations) if declarations else None
 
 
-def _run_script(script: str, base_dir: str) -> str:
+def _run_script(script: str, base_dir: str, qa_callback: Optional[Callable[[str], str]] = None) -> str:
     if confirm_and_run_callback is None:
         return "ERROR: no confirmation handler registered"
-    result = confirm_and_run_callback(script, base_dir)
+    result = confirm_and_run_callback(script, base_dir, qa_callback)
     if result is None:
         output = "USER REJECTED: command was not executed"
     else:
@@ -589,7 +669,9 @@ def _get_gemini_history_list(chat):
 
 def _send_gemini(md_content: str, user_message: str, base_dir: str,
                  file_items: list[dict] | None = None,
-                 discussion_history: str = "") -> str:
+                 discussion_history: str = "",
+                 pre_tool_callback = None,
+                 qa_callback: Optional[Callable[[str], str]] = None) -> str:
     global _gemini_chat, _gemini_cache, _gemini_cache_md_hash, _gemini_cache_created_at
 
     try:
@@ -597,7 +679,8 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str,
         # Only stable content (files + screenshots) goes in the cached system instruction.
         # Discussion history is sent as conversation messages so the cache isn't invalidated every turn.
         sys_instr = f"{_get_combined_system_prompt()}\n\n<context>\n{md_content}\n</context>"
-        tools_decl = [_gemini_tool_declaration()]
+        td = _gemini_tool_declaration()
+        tools_decl = [td] if td else None
 
         # DYNAMIC CONTEXT: Check if files/context changed mid-session
         current_md_hash = hashlib.md5(md_content.encode()).hexdigest()
@@ -617,7 +700,7 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str,
         if _gemini_chat and _gemini_cache and _gemini_cache_created_at:
             elapsed = time.time() - _gemini_cache_created_at
             if elapsed > _GEMINI_CACHE_TTL * 0.9:
-                old_history = list(_get_gemini_history_list(_gemini_chat)) if _get_gemini_history_list(_gemini_chat) else []
+                old_history = list(_get_gemini_history_list(_gemini_chat)) if _get_gemini_history_list(_get_gemini_history_list(_gemini_chat)) else []
                 try: _gemini_client.caches.delete(name=_gemini_cache.name)
                 except Exception as e: _append_comms("OUT", "request", {"message": f"[CACHE DELETE WARN] {e}"})
                 _gemini_chat = None
@@ -633,28 +716,42 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str,
                 max_output_tokens=_max_tokens,
                 safety_settings=[types.SafetySetting(category="HARM_CATEGORY_DANGEROUS_CONTENT", threshold="BLOCK_ONLY_HIGH")]
             )
+            
+            # Check if context is large enough to warrant caching (min 2048 tokens usually)
+            should_cache = False
             try:
-                # Gemini requires 1024 (Flash) or 4096 (Pro) tokens to cache.
-                _gemini_cache = _gemini_client.caches.create(
-                    model=_model,
-                    config=types.CreateCachedContentConfig(
-                        system_instruction=sys_instr,
-                        tools=tools_decl,
-                        ttl=f"{_GEMINI_CACHE_TTL}s",
-                    )
-                )
-                _gemini_cache_created_at = time.time()
-                chat_config = types.GenerateContentConfig(
-                    cached_content=_gemini_cache.name,
-                    temperature=_temperature,
-                    max_output_tokens=_max_tokens,
-                    safety_settings=[types.SafetySetting(category="HARM_CATEGORY_DANGEROUS_CONTENT", threshold="BLOCK_ONLY_HIGH")]
-                )
-                _append_comms("OUT", "request", {"message": f"[CACHE CREATED] {_gemini_cache.name}"})
+                count_resp = _gemini_client.models.count_tokens(model=_model, contents=[sys_instr])
+                # We use a 2048 threshold to be safe across models
+                if count_resp.total_tokens >= 2048:
+                    should_cache = True
+                else:
+                    _append_comms("OUT", "request", {"message": f"[CACHING SKIPPED] Context too small ({count_resp.total_tokens} tokens < 2048)"})
             except Exception as e:
-                _gemini_cache = None
-                _gemini_cache_created_at = None
-                _append_comms("OUT", "request", {"message": f"[CACHE FAILED] {type(e).__name__}: {e} — falling back to inline system_instruction"})
+                _append_comms("OUT", "request", {"message": f"[COUNT FAILED] {e}"})
+
+            if should_cache:
+                try:
+                    # Gemini requires 1024 (Flash) or 4096 (Pro) tokens to cache.
+                    _gemini_cache = _gemini_client.caches.create(
+                        model=_model,
+                        config=types.CreateCachedContentConfig(
+                            system_instruction=sys_instr,
+                            tools=tools_decl,
+                            ttl=f"{_GEMINI_CACHE_TTL}s",
+                        )
+                    )
+                    _gemini_cache_created_at = time.time()
+                    chat_config = types.GenerateContentConfig(
+                        cached_content=_gemini_cache.name,
+                        temperature=_temperature,
+                        max_output_tokens=_max_tokens,
+                        safety_settings=[types.SafetySetting(category="HARM_CATEGORY_DANGEROUS_CONTENT", threshold="BLOCK_ONLY_HIGH")]
+                    )
+                    _append_comms("OUT", "request", {"message": f"[CACHE CREATED] {_gemini_cache.name}"})
+                except Exception as e:
+                    _gemini_cache = None
+                    _gemini_cache_created_at = None
+                    _append_comms("OUT", "request", {"message": f"[CACHE FAILED] {type(e).__name__}: {e} \u2014 falling back to inline system_instruction"})
 
             kwargs = {"model": _model, "config": chat_config}
             if old_history:
@@ -674,7 +771,7 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str,
         _cumulative_tool_bytes = 0
 
         # Strip stale file refreshes and truncate old tool outputs ONCE before
-        # entering the tool loop (not per-round — history entries don't change).
+        # entering the tool loop (not per-round \u2014 history entries don't change).
         if _gemini_chat and _get_gemini_history_list(_gemini_chat):
             for msg in _get_gemini_history_list(_gemini_chat):
                 if msg.role == "user" and hasattr(msg, "parts"):
@@ -737,6 +834,16 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str,
             f_resps, log = [], []
             for i, fc in enumerate(calls):
                 name, args = fc.name, dict(fc.args)
+                
+                # Check for tool confirmation if callback is provided
+                if pre_tool_callback:
+                    payload_str = json.dumps({"tool": name, "args": args})
+                    if not pre_tool_callback(payload_str):
+                        out = "USER REJECTED: tool execution cancelled"
+                        f_resps.append(types.Part.from_function_response(name=name, response={"output": out}))
+                        log.append({"tool_use_id": name, "content": out})
+                        continue
+
                 events.emit("tool_execution", payload={"status": "started", "tool": name, "args": args, "round": r_idx})
                 if name in mcp_client.TOOL_NAMES:
                     _append_comms("OUT", "tool_call", {"name": name, "args": args})
@@ -744,7 +851,7 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str,
                 elif name == TOOL_NAME:
                     scr = args.get("script", "")
                     _append_comms("OUT", "tool_call", {"name": TOOL_NAME, "script": scr})
-                    out = _run_script(scr, base_dir)
+                    out = _run_script(scr, base_dir, qa_callback)
                 else: out = f"ERROR: unknown tool '{name}'"
 
                 if i == len(calls) - 1:
@@ -773,8 +880,158 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str,
         return "\n\n".join(all_text) if all_text else "(No text returned)"
     except Exception as e: raise _classify_gemini_error(e) from e
 
+def _send_gemini_cli(md_content: str, user_message: str, base_dir: str,
+                      file_items: list[dict] | None = None,
+                      discussion_history: str = "",
+                      pre_tool_callback = None,
+                      qa_callback: Optional[Callable[[str], str]] = None) -> str:
+    global _gemini_cli_adapter
+    try:
+        if _gemini_cli_adapter is None:
+            _gemini_cli_adapter = GeminiCliAdapter(binary_path="gemini")
         
+        adapter = _gemini_cli_adapter
 
+        mcp_client.configure(file_items or [], [base_dir])
+
+        # Construct the system instruction, combining the base system prompt and the current context.
+        sys_instr = f"{_get_combined_system_prompt()}\n\n<context>\n{md_content}\n</context>"
+        safety_settings = [{'category': 'HARM_CATEGORY_DANGEROUS_CONTENT', 'threshold': 'BLOCK_ONLY_HIGH'}]
+
+        # Initial payload for the first message
+        payload = user_message
+        if adapter.session_id is None:
+            if discussion_history:
+                payload = f"[DISCUSSION HISTORY]\n\n{discussion_history}\n\n---\n\n{user_message}"
+
+        all_text = []
+        _cumulative_tool_bytes = 0
+
+        for r_idx in range(MAX_TOOL_ROUNDS + 2):
+            if adapter is None:
+                break
+
+            events.emit("request_start", payload={"provider": "gemini_cli", "model": _model, "round": r_idx})
+            _append_comms("OUT", "request", {"message": f"[CLI] [round {r_idx}] [msg {len(payload)}]"})
+
+            resp_data = adapter.send(payload, safety_settings=safety_settings, system_instruction=sys_instr, model=_model)
+            
+            # Log any stderr from the CLI for transparency
+            cli_stderr = resp_data.get("stderr", "")
+            if cli_stderr:
+                sys.stderr.write(f"\n--- Gemini CLI stderr ---\n{cli_stderr}\n-------------------------\n")
+                sys.stderr.flush()
+
+            txt = resp_data.get("text", "")
+            if txt: all_text.append(txt)
+
+            calls = resp_data.get("tool_calls", [])
+            usage = adapter.last_usage or {}
+            latency = adapter.last_latency
+
+            events.emit("response_received", payload={"provider": "gemini_cli", "model": _model, "usage": usage, "latency": latency, "round": r_idx})
+
+            # Clean up the tool calls format to match comms log expectation
+            log_calls = []
+            for c in calls:
+                log_calls.append({"name": c.get("name"), "args": c.get("args"), "id": c.get("id")})
+
+            _append_comms("IN", "response", {
+                "round": r_idx,
+                "stop_reason": "TOOL_USE" if calls else "STOP",
+                "text": txt,
+                "tool_calls": log_calls,
+                "usage": usage
+            })
+
+            # If there's text and we're not done, push it to the history immediately
+            # so it appears as a separate entry in the GUI.
+            if txt and calls and comms_log_callback:
+                # Use kind='history_add' to push a new entry into the disc_entries list
+                comms_log_callback({
+                    "ts": project_manager.now_ts(),
+                    "direction": "IN",
+                    "kind": "history_add",
+                    "payload": {
+                        "role": "AI",
+                        "content": txt
+                    }
+                })
+
+            if not calls or r_idx > MAX_TOOL_ROUNDS:
+                break
+
+            tool_results_for_cli = []
+            for i, fc in enumerate(calls):
+                name = fc.get("name")
+                args = fc.get("args", {})
+                call_id = fc.get("id")
+                
+                # Check for tool confirmation if callback is provided
+                if pre_tool_callback:
+                    payload_str = json.dumps({"tool": name, "args": args})
+                    if not pre_tool_callback(payload_str):
+                        out = "USER REJECTED: tool execution cancelled"
+                        tool_results_for_cli.append({
+                            "role": "tool",
+                            "tool_call_id": call_id,
+                            "name": name,
+                            "content": out
+                        })
+                        _append_comms("IN", "tool_result", {"name": name, "id": call_id, "output": out})
+                        continue
+
+                events.emit("tool_execution", payload={"status": "started", "tool": name, "args": args, "round": r_idx})
+                if name in mcp_client.TOOL_NAMES:
+                    _append_comms("OUT", "tool_call", {"name": name, "id": call_id, "args": args})
+                    out = mcp_client.dispatch(name, args)
+                elif name == TOOL_NAME:
+                    scr = args.get("script", "")
+                    _append_comms("OUT", "tool_call", {"name": TOOL_NAME, "id": call_id, "script": scr})
+                    out = _run_script(scr, base_dir, qa_callback)
+                else:
+                    out = f"ERROR: unknown tool '{name}'"
+
+                if i == len(calls) - 1:
+                    if file_items:
+                        file_items, changed = _reread_file_items(file_items)
+                        ctx = _build_file_diff_text(changed)
+                        if ctx:
+                            out += f"\n\n[SYSTEM: FILES UPDATED]\n\n{ctx}"
+                    if r_idx == MAX_TOOL_ROUNDS:
+                        out += "\n\n[SYSTEM: MAX ROUNDS. PROVIDE FINAL ANSWER.]"
+
+                out = _truncate_tool_output(out)
+                _cumulative_tool_bytes += len(out)
+                
+                tool_results_for_cli.append({
+                    "role": "tool",
+                    "tool_call_id": call_id,
+                    "name": name,
+                    "content": out
+                })
+                
+                _append_comms("IN", "tool_result", {"name": name, "id": call_id, "output": out})
+                events.emit("tool_execution", payload={"status": "completed", "tool": name, "result": out, "round": r_idx})
+            
+            # CRITICAL: Update payload for the next round
+            payload = json.dumps(tool_results_for_cli)
+
+            if _cumulative_tool_bytes > _MAX_TOOL_OUTPUT_BYTES:
+                 _append_comms("OUT", "request", {"message": f"[TOOL OUTPUT BUDGET EXCEEDED: {_cumulative_tool_bytes} bytes]"})
+                 # We should ideally tell the model here, but for CLI we just append to payload
+
+            # For Gemini CLI, we send the tool results as a JSON array of messages (or similar)
+            # The adapter expects a string, so we'll pass the JSON string of the results.
+            payload = json.dumps(tool_results_for_cli)
+
+        # Return only the text from the last round, because intermediate 
+        # text chunks were already pushed to history via comms_log_callback.
+        final_text = all_text[-1] if all_text else "(No text returned)"
+        return final_text
+    except Exception as e:
+        # Basic error classification for CLI
+        raise ProviderError("unknown", "gemini_cli", e)
 
 # ------------------------------------------------------------------ anthropic history management
 
@@ -1024,13 +1281,13 @@ def _repair_anthropic_history(history: list[dict]):
     })
 
 
-def _send_anthropic(md_content: str, user_message: str, base_dir: str, file_items: list[dict] | None = None, discussion_history: str = "") -> str:
+def _send_anthropic(md_content: str, user_message: str, base_dir: str, file_items: list[dict] | None = None, discussion_history: str = "", pre_tool_callback = None, qa_callback: Optional[Callable[[str], str]] = None) -> str:
     try:
         _ensure_anthropic_client()
         mcp_client.configure(file_items or [], [base_dir])
 
         # Split system into two cache breakpoints:
-        # 1. Stable system prompt (never changes — always a cache hit)
+        # 1. Stable system prompt (never changes \u2014 always a cache hit)
         # 2. Dynamic file context (invalidated only when files change)
         stable_prompt = _get_combined_system_prompt()
         stable_blocks = [{"type": "text", "text": stable_prompt, "cache_control": {"type": "ephemeral"}}]
@@ -1155,6 +1412,19 @@ def _send_anthropic(md_content: str, user_message: str, base_dir: str, file_item
                 b_name = getattr(block, "name", None)
                 b_id   = getattr(block, "id",   "")
                 b_input = getattr(block, "input", {})
+
+                # Check for tool confirmation if callback is provided
+                if pre_tool_callback:
+                    payload_str = json.dumps({"tool": b_name, "args": b_input})
+                    if not pre_tool_callback(payload_str):
+                        output = "USER REJECTED: tool execution cancelled"
+                        tool_results.append({
+                            "type":        "tool_result",
+                            "tool_use_id": b_id,
+                            "content":     output,
+                        })
+                        continue
+
                 events.emit("tool_execution", payload={"status": "started", "tool": b_name, "args": b_input, "round": round_idx})
                 if b_name in mcp_client.TOOL_NAMES:
                     _append_comms("OUT", "tool_call", {"name": b_name, "id": b_id, "args": b_input})
@@ -1175,7 +1445,7 @@ def _send_anthropic(md_content: str, user_message: str, base_dir: str, file_item
                         "id":     b_id,
                         "script": script,
                     })
-                    output = _run_script(script, base_dir)
+                    output = _run_script(script, base_dir, qa_callback)
                     _append_comms("IN", "tool_result", {
                         "name":   TOOL_NAME,
                         "id":     b_id,
@@ -1238,6 +1508,304 @@ def _send_anthropic(md_content: str, user_message: str, base_dir: str, file_item
         raise _classify_anthropic_error(exc) from exc
 
 
+# ------------------------------------------------------------------ deepseek
+
+def _ensure_deepseek_client():
+    global _deepseek_client
+    if _deepseek_client is None:
+        creds = _load_credentials()
+        # Placeholder for Dedicated DeepSeek SDK instantiation
+        # import deepseek
+        # _deepseek_client = deepseek.DeepSeek(api_key=creds["deepseek"]["api_key"])
+        pass
+
+
+def _send_deepseek(md_content: str, user_message: str, base_dir: str,
+                   file_items: list[dict] | None = None,
+                   discussion_history: str = "",
+                   stream: bool = False,
+                   pre_tool_callback = None,
+                   qa_callback: Optional[Callable[[str], str]] = None) -> str:
+    """
+    Sends a message to the DeepSeek API, handling tool calls and history.
+    Supports streaming responses.
+    """
+    try:
+        mcp_client.configure(file_items or [], [base_dir])
+        creds = _load_credentials()
+        api_key = creds.get("deepseek", {}).get("api_key")
+        if not api_key:
+            raise ValueError("DeepSeek API key not found in credentials.toml")
+
+        # DeepSeek API details
+        api_url = "https://api.deepseek.com/chat/completions"
+        headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+        }
+
+        # Build the messages for the current API call
+        current_api_messages = []
+        with _deepseek_history_lock:
+            for msg in _deepseek_history:
+                current_api_messages.append(msg)
+
+        # Add the current user's input for this turn
+        initial_user_message_content = user_message
+        if discussion_history:
+            initial_user_message_content = f"[DISCUSSION HISTORY]\n\n{discussion_history}\n\n---\n\n{user_message}"
+        current_api_messages.append({"role": "user", "content": initial_user_message_content})
+
+        # Construct the full request payload
+        request_payload = {
+            "model": _model,
+            "messages": current_api_messages,
+            "temperature": _temperature,
+            "max_tokens": _max_tokens,
+            "stream": stream,
+        }
+
+        # Insert system prompt at the beginning
+        sys_msg = {"role": "system", "content": f"{_get_combined_system_prompt()}\n\n<context>\n{md_content}\n</context>"}
+        request_payload["messages"].insert(0, sys_msg)
+
+        all_text_parts = []
+        _cumulative_tool_bytes = 0
+        round_idx = 0
+
+        while round_idx <= MAX_TOOL_ROUNDS + 1:
+            events.emit("request_start", payload={"provider": "deepseek", "model": _model, "round": round_idx, "streaming": stream})
+
+            try:
+                response = requests.post(api_url, headers=headers, json=request_payload, timeout=60, stream=stream)
+                response.raise_for_status()
+            except requests.exceptions.RequestException as e:
+                raise _classify_deepseek_error(e) from e
+
+            # Process response
+            if stream:
+                aggregated_content = ""
+                aggregated_tool_calls = []
+                aggregated_reasoning = ""
+                current_usage = {}
+                final_finish_reason = "stop"
+
+                for line in response.iter_lines():
+                    if not line:
+                        continue
+                    decoded = line.decode('utf-8')
+                    if decoded.startswith('data: '):
+                        chunk_str = decoded[len('data: '):]
+                        if chunk_str.strip() == '[DONE]':
+                            continue
+                        try:
+                            chunk = json.loads(chunk_str)
+                            delta = chunk.get("choices", [{}])[0].get("delta", {})
+                            
+                            if delta.get("content"):
+                                aggregated_content += delta["content"]
+                            
+                            if delta.get("reasoning_content"):
+                                aggregated_reasoning += delta["reasoning_content"]
+                            
+                            if delta.get("tool_calls"):
+                                # Simple aggregation of tool call deltas
+                                for tc_delta in delta["tool_calls"]:
+                                    idx = tc_delta.get("index", 0)
+                                    while len(aggregated_tool_calls) <= idx:
+                                        aggregated_tool_calls.append({"id": "", "type": "function", "function": {"name": "", "arguments": ""}})
+                                    
+                                    target = aggregated_tool_calls[idx]
+                                    if tc_delta.get("id"):
+                                        target["id"] = tc_delta["id"]
+                                    if tc_delta.get("function", {}).get("name"):
+                                        target["function"]["name"] += tc_delta["function"]["name"]
+                                    if tc_delta.get("function", {}).get("arguments"):
+                                        target["function"]["arguments"] += tc_delta["function"]["arguments"]
+
+                            if chunk.get("choices", [{}])[0].get("finish_reason"):
+                                final_finish_reason = chunk["choices"][0]["finish_reason"]
+                            
+                            if chunk.get("usage"):
+                                current_usage = chunk["usage"]
+                        except json.JSONDecodeError:
+                            continue
+
+                assistant_text = aggregated_content
+                tool_calls_raw = aggregated_tool_calls
+                reasoning_content = aggregated_reasoning
+                finish_reason = final_finish_reason
+                usage = current_usage
+            else:
+                response_data = response.json()
+                choices = response_data.get("choices", [])
+                if not choices:
+                    _append_comms("IN", "response", {"round": round_idx, "text": "(No choices returned)", "usage": response_data.get("usage", {})})
+                    break
+                
+                choice = choices[0]
+                message = choice.get("message", {})
+                assistant_text = message.get("content", "")
+                tool_calls_raw = message.get("tool_calls", [])
+                reasoning_content = message.get("reasoning_content", "")
+                finish_reason = choice.get("finish_reason", "stop")
+                usage = response_data.get("usage", {})
+
+            # Format reasoning content if it exists
+            thinking_tags = ""
+            if reasoning_content:
+                thinking_tags = f"<thinking>\n{reasoning_content}\n</thinking>\n"
+
+            full_assistant_text = thinking_tags + assistant_text
+
+            # Update history
+            with _deepseek_history_lock:
+                msg_to_store = {"role": "assistant", "content": assistant_text}
+                if reasoning_content:
+                    msg_to_store["reasoning_content"] = reasoning_content
+                if tool_calls_raw:
+                    msg_to_store["tool_calls"] = tool_calls_raw
+                _deepseek_history.append(msg_to_store)
+
+            if full_assistant_text:
+                all_text_parts.append(full_assistant_text)
+
+            _append_comms("IN", "response", {
+                "round": round_idx,
+                "stop_reason": finish_reason,
+                "text": full_assistant_text,
+                "tool_calls": tool_calls_raw,
+                "usage": usage,
+                "streaming": stream
+            })
+
+            if finish_reason != "tool_calls" and not tool_calls_raw:
+                break
+
+            if round_idx > MAX_TOOL_ROUNDS:
+                break
+
+            tool_results_for_history = []
+            for i, tc_raw in enumerate(tool_calls_raw):
+                tool_info = tc_raw.get("function", {})
+                tool_name = tool_info.get("name")
+                tool_args_str = tool_info.get("arguments", "{}")
+                tool_id = tc_raw.get("id")
+
+                try:
+                    tool_args = json.loads(tool_args_str)
+                except:
+                    tool_args = {}
+
+                # Check for tool confirmation if callback is provided
+                if pre_tool_callback:
+                    payload_str = json.dumps({"tool": tool_name, "args": tool_args})
+                    if not pre_tool_callback(payload_str):
+                        tool_output = "USER REJECTED: tool execution cancelled"
+                        tool_results_for_history.append({
+                            "role": "tool",
+                            "tool_call_id": tool_id,
+                            "content": tool_output,
+                        })
+                        _append_comms("IN", "tool_result", {"name": tool_name, "id": tool_id, "output": tool_output})
+                        continue
+
+                events.emit("tool_execution", payload={"status": "started", "tool": tool_name, "args": tool_args, "round": round_idx})
+
+                if tool_name in mcp_client.TOOL_NAMES:
+                    _append_comms("OUT", "tool_call", {"name": tool_name, "id": tool_id, "args": tool_args})
+                    tool_output = mcp_client.dispatch(tool_name, tool_args)
+                elif tool_name == TOOL_NAME:
+                    script = tool_args.get("script", "")
+                    _append_comms("OUT", "tool_call", {"name": TOOL_NAME, "id": tool_id, "script": script})
+                    tool_output = _run_script(script, base_dir, qa_callback)
+                else:
+                    tool_output = f"ERROR: unknown tool '{tool_name}'"
+
+                if i == len(tool_calls_raw) - 1:
+                    if file_items:
+                        file_items, changed = _reread_file_items(file_items)
+                        ctx = _build_file_diff_text(changed)
+                        if ctx:
+                            tool_output += f"\n\n[SYSTEM: FILES UPDATED]\n\n{ctx}"
+                    if round_idx == MAX_TOOL_ROUNDS:
+                        tool_output += "\n\n[SYSTEM: MAX ROUNDS. PROVIDE FINAL ANSWER.]"
+
+                tool_output = _truncate_tool_output(tool_output)
+                _cumulative_tool_bytes += len(tool_output)
+
+                tool_results_for_history.append({
+                    "role": "tool",
+                    "tool_call_id": tool_id,
+                    "content": tool_output,
+                })
+                
+                _append_comms("IN", "tool_result", {"name": tool_name, "id": tool_id, "output": tool_output})
+                events.emit("tool_execution", payload={"status": "completed", "tool": tool_name, "result": tool_output, "round": round_idx})
+
+            if _cumulative_tool_bytes > _MAX_TOOL_OUTPUT_BYTES:
+                 tool_results_for_history.append({
+                    "role": "user",
+                    "content": f"SYSTEM WARNING: Cumulative tool output exceeded {_MAX_TOOL_OUTPUT_BYTES // 1000}KB budget. Provide your final answer now."
+                 })
+                 _append_comms("OUT", "request", {"message": f"[TOOL OUTPUT BUDGET EXCEEDED: {_cumulative_tool_bytes} bytes]"})
+
+            with _deepseek_history_lock:
+                for tr in tool_results_for_history:
+                    _deepseek_history.append(tr)
+
+            # Update for next round
+            next_messages = []
+            with _deepseek_history_lock:
+                for msg in _deepseek_history:
+                    next_messages.append(msg)
+            next_messages.insert(0, sys_msg)
+            request_payload["messages"] = next_messages
+            round_idx += 1
+
+        return "\n\n".join(all_text_parts) if all_text_parts else "(No text returned)"
+
+    except Exception as e:
+        raise _classify_deepseek_error(e) from e
+
+
+def run_tier4_analysis(stderr: str) -> str:
+    """
+    Stateless Tier 4 QA analysis of an error message.
+    Uses gemini-2.5-flash-lite to summarize the error and suggest a fix.
+    """
+    if not stderr or not stderr.strip():
+        return ""
+
+    try:
+        _ensure_gemini_client()
+        prompt = (
+            f"You are a Tier 4 QA Agent specializing in error analysis.\n"
+            f"Analyze the following stderr output from a PowerShell command:\n\n"
+            f"```\n{stderr}\n```\n\n"
+            f"Provide a concise summary of the failure and suggest a fix in approximately 20 words."
+        )
+
+        # Use flash-lite for cost-effective stateless analysis
+        model_name = "gemini-2.5-flash-lite"
+
+        # We don't use the chat session here to keep it stateless
+        resp = _gemini_client.models.generate_content(
+            model=model_name,
+            contents=prompt,
+            config=types.GenerateContentConfig(
+                temperature=0.0,
+                max_output_tokens=150,
+            )
+        )
+
+        analysis = resp.text.strip()
+        return analysis
+    except Exception as e:
+        # We don't want to crash the main loop if QA analysis fails
+        return f"[QA ANALYSIS FAILED] {e}"
+
+
 # ------------------------------------------------------------------ unified send
 
 def send(
@@ -1246,6 +1814,9 @@ def send(
     base_dir: str = ".",
     file_items: list[dict] | None = None,
     discussion_history: str = "",
+    stream: bool = False,
+    pre_tool_callback = None,
+    qa_callback: Optional[Callable[[str], str]] = None,
 ) -> str:
     """
     Send a message to the active provider.
@@ -1258,23 +1829,33 @@ def send(
                          dynamic context refresh after tool calls
     discussion_history : discussion history text (used by Gemini to inject as
                          conversation message instead of caching it)
+    stream             : Whether to use streaming (supported by DeepSeek)
+    pre_tool_callback  : Optional callback (payload: str) -> bool called before tool execution
+    qa_callback        : Optional callback (stderr: str) -> str called for Tier 4 error analysis
     """
     with _send_lock:
         if _provider == "gemini":
-            return _send_gemini(md_content, user_message, base_dir, file_items, discussion_history)
+            return _send_gemini(md_content, user_message, base_dir, file_items, discussion_history, pre_tool_callback, qa_callback)
+        elif _provider == "gemini_cli":
+            return _send_gemini_cli(md_content, user_message, base_dir, file_items, discussion_history, pre_tool_callback, qa_callback)
         elif _provider == "anthropic":
-            return _send_anthropic(md_content, user_message, base_dir, file_items, discussion_history)
+            return _send_anthropic(md_content, user_message, base_dir, file_items, discussion_history, pre_tool_callback, qa_callback)
+        elif _provider == "deepseek":
+            return _send_deepseek(md_content, user_message, base_dir, file_items, discussion_history, stream=stream, pre_tool_callback=pre_tool_callback, qa_callback=qa_callback)
         raise ValueError(f"unknown provider: {_provider}")
 
-def get_history_bleed_stats() -> dict:
+def get_history_bleed_stats(md_content: str | None = None) -> dict:
     """
     Calculates how close the current conversation history is to the token limit.
+    If md_content is provided and no chat session exists, it estimates based on md_content.
     """
     if _provider == "anthropic":
         # For Anthropic, we have a robust estimator
         with _anthropic_history_lock:
             history_snapshot = list(_anthropic_history)
         current_tokens = _estimate_prompt_tokens([], history_snapshot)
+        if md_content:
+            current_tokens += max(1, int(len(md_content) / _CHARS_PER_TOKEN))
         limit_tokens = _ANTHROPIC_MAX_PROMPT_TOKENS
         percentage = (current_tokens / limit_tokens) * 100 if limit_tokens > 0 else 0
         return {
@@ -1284,33 +1865,126 @@ def get_history_bleed_stats() -> dict:
             "percentage": percentage,
         }
     elif _provider == "gemini":
+        effective_limit = _history_trunc_limit if _history_trunc_limit > 0 else _GEMINI_MAX_INPUT_TOKENS
         if _gemini_chat:
             try:
                 _ensure_gemini_client()
-                history = _get_gemini_history_list(_gemini_chat)
-                if history:
-                    resp = _gemini_client.models.count_tokens(
-                        model=_model,
-                        contents=history
-                    )
-                    current_tokens = resp.total_tokens
-                    limit_tokens = _GEMINI_MAX_INPUT_TOKENS
-                    percentage = (current_tokens / limit_tokens) * 100 if limit_tokens > 0 else 0
+                raw_history = list(_get_gemini_history_list(_gemini_chat))
+                
+                # Copy and correct roles for counting
+                history = []
+                for c in raw_history:
+                    # Gemini roles MUST be 'user' or 'model'
+                    role = "model" if c.role in ["assistant", "model"] else "user"
+                    history.append(types.Content(role=role, parts=c.parts))
+
+                if md_content:
+                    # Prepend context as a user part for counting
+                    history.insert(0, types.Content(role="user", parts=[types.Part.from_text(text=md_content)]))
+                
+                if not history:
+                    print("[DEBUG] Gemini count_tokens skipped: no history or md_content")
                     return {
                         "provider": "gemini",
-                        "limit": limit_tokens,
-                        "current": current_tokens,
-                        "percentage": percentage,
+                        "limit": effective_limit,
+                        "current": 0,
+                        "percentage": 0,
                     }
-            except Exception:
+
+                print(f"[DEBUG] Gemini count_tokens on {len(history)} messages using model {_model}")
+                resp = _gemini_client.models.count_tokens(
+                    model=_model,
+                    contents=history
+                )
+                current_tokens = resp.total_tokens
+                percentage = (current_tokens / effective_limit) * 100 if effective_limit > 0 else 0
+                print(f"[DEBUG] Gemini current_tokens={current_tokens}, percentage={percentage:.4f}%")
+                return {
+                    "provider": "gemini",
+                    "limit": effective_limit,
+                    "current": current_tokens,
+                    "percentage": percentage,
+                }
+            except Exception as e:
+                print(f"[DEBUG] Gemini count_tokens error: {e}")
+                pass
+        elif md_content:
+            try:
+                _ensure_gemini_client()
+                print(f"[DEBUG] Gemini count_tokens (MD ONLY) using model {_model}")
+                resp = _gemini_client.models.count_tokens(
+                    model=_model,
+                    contents=[types.Content(role="user", parts=[types.Part.from_text(text=md_content)])]
+                )
+                current_tokens = resp.total_tokens
+                percentage = (current_tokens / effective_limit) * 100 if effective_limit > 0 else 0
+                print(f"[DEBUG] Gemini (MD ONLY) current_tokens={current_tokens}, percentage={percentage:.4f}%")
+                return {
+                    "provider": "gemini",
+                    "limit": effective_limit,
+                    "current": current_tokens,
+                    "percentage": percentage,
+                }
+            except Exception as e:
+                print(f"[DEBUG] Gemini count_tokens (MD ONLY) error: {e}")
                 pass
         
         return {
             "provider": "gemini",
-            "limit": _GEMINI_MAX_INPUT_TOKENS,
+            "limit": effective_limit,
             "current": 0,
             "percentage": 0,
         }
+    elif _provider == "gemini_cli":
+        effective_limit = _history_trunc_limit if _history_trunc_limit > 0 else _GEMINI_MAX_INPUT_TOKENS
+        # For Gemini CLI, we don't have direct count_tokens access without making a call,
+        # so we report the limit and current usage from the last run if available.
+        limit_tokens = effective_limit
+        current_tokens = 0
+        if _gemini_cli_adapter and _gemini_cli_adapter.last_usage:
+            # Stats from CLI use 'input_tokens' or 'input'
+            u = _gemini_cli_adapter.last_usage
+            current_tokens = u.get("input_tokens") or u.get("input", 0)
+
+        percentage = (current_tokens / limit_tokens) * 100 if limit_tokens > 0 else 0
+        return {
+            "provider": "gemini_cli",
+            "limit": limit_tokens,
+            "current": current_tokens,
+            "percentage": percentage,
+        }
+    elif _provider == "deepseek":
+        limit_tokens = 64000
+        current_tokens = 0
+        with _deepseek_history_lock:
+            for msg in _deepseek_history:
+                content = msg.get("content", "")
+                if isinstance(content, str):
+                    current_tokens += len(content)
+                elif isinstance(content, list):
+                    for block in content:
+                        if isinstance(block, dict):
+                            text = block.get("text", "")
+                            if isinstance(text, str):
+                                current_tokens += len(text)
+                            inp = block.get("input")
+                            if isinstance(inp, dict):
+                                import json as _json
+                                current_tokens += len(_json.dumps(inp, ensure_ascii=False))
+
+        if md_content:
+            current_tokens += len(md_content)
+        if user_message:
+            current_tokens += len(user_message)
+
+        current_tokens = max(1, int(current_tokens / _CHARS_PER_TOKEN))
+        percentage = (current_tokens / limit_tokens) * 100 if limit_tokens > 0 else 0
+        return {
+            "provider": "deepseek",
+            "limit": limit_tokens,
+            "current": current_tokens,
+            "percentage": percentage,
+        }
 
     # Default empty state
     return {

api_hook_client.py

@@ -3,12 +3,12 @@ import json
 import time
 
 class ApiHookClient:
-    def __init__(self, base_url="http://127.0.0.1:8999", max_retries=3, retry_delay=1):
+    def __init__(self, base_url="http://127.0.0.1:8999", max_retries=5, retry_delay=0.2):
         self.base_url = base_url
         self.max_retries = max_retries
         self.retry_delay = retry_delay
 
-    def wait_for_server(self, timeout=10):
+    def wait_for_server(self, timeout=3):
         """
         Polls the /status endpoint until the server is ready or timeout is reached.
         """
@@ -18,20 +18,23 @@ class ApiHookClient:
                 if self.get_status().get('status') == 'ok':
                     return True
             except (requests.exceptions.ConnectionError, requests.exceptions.Timeout):
-                time.sleep(0.5)
+                time.sleep(0.1)
         return False
 
-    def _make_request(self, method, endpoint, data=None):
+    def _make_request(self, method, endpoint, data=None, timeout=None):
         url = f"{self.base_url}{endpoint}"
         headers = {'Content-Type': 'application/json'}
         
         last_exception = None
+        # Increase default request timeout for local server
+        req_timeout = timeout if timeout is not None else 2.0 
+        
         for attempt in range(self.max_retries + 1):
             try:
                 if method == 'GET':
-                    response = requests.get(url, timeout=2)
+                    response = requests.get(url, timeout=req_timeout)
                 elif method == 'POST':
-                    response = requests.post(url, json=data, headers=headers, timeout=2)
+                    response = requests.post(url, json=data, headers=headers, timeout=req_timeout)
                 else:
                     raise ValueError(f"Unsupported HTTP method: {method}")
 
@@ -59,7 +62,7 @@ class ApiHookClient:
         """Checks the health of the hook server."""
         url = f"{self.base_url}/status"
         try:
-            response = requests.get(url, timeout=1)
+            response = requests.get(url, timeout=0.2)
             response.raise_for_status()
             return response.json()
         except Exception:
@@ -74,6 +77,17 @@ class ApiHookClient:
     def get_session(self):
         return self._make_request('GET', '/api/session')
 
+    def get_mma_status(self):
+        """Retrieves current MMA status (track, tickets, tier, etc.)"""
+        return self._make_request('GET', '/api/gui/mma_status')
+
+    def push_event(self, event_type, payload):
+        """Pushes an event to the GUI's AsyncEventQueue via the /api/gui endpoint."""
+        return self.post_gui({
+            "action": event_type,
+            "payload": payload
+        })
+
     def get_performance(self):
         """Retrieves UI performance metrics."""
         return self._make_request('GET', '/api/performance')
@@ -83,3 +97,132 @@ class ApiHookClient:
 
     def post_gui(self, gui_data):
         return self._make_request('POST', '/api/gui', data=gui_data)
+
+    def select_tab(self, tab_bar, tab):
+        """Tells the GUI to switch to a specific tab in a tab bar."""
+        return self.post_gui({
+            "action": "select_tab",
+            "tab_bar": tab_bar,
+            "tab": tab
+        })
+
+    def select_list_item(self, listbox, item_value):
+        """Tells the GUI to select an item in a listbox by its value."""
+        return self.post_gui({
+            "action": "select_list_item",
+            "listbox": listbox,
+            "item_value": item_value
+        })
+
+    def set_value(self, item, value):
+        """Sets the value of a GUI item."""
+        return self.post_gui({
+            "action": "set_value",
+            "item": item,
+            "value": value
+        })
+
+    def get_value(self, item):
+        """Gets the value of a GUI item via its mapped field."""
+        try:
+            # First try direct field querying via POST
+            res = self._make_request('POST', '/api/gui/value', data={"field": item})
+            if res and "value" in res:
+                v = res.get("value")
+                if v is not None:
+                    return v
+        except Exception:
+            pass
+
+        try:
+            # Try GET fallback
+            res = self._make_request('GET', f'/api/gui/value/{item}')
+            if res and "value" in res:
+                v = res.get("value")
+                if v is not None:
+                    return v
+        except Exception:
+            pass
+
+        try:
+            # Fallback for thinking/live/prior which are in diagnostics
+            diag = self._make_request('GET', '/api/gui/diagnostics')
+            if item in diag:
+                return diag[item]
+            # Map common indicator tags to diagnostics keys
+            mapping = {
+                "thinking_indicator": "thinking",
+                "operations_live_indicator": "live",
+                "prior_session_indicator": "prior"
+            }
+            key = mapping.get(item)
+            if key and key in diag:
+                return diag[key]
+        except Exception:
+            pass
+        return None
+
+    def click(self, item, *args, **kwargs):
+        """Simulates a click on a GUI button or item."""
+        user_data = kwargs.pop('user_data', None)
+        return self.post_gui({
+            "action": "click",
+            "item": item,
+            "args": args,
+            "kwargs": kwargs,
+            "user_data": user_data
+        })
+
+    def get_indicator_state(self, tag):
+        """Checks if an indicator is shown using the diagnostics endpoint."""
+        # Mapping tag to the keys used in diagnostics endpoint
+        mapping = {
+            "thinking_indicator": "thinking",
+            "operations_live_indicator": "live",
+            "prior_session_indicator": "prior"
+        }
+        key = mapping.get(tag, tag)
+        try:
+            diag = self._make_request('GET', '/api/gui/diagnostics')
+            return {"tag": tag, "shown": diag.get(key, False)}
+        except Exception as e:
+            return {"tag": tag, "shown": False, "error": str(e)}
+
+    def get_events(self):
+        """Fetches and clears the event queue from the server."""
+        try:
+            return self._make_request('GET', '/api/events').get("events", [])
+        except Exception:
+            return []
+
+    def wait_for_event(self, event_type, timeout=5):
+        """Polls for a specific event type."""
+        start = time.time()
+        while time.time() - start < timeout:
+            events = self.get_events()
+            for ev in events:
+                if ev.get("type") == event_type:
+                    return ev
+            time.sleep(0.1) # Fast poll
+        return None
+
+    def wait_for_value(self, item, expected, timeout=5):
+        """Polls until get_value(item) == expected."""
+        start = time.time()
+        while time.time() - start < timeout:
+            if self.get_value(item) == expected:
+                return True
+            time.sleep(0.1) # Fast poll
+        return False
+
+    def reset_session(self):
+        """Simulates clicking the 'Reset Session' button in the GUI."""
+        return self.click("btn_reset")
+
+    def request_confirmation(self, tool_name, args):
+        """Asks the user for confirmation via the GUI (blocking call)."""
+        # Using a long timeout as this waits for human input (60 seconds)
+        res = self._make_request('POST', '/api/ask',
+                                data={'type': 'tool_approval', 'tool': tool_name, 'args': args},
+                                timeout=60.0)
+        return res.get('response')

api_hooks.py

@@ -1,10 +1,11 @@
 import json
 import threading
-from http.server import HTTPServer, BaseHTTPRequestHandler
+import uuid
+from http.server import ThreadingHTTPServer, BaseHTTPRequestHandler
 import logging
 import session_logger
 
-class HookServerInstance(HTTPServer):
+class HookServerInstance(ThreadingHTTPServer):
     """Custom HTTPServer that carries a reference to the main App instance."""
     def __init__(self, server_address, RequestHandlerClass, app):
         super().__init__(server_address, RequestHandlerClass)
@@ -21,11 +22,12 @@ class HookHandler(BaseHTTPRequestHandler):
             self.end_headers()
             self.wfile.write(json.dumps({'status': 'ok'}).encode('utf-8'))
         elif self.path == '/api/project':
+            import project_manager
             self.send_response(200)
             self.send_header('Content-Type', 'application/json')
             self.end_headers()
-            self.wfile.write(
-                json.dumps({'project': app.project}).encode('utf-8'))
+            flat = project_manager.flat_config(app.project)
+            self.wfile.write(json.dumps({'project': flat}).encode('utf-8'))
         elif self.path == '/api/session':
             self.send_response(200)
             self.send_header('Content-Type', 'application/json')
@@ -41,6 +43,141 @@ class HookHandler(BaseHTTPRequestHandler):
             if hasattr(app, 'perf_monitor'):
                 metrics = app.perf_monitor.get_metrics()
             self.wfile.write(json.dumps({'performance': metrics}).encode('utf-8'))
+        elif self.path == '/api/events':
+            # Long-poll or return current event queue
+            self.send_response(200)
+            self.send_header('Content-Type', 'application/json')
+            self.end_headers()
+            events = []
+            if hasattr(app, '_api_event_queue'):
+                with app._api_event_queue_lock:
+                    events = list(app._api_event_queue)
+                    app._api_event_queue.clear()
+            self.wfile.write(json.dumps({'events': events}).encode('utf-8'))
+        elif self.path == '/api/gui/value':
+            # POST with {"field": "field_tag"} to get value
+            content_length = int(self.headers.get('Content-Length', 0))
+            body = self.rfile.read(content_length)
+            data = json.loads(body.decode('utf-8'))
+            field_tag = data.get("field")
+            print(f"[DEBUG] Hook Server: get_value for {field_tag}")
+
+            event = threading.Event()
+            result = {"value": None}
+
+            def get_val():
+                try:
+                    if field_tag in app._settable_fields:
+                        attr = app._settable_fields[field_tag]
+                        val = getattr(app, attr, None)
+                        print(f"[DEBUG] Hook Server: attr={attr}, val={val}")
+                        result["value"] = val
+                    else:
+                        print(f"[DEBUG] Hook Server: {field_tag} NOT in settable_fields")
+                finally:
+                    event.set()
+
+            with app._pending_gui_tasks_lock:
+                app._pending_gui_tasks.append({
+                    "action": "custom_callback",
+                    "callback": get_val
+                })
+
+            if event.wait(timeout=2):
+                self.send_response(200)
+                self.send_header('Content-Type', 'application/json')
+                self.end_headers()
+                self.wfile.write(json.dumps(result).encode('utf-8'))
+            else:
+                self.send_response(504)
+                self.end_headers()
+        elif self.path.startswith('/api/gui/value/'):
+            # Generic endpoint to get the value of any settable field
+            field_tag = self.path.split('/')[-1]
+            event = threading.Event()
+            result = {"value": None}
+
+            def get_val():
+                try:
+                    if field_tag in app._settable_fields:
+                        attr = app._settable_fields[field_tag]
+                        result["value"] = getattr(app, attr, None)
+                finally:
+                    event.set()
+
+            with app._pending_gui_tasks_lock:
+                app._pending_gui_tasks.append({
+                    "action": "custom_callback",
+                    "callback": get_val
+                })
+
+            if event.wait(timeout=2):
+                self.send_response(200)
+                self.send_header('Content-Type', 'application/json')
+                self.end_headers()
+                self.wfile.write(json.dumps(result).encode('utf-8'))
+            else:
+                self.send_response(504)
+                self.end_headers()
+        elif self.path == '/api/gui/mma_status':
+            event = threading.Event()
+            result = {}
+
+            def get_mma():
+                try:
+                    result["mma_status"] = getattr(app, "mma_status", "idle")
+                    result["active_tier"] = getattr(app, "active_tier", None)
+                    result["active_track"] = getattr(app, "active_track", None)
+                    result["active_tickets"] = getattr(app, "active_tickets", [])
+                    result["mma_step_mode"] = getattr(app, "mma_step_mode", False)
+                    result["pending_approval"] = app._pending_mma_approval is not None
+                finally:
+                    event.set()
+
+            with app._pending_gui_tasks_lock:
+                app._pending_gui_tasks.append({
+                    "action": "custom_callback",
+                    "callback": get_mma
+                })
+
+            if event.wait(timeout=2):
+                self.send_response(200)
+                self.send_header('Content-Type', 'application/json')
+                self.end_headers()
+                self.wfile.write(json.dumps(result).encode('utf-8'))
+            else:
+                self.send_response(504)
+                self.end_headers()
+        elif self.path == '/api/gui/diagnostics':
+            # Safe way to query multiple states at once via the main thread queue
+            event = threading.Event()
+            result = {}
+
+            def check_all():
+                try:
+                    # Generic state check based on App attributes (works for both DPG and ImGui versions)
+                    status = getattr(app, "ai_status", "idle")
+                    result["thinking"] = status in ["sending...", "running powershell..."]
+                    result["live"] = status in ["running powershell...", "fetching url...", "searching web...", "powershell done, awaiting AI..."]
+                    result["prior"] = getattr(app, "is_viewing_prior_session", False)
+                finally:
+                    event.set()
+
+            with app._pending_gui_tasks_lock:
+                app._pending_gui_tasks.append({
+                    "action": "custom_callback",
+                    "callback": check_all
+                })
+
+            if event.wait(timeout=2):
+                self.send_response(200)
+                self.send_header('Content-Type', 'application/json')
+                self.end_headers()
+                self.wfile.write(json.dumps(result).encode('utf-8'))
+            else:
+                self.send_response(504)
+                self.end_headers()
+                self.wfile.write(json.dumps({'error': 'timeout'}).encode('utf-8'))
         else:
             self.send_response(404)
             self.end_headers()
@@ -70,11 +207,6 @@ class HookHandler(BaseHTTPRequestHandler):
                 self.wfile.write(
                     json.dumps({'status': 'updated'}).encode('utf-8'))
             elif self.path == '/api/gui':
-                if not hasattr(app, '_pending_gui_tasks'):
-                    app._pending_gui_tasks = []
-                if not hasattr(app, '_pending_gui_tasks_lock'):
-                    app._pending_gui_tasks_lock = threading.Lock()
-                    
                 with app._pending_gui_tasks_lock:
                     app._pending_gui_tasks.append(data)
 
@@ -83,6 +215,75 @@ class HookHandler(BaseHTTPRequestHandler):
                 self.end_headers()
                 self.wfile.write(
                     json.dumps({'status': 'queued'}).encode('utf-8'))
+            elif self.path == '/api/ask':
+                request_id = str(uuid.uuid4())
+                event = threading.Event()
+
+                if not hasattr(app, '_pending_asks'):
+                    app._pending_asks = {}
+                if not hasattr(app, '_ask_responses'):
+                    app._ask_responses = {}
+
+                app._pending_asks[request_id] = event
+
+                # Emit event for test/client discovery
+                with app._api_event_queue_lock:
+                    app._api_event_queue.append({
+                        "type": "ask_received",
+                        "request_id": request_id,
+                        "data": data
+                    })
+
+                with app._pending_gui_tasks_lock:
+                    app._pending_gui_tasks.append({
+                        "type": "ask",
+                        "request_id": request_id,
+                        "data": data
+                    })
+
+                if event.wait(timeout=60.0):
+                    response_data = app._ask_responses.get(request_id)
+                    # Clean up response after reading
+                    if request_id in app._ask_responses:
+                        del app._ask_responses[request_id]
+                    
+                    self.send_response(200)
+                    self.send_header('Content-Type', 'application/json')
+                    self.end_headers()
+                    self.wfile.write(json.dumps({'status': 'ok', 'response': response_data}).encode('utf-8'))
+                else:
+                    if request_id in app._pending_asks:
+                        del app._pending_asks[request_id]
+                    self.send_response(504)
+                    self.end_headers()
+                    self.wfile.write(json.dumps({'error': 'timeout'}).encode('utf-8'))
+
+            elif self.path == '/api/ask/respond':
+                request_id = data.get('request_id')
+                response_data = data.get('response')
+
+                if request_id and hasattr(app, '_pending_asks') and request_id in app._pending_asks:
+                    app._ask_responses[request_id] = response_data
+                    event = app._pending_asks[request_id]
+                    event.set()
+
+                    # Clean up pending ask entry
+                    del app._pending_asks[request_id]
+
+                    # Queue GUI task to clear the dialog
+                    with app._pending_gui_tasks_lock:
+                        app._pending_gui_tasks.append({
+                            "action": "clear_ask",
+                            "request_id": request_id
+                        })
+
+                    self.send_response(200)
+                    self.send_header('Content-Type', 'application/json')
+                    self.end_headers()
+                    self.wfile.write(json.dumps({'status': 'ok'}).encode('utf-8'))
+                else:
+                    self.send_response(404)
+                    self.end_headers()
             else:
                 self.send_response(404)
                 self.end_headers()
@@ -103,8 +304,31 @@ class HookServer:
         self.thread = None
 
     def start(self):
-        if not getattr(self.app, 'test_hooks_enabled', False):
+        if self.thread and self.thread.is_alive():
             return
+
+        is_gemini_cli = getattr(self.app, 'current_provider', '') == 'gemini_cli'
+        if not getattr(self.app, 'test_hooks_enabled', False) and not is_gemini_cli:
+            return
+
+        # Ensure the app has the task queue and lock initialized
+        if not hasattr(self.app, '_pending_gui_tasks'):
+            self.app._pending_gui_tasks = []
+        if not hasattr(self.app, '_pending_gui_tasks_lock'):
+            self.app._pending_gui_tasks_lock = threading.Lock()
+
+        # Initialize ask-related dictionaries
+        if not hasattr(self.app, '_pending_asks'):
+            self.app._pending_asks = {}
+        if not hasattr(self.app, '_ask_responses'):
+            self.app._ask_responses = {}
+
+        # Event queue for test script subscriptions
+        if not hasattr(self.app, '_api_event_queue'):
+            self.app._api_event_queue = []
+        if not hasattr(self.app, '_api_event_queue_lock'):
+            self.app._api_event_queue_lock = threading.Lock()
+
         self.server = HookServerInstance(('127.0.0.1', self.port), HookHandler, self.app)
         self.thread = threading.Thread(target=self.server.serve_forever, daemon=True)
         self.thread.start()

conductor/archive/deepseek_support_20260225/index.md

@@ -0,0 +1,5 @@
+# Track deepseek_support_20260225 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/deepseek_support_20260225/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "deepseek_support_20260225",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-25T00:00:00Z",
+  "updated_at": "2026-02-25T00:00:00Z",
+  "description": "Add support for the deepseek api as a provider."
+}

conductor/archive/deepseek_support_20260225/plan.md

@@ -0,0 +1,27 @@
+# Implementation Plan: DeepSeek API Provider Support
+
+## Phase 1: Infrastructure & Common Logic [checkpoint: 0ec3720]
+- [x] Task: Initialize MMA Environment `activate_skill mma-orchestrator` 1b3ff23
+- [x] Task: Update `credentials.toml` schema and configuration logic in `project_manager.py` to support `deepseek` 1b3ff23
+- [x] Task: Define the `DeepSeekProvider` interface in `ai_client.py` and align with existing provider patterns 1b3ff23
+- [x] Task: Conductor - User Manual Verification 'Infrastructure & Common Logic' (Protocol in workflow.md) 1b3ff23
+
+## Phase 2: DeepSeek API Client Implementation
+- [x] Task: Write failing tests for `DeepSeekProvider` model selection and basic completion
+- [x] Task: Implement `DeepSeekProvider` using the dedicated SDK
+- [x] Task: Write failing tests for streaming and tool calling parity in `DeepSeekProvider`
+- [x] Task: Implement streaming and tool calling logic for DeepSeek models
+- [x] Task: Conductor - User Manual Verification 'DeepSeek API Client Implementation' (Protocol in workflow.md)
+
+## Phase 3: Reasoning Traces & Advanced Capabilities
+- [x] Task: Write failing tests for reasoning trace capture in `DeepSeekProvider` (DeepSeek-R1)
+- [x] Task: Implement reasoning trace processing and integration with discussion history
+- [x] Task: Write failing tests for token estimation and cost tracking for DeepSeek models
+- [x] Task: Implement token usage tracking according to DeepSeek pricing
+- [x] Task: Conductor - User Manual Verification 'Reasoning Traces & Advanced Capabilities' (Protocol in workflow.md)
+
+## Phase 4: GUI Integration & Final Verification
+- [x] Task: Update `gui_2.py` and `theme_2.py` (if necessary) to include DeepSeek in the provider selection UI
+- [x] Task: Implement automated regression tests for the full DeepSeek lifecycle (prompt, streaming, tool call, reasoning)
+- [x] Task: Verify overall performance and UI responsiveness with the new provider
+- [x] Task: Conductor - User Manual Verification 'GUI Integration & Final Verification' (Protocol in workflow.md)

conductor/archive/deepseek_support_20260225/spec.md

@@ -0,0 +1,31 @@
+# Specification: DeepSeek API Provider Support
+
+## Overview
+Implement a new AI provider module to support the DeepSeek API within the Manual Slop application. This integration will leverage a dedicated SDK to provide access to high-performance models (DeepSeek-V3 and DeepSeek-R1) with support for streaming, tool calling, and detailed reasoning traces.
+
+## Functional Requirements
+- **Dedicated SDK Integration:** Utilize a DeepSeek-specific Python client for API interactions.
+- **Model Support:** Initial support for `deepseek-v3` (general performance) and `deepseek-r1` (reasoning).
+- **Core Features:**
+    - **Streaming:** Support real-time response generation for a better user experience.
+    - **Tool Calling:** Integrate with Manual Slop's existing tool/function execution framework.
+    - **Reasoning Traces:** Capture and display reasoning paths if provided by the model (e.g., DeepSeek-R1).
+- **Configuration Management:**
+    - Add `[deepseek]` section to `credentials.toml` for `api_key`.
+    - Update `config.toml` to allow selecting DeepSeek as the active provider.
+
+## Non-Functional Requirements
+- **Parity:** Maintain consistency with existing Gemini and Anthropic provider implementations in `ai_client.py`.
+- **Error Handling:** Robust handling of API rate limits and connection issues specific to DeepSeek.
+- **Observability:** Track token usage and costs according to DeepSeek's pricing model.
+
+## Acceptance Criteria
+- [ ] User can select "DeepSeek" as a provider in the GUI.
+- [ ] Successful completion of prompts using both DeepSeek-V3 and DeepSeek-R1 models.
+- [ ] Tool calling works correctly for standard operations (e.g., `read_file`).
+- [ ] Reasoning traces from R1 are captured and visible in the discussion history.
+- [ ] Streaming responses function correctly without blocking the GUI.
+
+## Out of Scope
+- Support for OpenAI-compatible proxies for DeepSeek in this initial track.
+- Automated fine-tuning or custom model endpoints.

conductor/archive/gemini_cli_headless_20260224/index.md

@@ -0,0 +1,5 @@
+# Track gemini_cli_headless_20260224 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/gemini_cli_headless_20260224/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "gemini_cli_headless_20260224",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-24T23:45:00Z",
+  "updated_at": "2026-02-24T23:45:00Z",
+  "description": "Support gemini cli headless as an alternative to the raw client_api route. So that they user may use their gemini subscription and gemini cli features within manual slop for a more discliplined and visually enriched UX."
+}

conductor/archive/gemini_cli_headless_20260224/plan.md

@@ -0,0 +1,26 @@
+# Implementation Plan: Gemini CLI Headless Integration
+
+## Phase 1: IPC Infrastructure Extension [checkpoint: c0bccce]
+- [x] Task: Extend `api_hooks.py` to support synchronous "Ask" requests. This involves adding a way for a client to POST a request and wait for a user response from the GUI. (1792107)
+- [x] Task: Update `api_hook_client.py` with a `request_confirmation(tool_name, args)` method that blocks until the GUI responds. (93f640d)
+- [x] Task: Create a standalone test script `tests/test_sync_hooks.py` to verify that the CLI-to-GUI communication works as expected. (1792107)
+- [x] Task: Conductor - User Manual Verification 'Phase 1: IPC Infrastructure Extension' (Protocol in workflow.md) (c0bccce)
+
+## Phase 2: Gemini CLI Adapter & Tool Bridge
+- [x] Task: Implement `scripts/cli_tool_bridge.py`. This script will be called by the Gemini CLI `BeforeTool` hook and use `ApiHookClient` to talk to the GUI. (211000c)
+- [x] Task: Implement the `GeminiCliAdapter` in `ai_client.py` (or a new `gemini_cli_adapter.py`). It must handle the `subprocess` lifecycle and parse the `stream-json` output. (b762a80)
+- [x] Task: Integrate `GeminiCliAdapter` into the main `ai_client.send()` logic. (b762a80)
+- [x] Task: Write unit tests for the JSON parsing and subprocess management in `GeminiCliAdapter`. (b762a80)
+- [~] Task: Conductor - User Manual Verification 'Phase 2: Gemini CLI Adapter & Tool Bridge' (Protocol in workflow.md)
+
+## Phase 3: GUI Integration & Provider Support
+- [x] Task: Update `gui_2.py` to add "Gemini CLI" to the provider dropdown. (3ce4fa0)
+- [x] Task: Implement UI elements for "Gemini CLI Session Management" (Login button, session ID display). (3ce4fa0)
+- [x] Task: Update the `manual_slop.toml` logic to persist Gemini CLI specific settings (e.g., path to CLI, approval mode). (3ce4fa0)
+- [~] Task: Conductor - User Manual Verification 'Phase 3: GUI Integration & Provider Support' (Protocol in workflow.md)
+
+## Phase 4: Integration Testing & UX Polish
+- [x] Task: Create a comprehensive integration test `tests/test_gemini_cli_integration.py` that uses the `live_gui` fixture to simulate a full session. (d187a6c)
+- [x] Task: Verify tool confirmation flow: CLI Tool -> Bridge -> GUI Modal -> User Approval -> CLI Execution. (d187a6c)
+- [x] Task: Polish the display of CLI telemetry (tokens/latency) in the GUI diagnostics panel. (1e5b43e)
+- [x] Task: Conductor - User Manual Verification 'Phase 4: Integration Testing & UX Polish' (Protocol in workflow.md) (1e5b43e)

conductor/archive/gemini_cli_headless_20260224/spec.md

@@ -0,0 +1,45 @@
+# Specification: Gemini CLI Headless Integration
+
+## Overview
+This track integrates the `gemini` CLI as a headless backend provider for Manual Slop. This allows users to leverage their Gemini subscription and the CLI's advanced features (e.g., specialized sub-agents like `codebase_investigator`, structured JSON streaming, and robust session management) directly within the Manual Slop GUI.
+
+## Goals
+- Add "Gemini CLI" as a selectable AI provider in Manual Slop.
+- Support both persistent interactive sessions and one-off task-specific delegation (e.g., running `gemini investigate`).
+- Implement a secure "BeforeTool" hook to ensure all CLI-initiated tool calls are intercepted and confirmed via the Manual Slop GUI.
+- Capture and display the CLI's visually enriched output (via JSONL stream) within the existing discussion history.
+
+## Functional Requirements
+
+### 1. Gemini CLI Provider Adapter
+- **Implementation**: Create a `GeminiCliAdapter` class (or extend `ai_client.py`) that wraps the `gemini` CLI subprocess.
+- **Communication**: Use `--output-format stream-json` to receive real-time updates (text chunks, tool calls, status).
+- **Session Management**: Support session persistence by tracking the session ID and passing it to subsequent CLI calls.
+- **Authentication**: 
+    - Provide a "Login to Gemini CLI" action in the GUI that triggers `gemini login`.
+    - Support passing an API key via environment variables if configured in `manual_slop.toml`.
+
+### 2. GUI Intercepted Tool Execution
+- **Mechanism**: Use the Gemini CLI's `BeforeTool` hook.
+- **Hook Helper**: A small Python script `scripts/cli_tool_bridge.py` will be registered as the `BeforeTool` hook.
+- **IPC**: This bridge script will communicate with Manual Slop's `HookServer` (extending it to support synchronous "ask" requests).
+- **Confirmation**: When a tool is requested, the bridge blocks until the user confirms/denies the action in the GUI, returning the decision as JSON to the CLI.
+
+### 3. Visual & Telemetry Integration
+- **Rich Output**: Parse the `stream-json` events to display markdown content and tool status in the GUI.
+- **Telemetry**: Extract and display token usage and latency metrics provided by the CLI's `result` event.
+
+## Non-Functional Requirements
+- **Performance**: The subprocess bridge should introduce minimal latency (<100ms overhead for communication).
+- **Reliability**: Gracefully handle CLI crashes or timeouts by reporting errors in the GUI and allowing session resets.
+
+## Acceptance Criteria
+- [ ] User can select "Gemini CLI" in the Provider dropdown.
+- [ ] User can successfully send messages and receive streamed responses from the CLI.
+- [ ] Any tool call (PowerShell/MCP) initiated by the CLI triggers the standard Manual Slop confirmation modal.
+- [ ] Tools only execute after user approval; rejection correctly notifies the CLI agent.
+- [ ] Session history is maintained correctly across multiple turns when using the CLI provider.
+
+## Out of Scope
+- Full terminal emulation (ANSI color support) within the GUI; the focus is on structured text and data.
+- Migrating existing raw `client_api` sessions to CLI sessions.

conductor/archive/gemini_cli_parity_20260225/index.md

@@ -0,0 +1,5 @@
+# Track gemini_cli_parity_20260225 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/gemini_cli_parity_20260225/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "gemini_cli_parity_20260225",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-25T00:00:00Z",
+  "updated_at": "2026-02-25T00:00:00Z",
+  "description": "Make sure gemini cli behavior and feature set have full parity with regular direct gemini api usage in ai_client.py and elsewhere"
+}

conductor/archive/gemini_cli_parity_20260225/plan.md

@@ -0,0 +1,32 @@
+# Implementation Plan: Gemini CLI Parity
+
+## Phase 1: Infrastructure & Common Logic
+- [x] Task: Initialize MMA Environment `activate_skill mma-orchestrator`
+- [x] Task: Audit `gemini_cli_adapter.py` and `ai_client.py` for parity gaps (Findings: missing count_tokens, safety settings, and robust system prompt handling in CLI adapter)
+- [x] Task: Implement common logging utilities for CLI bridge observability
+- [x] Task: Conductor - User Manual Verification 'Infrastructure & Common Logic' (Protocol in workflow.md)
+
+## Phase 2: Token Counting & Safety Settings
+- [x] Task: Write failing tests for token estimation in `GeminiCLIAdapter`
+- [x] Task: Implement token counting parity in `GeminiCLIAdapter`
+- [x] Task: Write failing tests for safety setting application in `GeminiCLIAdapter`
+- [x] Task: Implement safety filter application in `GeminiCLIAdapter`
+- [x] Task: Conductor - User Manual Verification 'Token Counting & Safety Settings' (Protocol in workflow.md)
+
+## Phase 3: Tool Calling Parity & System Instructions
+- [x] Task: Write failing tests for system instruction usage in `GeminiCLIAdapter`
+- [x] Task: Implement system instruction propagation in `GeminiCLIAdapter`
+- [x] Task: Write failing tests for tool call/response mapping in `cli_tool_bridge.py`
+- [x] Task: Synchronize tool call handling between bridge and `ai_client.py`
+- [x] Task: Conductor - User Manual Verification 'Tool Calling Parity & System Instructions' (Protocol in workflow.md)
+
+## Phase 4: Final Verification & Performance Diagnostics
+- [x] Task: Implement automated parity regression tests comparing CLI vs Direct API outputs
+- [x] Task: Verify bridge latency and error handling robustness
+- [x] Task: Conductor - User Manual Verification 'Final Verification & Performance Diagnostics' (Protocol in workflow.md)
+
+## Phase 5: Edge Case Resilience & GUI Integration Tests
+- [x] Task: Implement tests for context bleed prevention (filtering non-assistant messages)
+- [x] Task: Implement tests for parameter name resilience (dir_path/file_path aliases)
+- [x] Task: Implement tests for tool call loop termination and payload persistence
+- [x] Task: Conductor - User Manual Verification 'Edge Case Resilience' (Protocol in workflow.md)

conductor/archive/gemini_cli_parity_20260225/spec.md

@@ -0,0 +1,27 @@
+# Specification: Gemini CLI Parity
+
+## Overview
+Achieve full functional and behavioral parity between the Gemini CLI integration (`gemini_cli_adapter.py`, `cli_tool_bridge.py`) and the direct Gemini API implementation (`ai_client.py`). This ensures that users leveraging the Gemini CLI as a headless backend provider experience the same level of capability, reliability, and observability as direct API users.
+
+## Functional Requirements
+- **Token Estimation Parity:** Implement accurate token counting for both input and output in the Gemini CLI adapter to match the precision of the direct API.
+- **Safety Settings Parity:** Enable full configuration and enforcement of Gemini safety filters when using the CLI provider.
+- **Tool Calling Parity:** Synchronize tool definition mapping, call handling, and response processing between the CLI bridge and the direct SDK.
+- **System Instructions Parity:** Ensure system prompts and instructions are consistently passed and handled across both providers.
+- **Bridge Robustness:** Enhance the `cli_tool_bridge.py` and adapter to improve latency, error handling (retries), and detailed subprocess observability.
+
+## Non-Functional Requirements
+- **Observability:** Detailed logging of CLI subprocess interactions for debugging.
+- **Performance:** Minimize the overhead introduced by the bridge mechanism.
+- **Maintainability:** Ensure that future changes to `ai_client.py` can be easily mirrored in the CLI adapter.
+
+## Acceptance Criteria
+- [ ] Token counts for identical prompts match within a 5% margin between CLI and Direct API.
+- [ ] Safety settings configured in the GUI are correctly applied to CLI sessions.
+- [ ] Tool calls from the CLI are successfully executed and returned via the bridge without loss of context.
+- [ ] System instructions are correctly utilized by the model when using the CLI.
+- [ ] Automated tests verify that responses and tool execution flows are identical for both providers.
+
+## Out of Scope
+- Performance optimizations for the `gemini` CLI binary itself.
+- Support for non-Gemini CLI providers in this track.

conductor/archive/gui2_feature_parity_20260223/index.md

@@ -0,0 +1,5 @@
+# Track gui2_feature_parity_20260223 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/gui2_feature_parity_20260223/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "gui2_feature_parity_20260223",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-23T20:15:30Z",
+  "updated_at": "2026-02-23T20:15:30Z",
+  "description": "get gui_2 working with latest changes to the project."
+}

conductor/archive/gui2_feature_parity_20260223/plan.md

@@ -0,0 +1,82 @@
+# Implementation Plan: GUIv2 Feature Parity
+
+## Phase 1: Core Architectural Integration [checkpoint: 712d5a8]
+
+- [x] **Task:** Integrate `events.py` into `gui_2.py`. [24b831c]
+    - [x] Sub-task: Import the `events` module in `gui_2.py`.
+    - [x] Sub-task: Refactor the `ai_client` call in `_do_send` to use the event-driven `send` method.
+    - [x] Sub-task: Create event handlers in `App` class for `request_start`, `response_received`, and `tool_execution`.
+    - [x] Sub-task: Subscribe the handlers to `ai_client.events` upon `App` initialization.
+- [x] **Task:** Integrate `mcp_client.py` for native file tools. [ece84d4]
+    - [x] Sub-task: Import `mcp_client` in `gui_2.py`.
+    - [x] Sub-task: Add `mcp_client.perf_monitor_callback` to the `App` initialization.
+    - [x] Sub-task: In `ai_client`, ensure the MCP tools are registered and available for the AI to call when `gui_2.py` is the active UI.
+- [x] **Task:** Write tests for new core integrations. [ece84d4]
+    - [x] Sub-task: Create `tests/test_gui2_events.py` to verify that `gui_2.py` correctly handles AI lifecycle events.
+    - [x] Sub-task: Create `tests/test_gui2_mcp.py` to verify that the AI can use MCP tools through `gui_2.py`.
+- [x] **Task:** Conductor - User Manual Verification 'Core Architectural Integration' (Protocol in workflow.md)
+
+## Phase 2: Major Feature Implementation
+
+- [x] **Task:** Port the API Hooks System. [merged]
+    - [x] Sub-task: Import `api_hooks` in `gui_2.py`.
+    - [x] Sub-task: Instantiate `HookServer` in the `App` class.
+    - [x] Sub-task: Implement the logic to start the server based on a CLI flag (e.g., `--enable-test-hooks`).
+    - [x] Sub-task: Implement the queue and lock for pending GUI tasks from the hook server, similar to `gui.py`.
+    - [x] Sub-task: Add a main loop task to process the GUI task queue.
+- [x] **Task:** Port the Performance & Diagnostics feature. [merged]
+    - [x] Sub-task: Import `PerformanceMonitor` in `gui_2.py`.
+    - [x] Sub-task: Instantiate `PerformanceMonitor` in the `App` class.
+    - [x] Sub-task: Create a new "Diagnostics" window in `gui_2.py`.
+    - [x] Sub-task: Add UI elements (plots, labels) to the Diagnostics window to display FPS, CPU, frame time, etc.
+    - [x] Sub-task: Add a throttled update mechanism in the main loop to refresh diagnostics data.
+- [x] **Task:** Implement the Prior Session Viewer. [merged]
+    - [x] Sub-task: Add a "Load Prior Session" button to the UI.
+    - [x] Sub-task: Implement the file dialog logic to select a `.log` file.
+    - [x] Sub-task: Implement the logic to parse the log file and populate the comms history view.
+    - [x] Sub-task: Implement the "tinted" theme application when in viewing mode and a way to exit this mode.
+- [x] **Task:** Write tests for major features.
+    - [x] Sub-task: Create `tests/test_gui2_api_hooks.py` to test the hook server integration.
+    - [x] Sub-task: Create `tests/test_gui2_diagnostics.py` to verify the diagnostics panel displays data.
+- [x] **Task:** Conductor - User Manual Verification 'Major Feature Implementation' (Protocol in workflow.md)
+
+## Phase 3: UI/UX Refinement [checkpoint: cc5074e]
+
+- [x] **Task:** Refactor UI to a "Hub" based layout. [ddb53b2]
+    - [x] Sub-task: Analyze the docking layout of `gui.py`.
+    - [x] Sub-task: Create wrapper windows for "Context Hub", "AI Settings Hub", "Discussion Hub", and "Operations Hub" in `gui_2.py`.
+    - [x] Sub-task: Move existing windows into their respective Hubs using the `imgui-bundle` docking API.
+    - [x] Sub-task: Ensure the default layout is saved to and loaded from `manualslop_layout.ini`.
+- [x] **Task:** Add Agent Capability Toggles to the UI. [merged]
+    - [x] Sub-task: In the "Projects" or a new "Agent" panel, add checkboxes for each agent tool (e.g., `run_powershell`, `read_file`).
+    - [x] Sub-task: Ensure these UI toggles are saved to the project\'s `.toml` file.
+    - [x] Sub-task: Ensure `ai_client` respects these settings when determining which tools are available to the AI.
+- [x] **Task:** Full Theme Integration. [merged]
+    - [x] Sub-task: Review all newly added windows and controls.
+    - [x] Sub-task: Ensure that colors, fonts, and scaling from `theme_2.py` are correctly applied everywhere.
+    - [x] Sub-task: Test theme switching to confirm all elements update correctly.
+- [x] **Task:** Write tests for UI/UX changes. [ddb53b2]
+    - [x] Sub-task: Create `tests/test_gui2_layout.py` to verify the hub structure is created.
+    - [x] Sub-task: Add tests to verify agent capability toggles are respected.
+- [x] **Task:** Conductor - User Manual Verification 'UI/UX Refinement' (Protocol in workflow.md)
+
+## Phase 4: Finalization and Verification
+
+- [x] **Task:** Conduct full manual testing against `spec.md` Acceptance Criteria. (Note: Some UI display issues for text panels persist and will be addressed in a future track.)
+    - [x] Sub-task: Verify AC1: `gui_2.py` launches.
+    - [x] Sub-task: Verify AC2: Hub layout is correct.
+    - [x] Sub-task: Verify AC3: Diagnostics panel works.
+    - [x] Sub-task: Verify AC4: API hooks server runs.
+    - [x] Sub-task: Verify AC5: MCP tools are usable by AI.
+    - [x] Sub-task: Verify AC6: Prior Session Viewer works.
+    - [x] Sub-task: Verify AC7: Theming is consistent.
+- [x] **Task:** Run the full project test suite.
+    - [x] Sub-task: Execute `uv run run_tests.py` (or equivalent).
+    - [x] Sub-task: Ensure all existing and new tests pass.
+- [x] **Task:** Code Cleanup and Refactoring.
+    - [x] Sub-task: Remove any dead code or temporary debug statements.
+    - [x] Sub-task: Ensure code follows project style guides.
+- [x] **Task:** Conductor - User Manual Verification 'Finalization and Verification' (Protocol in workflow.md)
+
+---
+**Note:** This track is being closed. Remaining UI display issues for text panels in the comms and tool call history will be addressed in a subsequent track. Please see the project's issue tracker for details on the new track.

conductor/archive/gui2_feature_parity_20260223/spec.md

@@ -0,0 +1,45 @@
+# Specification: GUIv2 Feature Parity
+
+## 1. Overview
+
+This track aims to bring `gui_2.py` (the `imgui-bundle` based UI) to feature parity with the existing `gui.py` (the `dearpygui` based UI). This involves porting several major systems and features to ensure `gui_2.py` can serve as a viable replacement and support the latest project capabilities like automated testing and advanced diagnostics.
+
+## 2. Functional Requirements
+
+### FR1: Port Core Architectural Systems
+- **FR1.1: Event-Driven Architecture:** `gui_2.py` MUST be refactored to use the `events.py` module for handling API lifecycle events, decoupling the UI from the AI client.
+- **FR1.2: MCP File Tools Integration:** `gui_2.py` MUST integrate and use `mcp_client.py` to provide the AI with native, sandboxed file system capabilities (read, list, search).
+
+### FR2: Port Major Features
+- **FR2.1: API Hooks System:** The full API hooks system, including `api_hooks.py` and `api_hook_client.py`, MUST be integrated into `gui_2.py`. This will enable external test automation and state inspection.
+- **FR2.2: Performance & Diagnostics:** The performance monitoring capabilities from `performance_monitor.py` MUST be integrated. A new "Diagnostics" panel, mirroring the one in `gui.py`, MUST be created to display real-time metrics (FPS, CPU, Frame Time, etc.).
+- **FR2.3: Prior Session Viewer:** The functionality to load and view previous session logs (`.log` files from the `/logs` directory) MUST be implemented, including the distinctive "tinted" UI theme when viewing a prior session.
+
+### FR3: UI/UX Alignment
+- **FR3.1: 'Hub' UI Layout:** The windowing layout of `gui_2.py` MUST be refactored to match the "Hub" paradigm of `gui.py`. This includes creating:
+    - `Context Hub`
+    - `AI Settings Hub`
+    - `Discussion Hub`
+    - `Operations Hub`
+- **FR3.2: Agent Capability Toggles:** The UI MUST include checkboxes or similar controls to allow the user to enable or disable the AI's agent-level tools (e.g., `run_powershell`, `read_file`).
+- **FR3.3: Full Theme Integration:** All new UI components, windows, and controls MUST correctly apply and respond to the application's theming system (`theme_2.py`).
+
+## 3. Non-Functional Requirements
+
+- **NFR1: Stability:** The application must remain stable and responsive during and after the feature porting.
+- **NFR2: Maintainability:** The new code should follow existing project conventions and be well-structured to ensure maintainability.
+
+## 4. Acceptance Criteria
+
+- **AC1:** `gui_2.py` successfully launches without errors.
+- **AC2:** The "Hub" layout is present and organizes the UI elements as specified.
+- **AC3:** The Diagnostics panel is present and displays updating performance metrics.
+- **AC4:** The API hooks server starts and is reachable when `gui_2.py` is run with the appropriate flag.
+- **AC5:** The AI can successfully use file system tools provided by `mcp_client.py`.
+- **AC6:** The "Prior Session Viewer" can successfully load and display a log file.
+- **AC7:** All new UI elements correctly reflect the selected theme.
+
+## 5. Out of Scope
+
+- Deprecating or removing `gui.py`. Both will coexist for now.
+- Any new features not already present in `gui.py`. This is strictly a porting and alignment task.

conductor/archive/gui2_parity_20260224/index.md

@@ -0,0 +1,5 @@
+# Track gui2_parity_20260224 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/gui2_parity_20260224/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "gui2_parity_20260224",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-24T18:38:00Z",
+  "updated_at": "2026-02-24T18:38:00Z",
+  "description": "Investigate differences left between gui.py and gui_2.py. Needs to reach full parity, so we can sunset guy.py"
+}

conductor/archive/gui2_parity_20260224/plan.md

@@ -0,0 +1,43 @@
+# Implementation Plan: GUI 2.0 Feature Parity and Migration
+
+This plan follows the project's standard task workflow to ensure full feature parity and a stable transition to the ImGui-based `gui_2.py`.
+
+## Phase 1: Research and Gap Analysis [checkpoint: 36988cb]
+Identify and document the exact differences between `gui.py` and `gui_2.py`.
+
+- [x] Task: Audit `gui.py` and `gui_2.py` side-by-side to document specific visual and functional gaps. [fe33822]
+- [x] Task: Map existing `EventEmitter` and `ApiHookClient` integrations in `gui.py` to `gui_2.py`. [579b004]
+- [x] Task: Write failing tests in `tests/test_gui2_parity.py` that identify missing UI components or broken hooks in `gui_2.py`. [7c51674]
+- [x] Task: Verify failing parity tests. [0006f72]
+- [x] Task: Conductor - User Manual Verification 'Phase 1: Research and Gap Analysis' (Protocol in workflow.md) [9f99b77]
+
+## Phase 2: Visual and Functional Parity Implementation [checkpoint: ad84843]
+Address all identified gaps and ensure functional equivalence.
+
+- [x] Task: Implement missing panels and UX nuances (text sizing, font rendering) in `gui_2.py`. [a85293f]
+- [x] Task: Complete integration of all `EventEmitter` hooks in `gui_2.py` to match `gui.py`. [9d59a45]
+- [x] Task: Verify functional parity by running `tests/test_gui2_events.py` and `tests/test_gui2_layout.py`. [450820e]
+- [x] Task: Address any identified regressions or missing interactive elements. [2d8ee64]
+- [x] Task: Conductor - User Manual Verification 'Phase 2: Visual and Functional Parity Implementation' (Protocol in workflow.md) [ad84843]
+
+## Phase 3: Performance Optimization and Final Validation [checkpoint: 611c897]
+Ensure `gui_2.py` meets performance requirements and passes all quality gates.
+
+- [x] Task: Conduct performance benchmarking (FPS, CPU, Frame Time) for both `gui.py` and `gui_2.py`. [312b0ef]
+- [x] Task: Optimize rendering and docking logic in `gui_2.py` if performance targets are not met. [d647251]
+- [x] Task: Verify performance parity using `tests/test_gui2_performance.py`. [d647251]
+- [x] Task: Run full suite of automated GUI tests with `live_gui` fixture on `gui_2.py`. [d647251]
+- [x] Task: Conductor - User Manual Verification 'Phase 3: Performance Optimization and Final Validation' (Protocol in workflow.md) [14984c5]
+
+## Phase 4: Deprecation and Cleanup
+Finalize the migration and decommission the original `gui.py`.
+
+- [x] Task: Rename gui.py to gui_legacy.py. [c4c47b8]
+- [x] Task: Update project entry point or documentation to point to `gui_2.py` as the primary interface. [b92fa90]
+- [x] Task: Final project-wide link validation and documentation update. [14984c5]
+- [x] Task: Conductor - User Manual Verification 'Phase 4: Deprecation and Cleanup' (Protocol in workflow.md) [14984c5]
+
+## Phase: Review Fixes
+- [x] Task: Apply review suggestions [6f1e00b]
+---
+[checkpoint: 6f1e00b]

conductor/archive/gui2_parity_20260224/spec.md

@@ -0,0 +1,29 @@
+# Specification: GUI 2.0 Feature Parity and Migration
+
+## Overview
+The project is transitioning from `gui.py` (Dear PyGui-based) to `gui_2.py` (ImGui Bundle-based) to leverage advanced multi-viewport and docking features not natively supported by Dear PyGui. This track focuses on achieving full visual, functional, and performance parity between the two implementations, ultimately enabling the decommissioning of the original `gui.py`.
+
+## Functional Requirements
+1.  **Visual Parity:**
+    -   Ensure all panels, layouts, and interactive elements in `gui_2.py` match the established UX of `gui.py`.
+    -   Address nuances in UX, such as text panel sizing and font rendering, to ensure a seamless transition for existing users.
+2.  **Functional Parity:**
+    -   Verify that all backend hooks (API metrics, context management, MCP tools, shell execution) work identically in `gui_2.py`.
+    -   Ensure all interactive controls (buttons, inputs, dropdowns) trigger the correct application state changes.
+3.  **Performance Parity:**
+    -   Benchmark `gui_2.py` against `gui.py` for FPS, frame time, and CPU/memory usage.
+    -   Optimize `gui_2.py` to meet or exceed the performance metrics of the original implementation.
+
+## Non-Functional Requirements
+-   **Multi-Viewport Stability:** Ensure the ImGui-bundle implementation is stable across multiple windows and docking configurations.
+-   **Deprecation Workflow:** Establish a clear path for renaming `gui.py` to `gui_legacy.py` for a transition period.
+
+## Acceptance Criteria
+-   [ ] `gui_2.py` successfully passes the full suite of GUI automated verification tests (e.g., `test_gui2_events.py`, `test_gui2_layout.py`).
+-   [ ] A side-by-side audit confirms visual and functional parity for all core Hub panels.
+-   [ ] Performance benchmarks show `gui_2.py` is within +/- 5% of `gui.py` metrics.
+-   [ ] `gui.py` is renamed to `gui_legacy.py`.
+
+## Out of Scope
+-   Introducing new UI features or backend capabilities not present in `gui.py`.
+-   Modifying the core `EventEmitter` or `AiClient` logic (unless required for GUI hook integration).

conductor/archive/gui_sim_extension_20260224/index.md

@@ -0,0 +1,5 @@
+# Track gui_sim_extension_20260224 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/gui_sim_extension_20260224/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "gui_sim_extension_20260224",
+  "type": "chore",
+  "status": "new",
+  "created_at": "2026-02-24T19:17:00Z",
+  "updated_at": "2026-02-24T19:17:00Z",
+  "description": "extend test simulation to have further in breadth test (not remove the original though as its a useful small test) to extensively test all facets of possible gui interaction."
+}

conductor/archive/gui_sim_extension_20260224/plan.md

@@ -0,0 +1,39 @@
+# Implementation Plan: Extended GUI Simulation Testing
+
+## Phase 1: Setup and Architecture [checkpoint: b255d4b]
+- [x] Task: Review the existing baseline simulation test to identify reusable components or fixtures without modifying the original. a0b1c2d
+- [x] Task: Design the modular structure for the new simulation scripts within the `simulation/` directory. e1f2g3h
+- [x] Task: Create a base test configuration or fixture that initializes the GUI with the `--enable-test-hooks` flag and the `ApiHookClient` for API testing. i4j5k6l
+- [x] Task: Conductor - User Manual Verification 'Phase 1: Setup and Architecture' (Protocol in workflow.md) m7n8o9p
+
+## Phase 2: Context and Chat Simulation [checkpoint: a77d0e7]
+- [x] Task: Create the test script `sim_context.py` focused on the Context and Discussion panels. q1r2s3t
+- [x] Task: Simulate file aggregation interactions and context limit verification. u4v5w6x
+- [x] Task: Implement history generation and test chat submission via API hooks. y7z8a9b
+- [x] Task: Conductor - User Manual Verification 'Phase 2: Context and Chat Simulation' (Protocol in workflow.md) c1d2e3f
+
+## Phase 3: AI Settings and Tools Simulation [checkpoint: 760eec2]
+- [x] Task: Create the test script `sim_ai_settings.py` for AI model configuration changes (Gemini/Anthropic). g1h2i3j
+- [x] Task: Create the test script `sim_tools.py` focusing on file exploration, search, and MCP-like tool triggers. k4l5m6n
+- [x] Task: Validate proper panel rendering and data updates via API hooks for both AI settings and tool results. o7p8q9r
+- [x] Task: Conductor - User Manual Verification 'Phase 3: AI Settings and Tools Simulation' (Protocol in workflow.md) s1t2u3v
+
+## Phase 4: Execution and Modals Simulation [checkpoint: e8959bf]
+- [x] Task: Create the test script `sim_execution.py`. w3x4y5z
+- [x] Task: Simulate the AI generating a PowerShell script that triggers the explicit confirmation modal. a1b2c3d
+- [x] Task: Assert the modal appears correctly and accepts input/approval from the simulated user. e4f5g6h
+- [x] Task: Validate the executed output via API hooks. i7j8k9l
+- [x] Task: Conductor - User Manual Verification 'Phase 4: Execution and Modals Simulation' (Protocol in workflow.md) m0n1o2p
+
+## Phase 5: Reactive Interaction and Final Polish [checkpoint: final]
+- [x] Task: Implement reactive `/api/events` endpoint for real-time GUI feedback. x1y2z3a
+- [x] Task: Add auto-scroll and fading blink effects to Tool and Comms history panels. b4c5d6e
+- [x] Task: Restrict simulation testing to `gui_2.py` and ensure full integration pass. f7g8h9i
+- [x] Task: Conductor - User Manual Verification 'Phase 5: Reactive Interaction and Final Polish' (Protocol in workflow.md) j0k1l2m
+
+## Phase 6: Multi-Turn & Stability Polish [checkpoint: pass]
+- [x] Task: Implement looping reactive simulation for multi-turn tool approvals. a1b2c3d
+- [x] Task: Fix Gemini 400 error by adding token threshold for context caching. e4f5g6h
+- [x] Task: Ensure `btn_reset` clears all relevant UI fields including `ai_input`. i7j8k9l
+- [x] Task: Run full test suite (70+ tests) and ensure 100% pass rate. m0n1o2p
+- [x] Task: Conductor - User Manual Verification 'Phase 6: Multi-Turn & Stability Polish' (Protocol in workflow.md) q1r2s3t

conductor/archive/gui_sim_extension_20260224/spec.md

@@ -0,0 +1,27 @@
+# Specification: Extended GUI Simulation Testing
+
+## Overview
+This track aims to expand the test simulation suite by introducing comprehensive, in-breadth tests that cover all facets of the GUI interaction. The original small test simulation will be preserved as a useful baseline. The new extended tests will be structured as multiple focused, modular scripts rather than a single long-running journey, ensuring maintainability and targeted coverage.
+
+## Scope
+The extended simulation tests will cover the following key GUI workflows and panels:
+- **Context & Chat:** Testing the core Context and Discussion panels, including history management and context aggregation.
+- **AI Settings:** Validating AI settings manipulation, model switching, and provider changes (Gemini/Anthropic).
+- **Tools & Search:** Exercising file exploration, MCP-like file tools, and web search capabilities.
+- **Execution & Modals:** Testing the generation, explicit confirmation via modals, and execution of PowerShell scripts.
+
+## Functional Requirements
+1. **Modular Test Architecture:** Implement a suite of independent simulation scripts under the `simulation/` or `tests/` directory (e.g., `sim_context.py`, `sim_tools.py`, `sim_execution.py`).
+2. **Preserve Baseline:** Ensure the existing small test simulation remains functional and untouched.
+3. **Comprehensive Coverage:** Each modular script must focus on a specific, complex interaction workflow, simulating human-like usage via the existing IPC/API hooks mechanism.
+4. **Validation and Checkpointing:** Each script must include assertions to verify the GUI state, confirming that the expected panels are rendered, inputs are accepted, and actions produce the correct results.
+
+## Non-Functional Requirements
+- **Maintainability:** The modular design should make it easy to add or update specific workflows in the future.
+- **Performance:** Tests should run reliably without causing the GUI framework to lock up, utilizing the event-driven architecture properly.
+
+## Acceptance Criteria
+- [ ] A new suite of modular simulation scripts is created.
+- [ ] The existing test simulation is untouched and remains functional.
+- [ ] The new tests run successfully and pass all verifications via the automated API hook mechanism.
+- [ ] The scripts cover all four major GUI areas identified in the scope.

conductor/archive/history_segregation_20260224/index.md

@@ -0,0 +1,5 @@
+# Track history_segregation_20260224 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/history_segregation_20260224/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "history_segregation_20260224",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-24T18:28:00Z",
+  "updated_at": "2026-02-24T18:28:00Z",
+  "description": "Move discussion histories to their own toml to prevent the ai agent from reading it (will be on a blacklist)."
+}

conductor/archive/history_segregation_20260224/plan.md

@@ -0,0 +1,33 @@
+# Implementation Plan: Discussion History Segregation and Blacklisting
+
+This plan follows the Test-Driven Development (TDD) workflow to move discussion history into a dedicated sibling TOML file and enforce a strict blacklist against AI agent tool access.
+
+## Phase 1: Foundation and Migration Logic
+This phase focuses on the structural changes needed to handle dual-file project configurations and the automatic migration of legacy history.
+
+- [x] Task: Research existing `ProjectManager` serialization and tool access points in `mcp_client.py`. (f400799)
+- [x] Task: Write TDD tests for migrating the `discussion` key from `manual_slop.toml` to a new sibling file. (7c18e11)
+- [x] Task: Implement automatic migration in `ProjectManager.load_project()`. (7c18e11)
+- [x] Task: Update `ProjectManager.save_project()` to persist history separately. (7c18e11)
+- [x] Task: Verify that existing history is correctly migrated and remains visible in the GUI. (ba02c8e)
+- [x] Task: Conductor - User Manual Verification 'Foundation and Migration' (Protocol in workflow.md)
+
+## Phase 2: Blacklist Enforcement
+This phase ensures the AI agent is strictly prevented from reading the history source files through its tools.
+
+- [x] Task: Write failing tests that attempt to read a known history file via the `mcp_client.py` and `aggregate.py` logic. (77f3e22)
+- [x] Task: Implement hardcoded exclusion for `*_history.toml` and `history.toml` in `mcp_client.py`. (77f3e22)
+- [x] Task: Implement hardcoded exclusion in `aggregate.py` to prevent history from being added as a raw file context. (77f3e22)
+- [x] Task: Verify that tool-based file reads for the history file return a "Permission Denied" or "Blacklisted" error. (77f3e22)
+- [x] Task: Conductor - User Manual Verification 'Blacklist Enforcement' (Protocol in workflow.md)
+
+## Phase 3: Integration and Final Validation
+This phase validates the full lifecycle, ensuring the application remains functional and secure.
+
+- [x] Task: Conduct a full walkthrough using the simulation scripts to verify history persistence across turns. (754fbe5)
+- [x] Task: Verify that the AI can still use the *curated* history provided in the prompt context but cannot access the raw file. (754fbe5)
+- [x] Task: Run full suite of automated GUI and API hook tests. (754fbe5)
+- [x] Task: Conductor - User Manual Verification 'Integration and Final Validation' (Protocol in workflow.md) [checkpoint: 754fbe5]
+
+## Phase: Review Fixes
+- [x] Task: Apply review suggestions (docstrings, annotations, import placement) (09df57d)

conductor/archive/history_segregation_20260224/spec.md

@@ -0,0 +1,32 @@
+# Specification: Discussion History Segregation and Blacklisting
+
+## Overview
+Currently, `manual_slop.toml` stores both project configuration and the entire discussion history. This leads to redundancy and potential context bloat if the AI agent reads the raw TOML file via its tools. This track will move the discussion history to a dedicated sibling TOML file (`history.toml`) and strictly blacklist it from the AI agent's file tools to ensure it only interacts with the curated context provided in the prompt.
+
+## Functional Requirements
+1.  **File Segregation:**
+    -   Create a dedicated history file (e.g., `manual_slop_history.toml`) in the same directory as the main project configuration.
+    -   The main `manual_slop.toml` will henceforth only store project settings, tracked files, and system prompts.
+2.  **Automatic Migration:**
+    -   On application startup or project load, detect if the `discussion` key exists in `manual_slop.toml`.
+    -   If found, automatically migrate all discussion entries to the new history sibling file and remove the key from the original file.
+3.  **Strict Blacklisting:**
+    -   Hardcode the exclusion of the history TOML file in `mcp_client.py` and `aggregate.py`.
+    -   The AI agent must be prevented from reading this file using the `read_file` or `search_files` tools.
+4.  **Backend Integration:**
+    -   Update `ProjectManager` in `project_manager.py` to manage two distinct TOML files per project.
+    -   Ensure the GUI correctly loads history from the new file while maintaining existing functionality.
+
+## Non-Functional Requirements
+-   **Data Integrity:** Ensure no history is lost during the migration process.
+-   **Performance:** Minimize I/O overhead when saving history entries after each AI turn.
+
+## Acceptance Criteria
+-   [ ] `manual_slop.toml` no longer contains the `discussion` array.
+-   [ ] A sibling `history.toml` (or similar) contains all historical and new discussion entries.
+-   [ ] The AI agent cannot access the history TOML file via its file tools (verification via tool call test).
+-   [ ] Discussion history remains visible in the GUI and is correctly included in the AI prompt context.
+
+## Out of Scope
+-   Customizable blacklist via the UI.
+-   Support for cloud-based history storage.

conductor/archive/live_ux_test_20260223/plan.md

@@ -0,0 +1,40 @@
+# Implementation Plan: Human-Like UX Interaction Test
+
+## Phase 1: Infrastructure & Automation Core [checkpoint: 7626531]
+Establish the foundation for driving the GUI via API hooks and simulation logic.
+
+- [x] Task: Extend `ApiHookClient` with methods for tab switching and listbox selection if missing. f36d539
+- [x] Task: Implement `TestUserAgent` class to manage dynamic response generation and action delays. d326242
+- [x] Task: Write Tests (Verify basic hook connectivity and simulated delays) f36d539
+- [x] Task: Implement basic 'ping-pong' interaction via hooks. bfe9ef0
+- [x] Task: Harden API hook thread-safety and simplify GUI state polling. 8bd280e
+- [x] Task: Conductor - User Manual Verification 'Phase 1: Infrastructure & Automation Core' (Protocol in workflow.md) 7626531
+
+## Phase 2: Workflow Simulation [checkpoint: 9c4a72c]
+Build the core interaction loop for project creation and AI discussion.
+
+- [x] Task: Implement 'New Project' scaffolding script (creating a tiny console program). bd5dc16
+- [x] Task: Implement 5-turn discussion loop logic with sub-agent responses. bd5dc16
+- [x] Task: Write Tests (Verify state changes in Discussion Hub during simulated chat) 6d16438
+- [x] Task: Implement 'Thinking' and 'Live' indicator verification logic. 6d16438
+- [x] Task: Conductor - User Manual Verification 'Phase 2: Workflow Simulation' (Protocol in workflow.md) 9c4a72c
+
+## Phase 3: History & Session Verification [checkpoint: 0f04e06]
+Simulate complex session management and historical audit features.
+
+- [x] Task: Implement discussion switching logic (creating/switching between named discussions). 5e1b965
+- [x] Task: Implement 'Load Prior Log' simulation and 'Tinted Mode' detection. 5e1b965
+- [x] Task: Write Tests (Verify log loading and tab navigation consistency) 5e1b965
+- [x] Task: Implement truncation limit verification (forcing a long history and checking bleed). 5e1b965
+- [x] Task: Conductor - User Manual Verification 'Phase 3: History & Session Verification' (Protocol in workflow.md) 0f04e06
+
+## Phase 4: Final Integration & Regression [checkpoint: 8e63b31]
+Consolidate the simulation into end-user artifacts and CI tests.
+
+- [x] Task: Create `live_walkthrough.py` with full visual feedback and manual sign-off. 8bd280e
+- [x] Task: Create `tests/test_live_workflow.py` for automated regression testing. 8bd280e
+- [x] Task: Perform a full visual walkthrough and verify 'human-readable' pace. 8e63b31
+- [x] Task: Conductor - User Manual Verification 'Phase 4: Final Integration & Regression' (Protocol in workflow.md) 8e63b31
+
+## Phase: Review Fixes
+- [x] Task: Apply review suggestions 064d7ba

conductor/archive/logging_refactor_20260226/index.md

@@ -0,0 +1,5 @@
+# Track logging_refactor_20260226 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/logging_refactor_20260226/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "logging_refactor_20260226",
+  "type": "chore",
+  "status": "new",
+  "created_at": "2026-02-26T08:45:00Z",
+  "updated_at": "2026-02-26T08:45:00Z",
+  "description": "Review logging used throughout the project. The log directory has several categories of logs and they are getting quite large in number. We need sub-directories and we need a way to prune logs that aren't valuable to keep."
+}

conductor/archive/logging_refactor_20260226/plan.md

@@ -0,0 +1,39 @@
+# Implementation Plan: Logging Reorganization and Automated Pruning
+
+## Phase 1: Session Organization & Registry Foundation
+- [x] Task: Initialize MMA Environment (Protocol: `activate_skill mma-orchestrator`) [9a66b76]
+- [x] Task: Implement `LogRegistry` to manage `log_registry.toml` [10fbfd0]
+    - [x] Define TOML schema for session metadata.
+    - [x] Create methods to register sessions and update whitelist status.
+- [x] Task: Implement Session-Based Directory Creation [3f4dc1a]
+    - [x] Create utility to generate Session IDs: `YYYYMMDD_HHMMSS[_Label]`.
+    - [x] Update logging initialization to create and use session sub-directories.
+- [x] Task: Conductor - User Manual Verification 'Phase 1: Foundation' (Protocol in workflow.md) [3f4dc1a]
+
+## Phase 2: Pruning Logic & Heuristics
+- [x] Task: Implement `LogPruner` Core Logic [bd2a79c]
+    - [x] Implement time-based filtering (older than 24h).
+    - [x] Implement size-based heuristic for "insignificance" (~2 KB).
+- [x] Task: Implement Auto-Whitelisting Heuristics [4e9c47f]
+    - [x] Implement content scanning for `ERROR`, `WARNING`, `EXCEPTION`.
+    - [x] Implement complexity detection (message count > 10).
+- [x] Task: Integrate Pruning into App Startup [8b75883]
+    - [x] Hook the pruner into `gui_2.py` startup sequence.
+    - [x] Ensure pruning runs asynchronously to prevent startup lag.
+- [x] Task: Conductor - User Manual Verification 'Phase 2: Pruning' (Protocol in workflow.md) [8b75883]
+
+## Phase 3: GUI Integration & Manual Control
+- [x] Task: Add "Log Management" UI Panel [7d52123]
+    - [x] Display a list of recent sessions from the registry.
+    - [x] Add "Star/Unstar" toggle for manual whitelisting.
+- [x] Task: Display Session Metrics in UI [7d52123]
+    - [x] Show size, message count, and status (Whitelisted/Pending Prune).
+- [x] Task: Conductor - User Manual Verification 'Phase 3: GUI' (Protocol in workflow.md) [7d52123]
+
+## Phase 4: Final Verification & Cleanup
+- [x] Task: Comprehensive Integration Testing [23c0f0a]
+    - [x] Verify that empty old logs are deleted.
+    - [x] Verify that complex/error-filled old logs are preserved.
+- [x] Task: Final Refactoring and Documentation [04a991e]
+    - [x] Ensure all new classes and methods follow project style.
+- [x] Task: Conductor - User Manual Verification 'Phase 4: Final' (Protocol in workflow.md) [04a991e]

conductor/archive/logging_refactor_20260226/spec.md

@@ -0,0 +1,42 @@
+# Specification: Logging Reorganization and Automated Pruning
+
+## Overview
+Currently, `gui_2.py` and the test suites generate a large number of log files in a flat `logs/` directory. These logs accumulate quickly, especially during incremental development and testing. This track aims to organize logs into session-based sub-directories and implement a heuristic-based pruning system to keep the log directory clean while preserving valuable sessions.
+
+## Functional Requirements
+1. **Session-Based Organization:**
+   - Logs must be stored in sub-directories within `logs/`.
+   - Sub-directory naming convention: `YYYYMMDD_HHMMSS[_Label]` (e.g., `20260226_143005_feature_x`).
+   - The "Label" should be included if a project or track is active at session start.
+2. **Central Registry:**
+   - A `logs/log_registry.toml` file will track session metadata, including:
+     - Session ID / Path
+     - Start Time
+     - Whitelist Status (Manual/Auto)
+     - Metrics (message count, errors detected, total size).
+3. **Automated Pruning Heuristic:**
+   - Pruning triggers on application startup (`gui_2.py`).
+   - **Target:** Logs older than 24 hours.
+   - **Exemption:** Whitelisted logs are never auto-pruned.
+   - **Insignificance Criteria:** Non-whitelisted logs under a specific size threshold (heuristic: ~2 KB) or with zero significant interactions will be purged.
+4. **Whitelisting System:**
+   - **Auto-Whitelisting:** Sessions are marked as "rich" if they meet any of these:
+     - Complexity: > 10 messages/interactions.
+     - Diagnostics: Contains `ERROR`, `WARNING`, `EXCEPTION`.
+     - Major Events: User created a new project or initialized a track.
+   - **Manual Whitelisting:** The user can "star" a session via the GUI (persisted in the registry).
+
+## Non-Functional Requirements
+- **Performance:** Pruning and registry updates must be asynchronous or extremely fast to avoid delaying app startup.
+- **Safety:** Ensure the pruning logic is conservative to prevent accidental data loss of important debug information.
+
+## Acceptance Criteria
+- [ ] New logs are created in session-specific folders.
+- [ ] The `log_registry.toml` correctly identifies and tracks sessions.
+- [ ] On startup, non-whitelisted logs older than 1 day are successfully pruned.
+- [ ] Whitelisted logs (due to complexity or errors) remain untouched.
+- [ ] (Bonus) The GUI displays a basic list of sessions with their "starred" status.
+
+## Out of Scope
+- Migrating the entire backlog of existing flat logs (focus is on new sessions).
+- Implementing a full-blown log viewer (basic metadata view only).

conductor/archive/manual_slop_headless_20260225/index.md

@@ -0,0 +1,5 @@
+# Track manual_slop_headless_20260225 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/manual_slop_headless_20260225/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "manual_slop_headless_20260225",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-25T12:00:00Z",
+  "updated_at": "2026-02-25T12:00:00Z",
+  "description": "Support headless manual_slop for making an unraid gui docker frontend and a unraid server backend down the line."
+}

conductor/archive/manual_slop_headless_20260225/plan.md

@@ -0,0 +1,52 @@
+# Implementation Plan: Manual Slop Headless Backend
+
+## Phase 1: Project Setup & Headless Scaffold [checkpoint: d5f056c]
+- [x] Task: Update dependencies (02fc847)
+    - [x] Add `fastapi` and `uvicorn` to `pyproject.toml` (and sync `requirements.txt` via `uv`).
+- [x] Task: Implement headless startup
+    - [x] Modify `gui_2.py` (or create `headless.py`) to parse a `--headless` CLI flag.
+    - [x] Update config parsing in `config.toml` to support headless configuration sections.
+    - [x] Bypass Dear PyGui initialization if headless mode is active.
+- [x] Task: Create foundational API application
+    - [x] Set up the core FastAPI application instance.
+    - [x] Implement `/health` and `/status` endpoints for Docker lifecycle checks.
+- [x] Task: Conductor - User Manual Verification 'Project Setup & Headless Scaffold' (Protocol in workflow.md) d5f056c
+
+## Phase 2: Core API Routes & Authentication [checkpoint: 4e0bcd5]
+- [x] Task: Implement API Key Security
+    - [x] Create a dependency/middleware in FastAPI to validate `X-API-KEY`.
+    - [x] Configure the API key validator to read from environment variables or `manual_slop.toml` (supporting Unraid template secrets).
+    - [x] Add tests for authorized and unauthorized API access.
+- [x] Task: Implement AI Generation Endpoint
+    - [x] Create a `/api/v1/generate` POST endpoint.
+    - [x] Map request payloads to `ai_client.py` unified wrappers.
+    - [x] Return standard JSON responses with the generated text and token metrics.
+- [x] Task: Conductor - User Manual Verification 'Core API Routes & Authentication' (Protocol in workflow.md) 4e0bcd5
+
+## Phase 3: Remote Tool Confirmation Mechanism [checkpoint: a6e184e]
+- [x] Task: Refactor Execution Engine for Async Wait
+    - [x] Modify `shell_runner.py` and tool-call loops to support a non-blocking "Pending Confirmation" state instead of launching a GUI modal.
+- [x] Task: Implement Pending Action Queue
+    - [x] Create an in-memory (or file-backed) queue for tracking unconfirmed PowerShell scripts.
+- [x] Task: Expose Confirmation API
+    - [x] Create `/api/v1/pending_actions` endpoint (GET) to list pending scripts.
+    - [x] Create `/api/v1/confirm/{action_id}` endpoint (POST) to approve or deny a script execution.
+    - [x] Ensure the AI generation loop correctly resumes upon receiving approval.
+- [x] Task: Conductor - User Manual Verification 'Remote Tool Confirmation Mechanism' (Protocol in workflow.md) a6e184e
+
+## Phase 4: Session & Context Management via API [checkpoint: 7f3a1e2]
+- [x] Task: Expose Session History
+    - [x] Create endpoints to list, retrieve, and delete session logs from the `project_history.toml`.
+- [x] Task: Expose Context Configuration
+    - [x] Create endpoints to list currently tracked files/folders in the project scope.
+- [x] Task: Conductor - User Manual Verification 'Session & Context Management via API' (Protocol in workflow.md) 7f3a1e2
+
+## Phase 5: Dockerization [checkpoint: 5176b8d]
+- [x] Task: Create Dockerfile
+    - [x] Write a `Dockerfile` using `python:3.11-slim` as a base.
+    - [x] Configure `uv` inside the container for fast dependency installation.
+    - [x] Expose the API port (e.g., 8000) and set the container entrypoint.
+- [x] Task: Conductor - User Manual Verification 'Dockerization' (Protocol in workflow.md) 5176b8d
+
+## Phase: Review Fixes
+- [x] Task: Apply review suggestions (docstrings and security fix) 9b50bfa

conductor/archive/manual_slop_headless_20260225/spec.md

@@ -0,0 +1,48 @@
+# Specification: Manual Slop Headless Backend
+
+## Overview
+Transform Manual Slop into a decoupled, container-friendly backend service. This track enables the core AI orchestration and tool execution logic to run without a GUI, exposing its capabilities via a secured REST API optimized for an Unraid Docker environment.
+
+## Goals
+- Decouple the GUI logic (`Dear PyGui`, `ImGui`) from the core AI and Tool logic.
+- Implement a lightweight REST API server (FastAPI) to handle AI interactions.
+- Ensure full compatibility with Unraid Docker networking and configuration patterns.
+- Maintain the "Human-in-the-Loop" safety model through a remote confirmation mechanism.
+
+## Functional Requirements
+
+### 1. Headless Mode Lifecycle
+- **Startup**: Provide a `--headless` flag or `[headless]` section in `manual_slop.toml` to skip GUI initialization.
+- **Dependencies**: Ensure the app can start in environments without an X11/Wayland display or GPU.
+- **Service Mode**: Support running as a persistent background daemon/service.
+
+### 2. REST API (FastAPI)
+- **Status/Health**: `/status` and `/health` endpoints for Docker/Unraid monitoring.
+- **AI Interface**: `/generate` and `/stream` endpoints to interact with configured AI providers.
+- **Tool Management**: Endpoints to list and execute tools (PowerShell/MCP).
+- **Session Support**: Manage conversation history and project context via API.
+
+### 3. Security & Authentication
+- **API Key**: Require a `X-API-KEY` header for all sensitive endpoints.
+- **Unraid Integration**: API keys should be configurable via Environment Variables (standard for Unraid templates).
+
+### 4. Remote Confirmation Mechanism
+- **Challenge/Response**: When a tool requires execution, the API should return a "Pending Confirmation" state.
+- **Webhook/Poll**: Support a mechanism (e.g., a `/confirm/{id}` endpoint) for the future frontend to approve/deny actions.
+
+## Non-Functional Requirements
+- **Performance**: Headless mode should use significantly less memory/CPU than the GUI version.
+- **Logging**: Use standard Python `logging` for Docker-compatible stdout/stderr output.
+- **Portability**: Must run reliably inside a standard `python:3.11-slim` or similar Docker image.
+
+## Acceptance Criteria
+- [ ] Manual Slop starts successfully with `--headless` and no display environment.
+- [ ] API is accessible via a configurable port (e.g., 8000).
+- [ ] All API requests are rejected without a valid API Key.
+- [ ] AI generation works via REST endpoints, returning structured JSON or a stream.
+- [ ] Tool execution is successfully blocked until a separate "Confirm" API call is made.
+
+## Out of Scope
+- Building the actual Unraid GUI frontend (React/Vue/etc.).
+- Multi-user authentication (OIDC/OAuth2).
+- Native Unraid `.plg` plugin development (focusing on Docker).

conductor/archive/mma_core_engine_20260224/index.md

@@ -0,0 +1,9 @@
+# MMA Core Engine Implementation
+
+This track implements the 5 Core Epics defined during the MMA Architecture Evaluation.
+
+### Navigation
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Original Architecture Proposal / Meta-Track](../mma_implementation_20260224/index.md)
+- [MMA Support Directory (Source of Truth)](../../../MMA_Support/)

conductor/archive/mma_core_engine_20260224/metadata.json

@@ -0,0 +1,6 @@
+{
+  "id": "mma_core_engine_20260224",
+  "title": "MMA Core Engine Implementation",
+  "status": "planning",
+  "created_at": "2026-02-24T00:00:00.000000"
+}

conductor/archive/mma_core_engine_20260224/plan.md

@@ -0,0 +1,85 @@
+# Implementation Plan: MMA Core Engine Implementation
+
+## Phase 1: Track 1 - The Memory Foundations (AST Parser) [checkpoint: ac31e41]
+- [x] Task: Dependency Setup (8fb75cc)
+    - [x] Add `tree-sitter` and `tree-sitter-python` to `pyproject.toml` / `requirements.txt` (8fb75cc)
+- [x] Task: Core Parser Class (7a609ca)
+    - [x] Create `ASTParser` in `file_cache.py` (7a609ca)
+- [x] Task: Skeleton View Extraction (7a609ca)
+    - [x] Write query to extract `function_definition` and `class_definition` (7a609ca)
+    - [x] Replace bodies with `pass`, keep type hints and signatures (7a609ca)
+- [x] Task: Curated View Extraction (7a609ca)
+    - [x] Keep class structures, module docstrings (7a609ca)
+    - [x] Preserve `@core_logic` or `# [HOT]` function bodies, hide others (7a609ca)
+
+## Phase 2: Track 2 - State Machine & Data Structures [checkpoint: a518a30]
+- [x] Task: The Dataclasses (f9b5a50)
+    - [x] Create `models.py` defining `Ticket` and `Track` (f9b5a50)
+- [x] Task: Worker Context Definition (ee71929)
+    - [x] Define `WorkerContext` holding `Ticket` ID, model config, and ephemeral messages (ee71929)
+- [x] Task: State Mutator Methods (e925b21)
+    - [x] Implement `ticket.mark_blocked()`, `ticket.mark_complete()`, `track.get_executable_tickets()` (e925b21)
+
+## Phase 3: Track 3 - The Linear Orchestrator & Execution Clutch [checkpoint: e6c8d73]
+- [x] Task: The Engine Core (7a30168)
+    - [x] Create `multi_agent_conductor.py` containing `ConductorEngine` and `run_worker_lifecycle` (7a30168)
+- [x] Task: Context Injection (9d6d174)
+    - [x] Format context strings using `file_cache.py` target AST views (9d6d174)
+- [x] Task: The HITL Execution Clutch (1afd9c8)
+    - [x] Before executing `write_file`/`shell_runner.py` tools in step-mode, prompt user for confirmation (1afd9c8)
+    - [x] Provide functionality to mutate the history JSON before resuming execution (1afd9c8)
+
+## Phase 4: Track 4 - Tier 4 QA Interception [checkpoint: 61d17ad]
+- [x] Task: The Interceptor Loop (bc654c2)
+    - [x] Catch `subprocess.run()` execution errors inside `shell_runner.py` (bc654c2)
+- [x] Task: Tier 4 Instantiation (8e4e326)
+    - [x] Make a secondary API call to `default_cheap` model passing `stderr` and snippet (8e4e326)
+- [x] Task: Payload Formatting (fb3da4d)
+    - [x] Inject the 20-word fix summary into the Tier 3 worker history (fb3da4d)
+
+## Phase 5: Track 5 - UI Decoupling & Tier 1/2 Routing (The Final Boss) [checkpoint: 3982fda]
+- [x] Task: The Event Bus (695cb4a)
+    - [x] Implement an `asyncio.Queue` linking GUI actions to the backend engine (695cb4a)
+- [x] Task: Tier 1 & 2 System Prompts (a28d71b)
+    - [x] Create structured system prompts for Epic routing and Ticket creation (a28d71b)
+- [x] Task: The Dispatcher Loop (1dacd36)
+    - [x] Read Tier 2 JSON flat-lists, construct Tickets, execute Stub resolution paths (1dacd36)
+- [x] Task: UI Component Update (68861c0)
+    - [x] Refactor `gui_2.py` to push `UserRequestEvent` instead of blocking on API generation (68861c0)
+
+## Phase 6: Live & Headless Verification
+- [x] Task: Headless Engine Verification
+    - [x] Run a comprehensive headless test scenario (e.g., using a mock or dedicated test script).
+    - [x] Verify Ticket execution, "Context Amnesia" (statelessness), and Tier 4 error interception.
+- [x] Task: Live GUI Integration Verification
+    - [x] Launch `gui_2.py` and verify Event Bus responsiveness.
+    - [x] Confirm UI updates and async event handling during multi-model generation.
+- [x] Task: Comprehensive Regression Suite
+    - [x] Run all tests in `tests/` related to MMA, Conductor, and Async Events.
+    - [x] Verify that no regressions were introduced in existing functionality.
+
+## Phase 7: MMA Observability & UX
+- [x] Task: MMA Dashboard Implementation
+    - [x] Create a new dockable panel in `gui_2.py` for "MMA Dashboard".
+    - [x] Display active `Track` and `Ticket` queue status.
+- [x] Task: Execution Clutch UI
+    - [x] Implement Step Mode toggle and Pause Points logic in the GUI.
+    - [x] Add `[Approve]`, `[Edit Payload]`, and `[Abort]` buttons for tool execution.
+- [x] Task: Memory Mutator Modal
+    - [x] Create a modal for editing raw JSON conversation history of paused workers.
+- [x] Task: Tiered Metrics & Log Links
+    - [x] Add visual indicators for the active model Tier.
+    - [x] Provide clickable links to sub-agent logs.
+
+## Phase 8: Visual Verification & Interaction Tests
+- [x] Task: Visual Verification Script
+    - [x] Create `tests/visual_mma_verification.py` to drive the GUI into various MMA states.
+    - [x] Verify MMA Dashboard visibility and progress bar.
+    - [x] Verify Ticket Queue rendering with correct status colors.
+- [x] Task: HITL Interaction Verification
+    - [x] Drive a simulated HITL pause through the verification script.
+    - [x] Manually verify the "MMA Step Approval" modal appearance.
+    - [x] Manually verify "Edit Payload" (Memory Mutator) functionality.
+- [~] Task: Final Polish & Fixes
+    - [ ] Fix any visual glitches or layout issues discovered during manual testing.
+    - [ ] Fix any visual glitches or layout issues discovered during manual testing.

conductor/archive/mma_core_engine_20260224/spec.md

@@ -0,0 +1,39 @@
+# Specification: MMA Core Engine Implementation
+
+## 1. Overview
+This track consolidates the implementation of the 4-Tier Hierarchical Multi-Model Architecture into the `manual_slop` codebase. The architecture transitions the current monolithic single-agent loop into a compartmentalized, token-efficient, and fully debuggable state machine.
+
+## 2. Functional Requirements
+
+### Phase 1: The Memory Foundations (AST Parser)
+- Integrate `tree-sitter` and `tree-sitter-python` into `pyproject.toml` / `requirements.txt`.
+- Implement `ASTParser` in `file_cache.py` to extract strict memory views (Skeleton View, Curated View).
+- Strip function bodies from dependencies while preserving `@core_logic` or `# [HOT]` logic for the target modules.
+
+### Phase 2: State Machine & Data Structures
+- Create `models.py` incorporating strict Pydantic/Dataclass schemas for `Ticket`, `Track`, and `WorkerContext`.
+- Enforce rigid state mutators governing dependencies between tickets (e.g., locking execution until a stub generation ticket completes).
+
+### Phase 3: The Linear Orchestrator & Execution Clutch
+- Build `multi_agent_conductor.py` and a `ConductorEngine` dispatcher loop.
+- Embed the "Execution Clutch" allowing developers to pause, review, and manually rewrite payloads (JSON history mutation) before applying changes to the local filesystem.
+
+### Phase 4: Tier 4 QA Interception
+- Augment `shell_runner.py` with try/except wrappers capturing process errors (`stderr`).
+- Rather than feeding raw stack traces to an expensive model, instantly forward them to a stateless `default_cheap` sub-agent for a 20-word summarization that is subsequently injected into the primary worker's context.
+
+### Phase 5: UI Decoupling & Tier 1/2 Routing (The Final Boss)
+- Disconnect `gui_2.py` from direct LLM inference requests.
+- Bind the GUI to a synchronous or `asyncio.Queue` Event Bus managed by the Orchestrator, allowing dynamic tracking of parallel worker executions without thread-locking the interface.
+
+## 3. Acceptance Criteria
+- [ ] A 1000-line script can be successfully parsed into a 100-line AST Skeleton.
+- [ ] Tickets properly block and resolve depending on stub-generation dependencies.
+- [ ] Shell errors are compressed into >50-token hints using the cheap utility model.
+- [ ] The GUI remains responsive during multi-model generation phases.
+
+## 4. Meta-Track Reference & Source of Truth
+For the original rationale, API formatting recommendations (e.g., Godot ECS schemas vs Nested JSON), and strict token firewall workflows, refer back to the architectural planning meta-track: `conductor/tracks/mma_implementation_20260224/`.
+
+**Fallback Source of Truth:**
+As a fallback, any track or sub-task should absolve its source of truth by referencing the `./MMA_Support/` directory. This directory contains the original design documents and raw discussions from which the entire `mma_implementation` track and 4-Tier Architecture were initially generated.

conductor/archive/mma_data_architecture_dag_engine/index.md

@@ -0,0 +1,7 @@
+# MMA Data Architecture & DAG Engine
+
+Restructures manual_slop state and execution into a per-track DAG model.
+
+### Navigation
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)

conductor/archive/mma_data_architecture_dag_engine/metadata.json

@@ -0,0 +1,6 @@
+{
+  "id": "mma_data_architecture_dag_engine",
+  "title": "MMA Data Architecture & DAG Engine",
+  "status": "planned",
+  "created_at": "2026-02-27T19:20:00.000000"
+}

conductor/archive/mma_data_architecture_dag_engine/plan.md

@@ -0,0 +1,19 @@
+# Implementation Plan: MMA Data Architecture & DAG Engine
+
+## Phase 1: Track-Scoped State Management
+- [x] Task: Define the data schema for a Track (Metadata, Discussion History, Task List). [2efe80e]
+- [x] Task: Update `project_manager.py` to create and read from `tracks/<track_id>/state.toml`. [e1a3712]
+- [x] Task: Ensure Tier 2 (Tech Lead) history is securely scoped to the active track's state file. [b845b89]
+
+## Phase 2: Python DAG Engine
+- [x] Task: Create a `Task` class with `status` (Blocked, Ready, In Progress, Review, Done) and `depends_on` fields. [a3cfeff]
+- [x] Task: Implement a topological sorting algorithm to resolve execution order. [f85ec9d]
+- [x] Task: Write robust unit tests verifying cycle detection and dependency resolution. [f85ec9d]
+
+## Phase 3: Execution State Machine
+- [x] Task: Implement the core loop that evaluates the DAG and identifies "Ready" tasks. [154957f]
+- [x] Task: Create configuration settings for "Auto-Queue" vs "Manual Step" execution modes. [154957f]
+- [x] Task: Connect the state machine to the backend dispatcher, preparing it for GUI integration. [2429b7c]
+
+## Phase: Review Fixes
+- [x] Task: Apply review suggestions [6548ce6]

conductor/archive/mma_data_architecture_dag_engine/spec.md

@@ -0,0 +1,34 @@
+# Track Specification: MMA Data Architecture & DAG Engine
+
+## Overview
+Restructure how `manual_slop` stores and executes work. The current implementation relies on global state and linear execution, which does not support the complexity of multi-agent, task-based workflows. This track establishes a robust, data-oriented foundation using track-scoped state and a native Python Directed Acyclic Graph (DAG) engine.
+
+## Goals
+1.  **Track-Scoped State:** Transition away from a single global `manual_slop_history.toml` to a per-track state structure (e.g., `tracks/<track_id>/state.toml`) to manage specific discussion history and context.
+2.  **Task DAG Engine:** Implement a topological sorter and DAG execution engine in Python to manage dependencies between tasks.
+3.  **Execution State Machine:** Build an internal state machine that governs whether the DAG auto-advances or waits for manual user intervention to spawn the next worker.
+
+## Constraints
+-   Must integrate seamlessly with the existing project TOML structure.
+-   Avoid external complex dependencies (like Steve Yegge's Beads) for now; rely on standard Python libraries or lightweight custom implementations to ensure tight coupling with the DearPyGui stack.
+
+## Context & Origins
+This track was born from the "Human Verification" phase of the initial MMA Orchestrator prototype (`mma_orchestrator_integration_20260226`). We realized that while the backend API plumbing for the hierarchical MMA tiers (Tiers 1-4) was technically functional, the product lacked the necessary state management, UX visualization, and human-in-the-loop security gates to be usable. 
+
+**Key Takeaways from the Prototype Phase:**
+- The Tier 2 (Tech Lead) needs its own track-scoped discussion history, rather than polluting the global project history.
+- Tasks within a track require a DAG (Directed Acyclic Graph) engine to manage complex dependencies and blocking states.
+- The GUI must visualize this DAG and stream the output of individual workers directly to their associated tasks.
+- We must enforce tiered context subsetting so that Tier 3/4 workers don't receive the massive global context blob, and we need a pre-spawn approval modal so the user can intercept, review, and modify worker prompts/contexts before they execute.
+
+**Instructions for the Implementing Agent:**
+As you execute this track, ensure you maintain alignment with the other Phase 2 tracks. If you learn something that impacts the dependent tracks, please append a similar "Context Summary" to their `spec.md` files before concluding your run.
+
+## Execution Order & Dependencies
+This is a multi-track phase. To ensure architectural integrity, these tracks **MUST** be executed in the following strict order:
+1. **[CURRENT] MMA Data Architecture & DAG Engine:** (Builds the state and execution foundation)
+2. **Tiered Context Scoping & HITL Approval:** (Builds the security and context subsetting on top of the state)
+3. **MMA Dashboard Visualization Overhaul:** (Builds the UI to visualize the state and subsets)
+4. **Robust Live Simulation Verification:** (Builds the tests to verify the UI and state)
+
+**Prerequisites for this track:** None. This must be executed FIRST.

conductor/archive/mma_formalization_20260225/index.md

@@ -0,0 +1,5 @@
+# Track mma_formalization_20260225 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/mma_formalization_20260225/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "mma_formalization_20260225",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-25T18:48:00Z",
+  "updated_at": "2026-02-25T18:48:00Z",
+  "description": "Improve conductors use of 4-tier mma architecture workflow, skills, subagents. Introduce a seaprate skill for each dedicated tier and a dedicated cli tool to execute the roles appropriate/gather context as defined for that role's domain."
+}

conductor/archive/mma_formalization_20260225/plan.md

@@ -0,0 +1,27 @@
+# Implementation Plan: 4-Tier MMA Architecture Formalization
+
+## Phase 1: Tiered Skills Implementation [checkpoint: 6ce3ea7]
+- [x] Task: Create `mma-tier1-orchestrator` skill in `.gemini/skills/` [fe1862a]
+- [x] Task: Create `mma-tier2-tech-lead` skill in `.gemini/skills/` [fe1862a]
+- [x] Task: Create `mma-tier3-worker` skill in `.gemini/skills/` [fe1862a]
+- [x] Task: Create `mma-tier4-qa` skill in `.gemini/skills/` [fe1862a]
+- [x] Task: Conductor - User Manual Verification 'Phase 1: Tiered Skills Implementation' (Protocol in workflow.md) [6ce3ea7]
+
+## Phase 2: `mma-exec` CLI - Core Scoping [checkpoint: dd7e591]
+- [x] Task: Scaffold `scripts/mma_exec.py` with basic CLI structure (argparse/click) [0b2cd32]
+- [x] Task: Implement Role-Scoped Document selection logic (mapping roles to `product.md`, `tech-stack.md`, etc.) [55c0fd1]
+- [x] Task: Implement the "Context Amnesia" bridge (ensuring a fresh subprocess session for each call) [f6e6d41]
+- [x] Task: Integrate `mma-exec` with the existing `ai_client.py` logic (SKIPPED - out of scope for Conductor)
+- [x] Task: Conductor - User Manual Verification 'Phase 2: mma-exec CLI - Core Scoping' (Protocol in workflow.md) [0195329]
+
+## Phase 3: Advanced Context Features [checkpoint: eb64e52]
+- [x] Task: Implement AST "Skeleton View" generator using `tree-sitter` in `scripts/mma_exec.py` [4e564aa]
+- [x] Task: Add dependency mapping to `mma-exec` (providing skeletons of imported files to Workers) [32ec14f]
+- [x] Task: Implement logging/auditing for all role hand-offs in `logs/mma_delegation.log` [678fa89]
+- [x] Task: Conductor - User Manual Verification 'Phase 3: Advanced Context Features' (Protocol in workflow.md) [eb64e52]
+
+## Phase 4: Workflow & Conductor Integration [checkpoint: 0d533ec]
+- [x] Task: Update `conductor/workflow.md` with new MMA role definitions and `mma-exec` commands [5e256d1]
+- [x] Task: Create a Conductor helper/alias in `scripts/` to simplify manual role triggering [df1c429]
+- [x] Task: Final end-to-end verification using a sample feature implementation [verified]
+- [x] Task: Conductor - User Manual Verification 'Phase 4: Workflow & Conductor Integration' (Protocol in workflow.md) [0d533ec]

conductor/archive/mma_formalization_20260225/spec.md

@@ -0,0 +1,43 @@
+# Specification: 4-Tier MMA Architecture Formalization
+
+## Overview
+This track aims to formalize and automate the 4-Tier Hierarchical Multi-Model Architecture (MMA) within the Conductor framework. It introduces specialized skills for each tier and a new specialized CLI tool (`mma-exec`) to handle role-specific context gathering and "Context Amnesia" protocols.
+
+## Goals
+- Isolate cognitive load for sub-agents by providing only domain-specific context.
+- Minimize token burn through "Context Amnesia" and AST-based skeleton views.
+- Formalize the Orchestrator (Tier 1), Tech Lead (Tier 2), Worker (Tier 3), and QA (Tier 4) roles.
+
+## Functional Requirements
+
+### 1. Specialized Tier Skills
+Create four new Gemini CLI skills located in `.gemini/skills/`:
+- **mma-tier1-orchestrator:** Focused on product alignment, high-level planning, and track management.
+- **mma-tier2-tech-lead:** Focused on architectural design, tech stack alignment, and code review.
+- **mma-tier3-worker:** Focused on TDD implementation, surgical code changes, and following specific specs.
+- **mma-tier4-qa:** Focused on test analysis, error summarization, and bug reproduction.
+
+### 2. Specialized CLI: `mma-exec`
+A new Python-based CLI tool to replace/extend `run_subagent.ps1`:
+- **Role Scoping:** Automatically determines which project documents (Product, Tech Stack, etc.) to include based on the active role.
+- **AST Skeleton Views:** Integrates with `tree-sitter` to generate and provide only the interface/signature skeletons of dependency files to Tier 3 Workers.
+- **Context Amnesia Protocol:** Ensures each role execution starts with a fresh, scoped context to prevent history-induced hallucinations.
+- **Conductor Integration:** Designed to be called by the Conductor agent or manually by the developer.
+
+### 3. Workflow Integration
+- Update `conductor/workflow.md` to formalize the use of `mma-exec` and the tiered skills.
+- Add specific commands/aliases within the Conductor context to trigger role hand-offs.
+
+## Non-Functional Requirements
+- **Performance:** Context gathering (including AST parsing) must be fast enough for interactive use.
+- **Transparency:** All hand-offs and context inclusions must be logged for developer auditing.
+
+## Acceptance Criteria
+- [ ] Four new skills are registered and accessible.
+- [ ] `mma-exec` tool can successfully spawn a worker with only AST skeleton views of requested dependencies.
+- [ ] A test task can be implemented using the tiered delegation flow without manual context curation.
+- [ ] `workflow.md` documentation is fully updated.
+
+## Out of Scope
+- Migrating existing tracks to the new architecture (only new tasks/tracks are required to use it).
+- Automating the *decision* of when to hand off (remains semi-automated/manual per user preference).

conductor/archive/mma_implementation_20260224/index.md

@@ -0,0 +1,5 @@
+# Track mma_implementation_20260224 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/mma_implementation_20260224/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "mma_implementation_20260224",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-24T00:00:00Z",
+  "updated_at": "2026-02-24T00:00:00Z",
+  "description": "4-Tier Architecture Implementation & Conductor Self-Improvement"
+}

conductor/archive/mma_implementation_20260224/migration_epics.md

@@ -0,0 +1,128 @@
+# MMA Migration: Epics and Detailed Tasks
+
+## Track 1: The Memory Foundations (AST Parser)
+
+**Goal:** Build the engine that prevents token-bloat by turning massive source files into curated memory views.
+
+### 1. TDD Approach for `tree-sitter` Integration
+- Create `tests/test_file_cache_ast.py`.
+- Define mock Python source files containing various structures (classes, functions, docstrings, `@core_logic` decorators, `# [HOT]` comments).
+- Write failing tests that instantiate `ASTParser` and assert that `get_skeleton_view()` and `get_curated_view()` return the precisely filtered strings.
+- **Red Phase:** Ensure tests fail because `ASTParser` does not exist.
+- **Green Phase:** Implement the tree-sitter logic iteratively until strings match exactly.
+
+### 2. `ASTParser` Extraction Rules (Tasks)
+- **Task 1.1: Dependency Setup**
+  - Add `tree-sitter` and `tree-sitter-python` to `pyproject.toml` / `requirements.txt`.
+- **Task 1.2: Core Parser Class**
+  - Create `ASTParser` in `file_cache.py` that initializes the language parser.
+- **Task 1.3: Skeleton View Extraction**
+  - Write query to extract `function_definition` and `class_definition`.
+  - Keep signatures, parameters, and return type hints.
+  - Replace all bodies with `pass`.
+- **Task 1.4: Curated View Extraction**
+  - Write query to keep class structures and `expression_statement` docstrings.
+  - Implement heuristic to preserve full bodies of functions decorated with `@core_logic` or containing `# [HOT]` comments.
+  - Replace all other function bodies with `... # Hidden`.
+
+### 3. Acceptance Testing Criteria
+- **Unit Tests:** All AST parsing tests pass with >90% coverage for `file_cache.py`.
+- **Integration Test:** Execute the parser on a large, complex project file (e.g., `ai_client.py`). The output `Skeleton View` must be less than 15% of the original token count. The `Curated View` must correctly retain docstrings and marked functions while stripping standard bodies.
+## Track 2: State Machine & Data Structures
+
+**Goal:** Define the rigid Python objects (Pydantic/Dataclasses) that AI agents will pass to each other, enforcing structured data over loose chat strings.
+
+### 1. TDD Approach for \models.py\
+- Create \	ests/test_models.py\.
+- Write failing tests that instantiate \Track\, \Ticket\, and \WorkerContext\ with various valid and invalid schemas.
+- Write tests that assert state transitions (e.g., from \pending\ to \locked\, from \step_paused\ to \completed\) correctly update internal flags and dependencies.
+- **Red Phase:** Tests fail because \models.py\ classes are undefined or lack transition methods.
+- **Green Phase:** Implement the dataclasses and state mutators.
+
+### 2. State Machine Tasks
+- **Task 2.1: The Dataclasses**
+  - Create \models.py\. Define \Ticket\ (id, target_file, prompt, worker_archetype, status, dependencies).
+  - Define \Track\ (id, title, description, status, tickets).
+- **Task 2.2: Worker Context Definition**
+  - Define \WorkerContext\ holding a \Ticket\ ID, assigned model, configuration injection, and an ephemeral \messages\ array.
+- **Task 2.3: State Mutator Methods**
+  - Implement methods like \	icket.mark_blocked(dependency_id)\, \	icket.mark_complete()\, and \	rack.get_executable_tickets()\. Ensure strict validation of valid state transitions.
+
+### 3. Acceptance Testing Criteria
+- **Unit Tests:** \models.py\ has 100% test coverage for all state transitions.
+- **Integration Test:** Instantiate a \Track\ with 3 dependent \Tickets\ in Python. Programmatically mark tickets as complete and assert that the subsequent dependent tickets transition from \locked\ to \pending\ without any AI involvement.
+
+## Track 3: The Linear Orchestrator & Execution Clutch
+
+**Goal:** Build the synchronous, debuggable core loop that runs a single Tier 3 Worker and pauses for human approval.
+
+### 1. TDD Approach for \multi_agent_conductor.py\
+- Create \	ests/test_conductor.py\.
+- Write tests that mock the AI client response (e.g., returning a mock tool call like \write_file\).
+- Test that \
+un_worker_lifecycle(ticket: Ticket)\ fetches the Raw View from \ile_cache.py\, formats messages, and processes the mock output.
+- Test that execution pauses (waits for a simulated human signal) when the \	rust_level\ dictates.
+- **Red Phase:** Failure occurs because \multi_agent_conductor.py\ lacks the lifecycle execution loop.
+- **Green Phase:** Implement the \ConductorEngine\ core execution block.
+
+### 2. Linear Orchestration Tasks
+- **Task 3.1: The Engine Core**
+  - Create \multi_agent_conductor.py\. Implement the \ConductorEngine\ class containing the \
+un_worker_lifecycle\ synchronous execution.
+- **Task 3.2: Context Injection**
+  - Implement logic reading the Ticket target, querying \ile_cache.py\ for the \Raw View\, and formatting the messages array for the API.
+- **Task 3.3: The HITL Execution Clutch**
+  - Before executing tools via \mcp_client.py\ or \shell_runner.py\, intercept the tool payload if the Worker's archetype dictates a \step\ mode.
+  - Wait for explicit user confirmation via a CLI prompt (or event block for UI future-proofing). Allow editing of the JSON payload.
+  - Flush history upon \TicketCompleted\.
+
+### 3. Acceptance Testing Criteria
+- **Unit Tests:** Context generation, API schema mapping, and event-blocking are tested for all Edge cases.
+- **Integration Test:** Manually execute a script pointing the \ConductorEngine\ at a dummy file. The CLI should pause before \write_file\ execution, display the diff, allow manual JSON editing via terminal input, execute the updated JSON file modification, and return \Task Complete\.
+
+## Track 4: Tier 4 QA Interception
+
+**Goal:** Stop error traces from destroying the Worker's token window by routing crashes through a cheap, stateless translator.
+
+### 1. TDD Approach for \shell_runner.py\
+- Create \	ests/test_shell_runner.py\.
+- Write tests that mock a local execution failure (e.g., returning a mock 3000-line Python stack trace).
+- Test that the error is intercepted and passed to a mock Tier 4 agent.
+- Test that the output is compressed into a 20-word fix before returning.
+- **Red Phase:** Fails because no interception loop exists in \shell_runner.py\.
+- **Green Phase:** Implement the try/except logic handling \subprocess.run()\ with \
+eturncode != 0\.
+
+### 2. QA Interception Tasks
+- **Task 4.1: The Interceptor Loop**
+  - Open \shell_runner.py\ and catch execution errors.
+- **Task 4.2: Tier 4 Instantiation**
+  - Construct a secondary, synchronous API call directly to the \default_cheap\ model, sending the raw \stderr\ and the offending code snippet.
+- **Task 4.3: Payload Formatting**
+  - Inject the 20-word fix response from the Tier 4 agent back into the main Tier 3 worker's history context as a system hint.
+
+### 3. Acceptance Testing Criteria
+- **Unit Tests:** Verify that massive error outputs never leak uncompressed into the main history logs.
+- **Integration Test:** Purposely introduce a syntax error in a local script. Ensure the orchestrator catches it, pings the mock/cheap API, and the history log receives the 20-word hint instead of the 200-line stack trace.
+
+## Track 5: UI Decoupling & Tier 1/2 Routing (The Final Boss)
+
+**Goal:** Bring the whole system online by letting Tier 1 and Tier 2 generate Tickets dynamically, managed via an asynchronous Event Bus.
+
+### 1. TDD Approach for \gui_2.py\ Decoupling
+- Create \	ests/test_gui_decoupling.py\.
+- Write tests that instantiate a mocked GUI instance listening to an \syncio.Queue\.
+- Mock pushing \TrackStateUpdated\ and \TicketStarted\ events into the queue and ensure the GUI updates its view state rather than calling LLM endpoints directly.
+- **Red Phase:** Failure occurs because \gui_2.py\ is tightly coupled with \i_client.py\ logic.
+- **Green Phase:** Implement the \AgentBus\ messaging system linking \multi_agent_conductor.py\ to \gui_2.py\.
+
+### 2. Tier 1/2 Routing Tasks
+- **Task 5.1: The Event Bus**
+  - Implement an \syncio.Queue\ in \multi_agent_conductor.py\.
+- **Task 5.2: Tier 1 & 2 System Prompts**
+  - Define system prompts that force the 3.1 Pro/3.5 Sonnet models to output strict JSON arrays defining the Tracks and Tickets (utilizing native Structured Outputs).
+- **Task 5.3: The Dispatcher**
+  - Write an async loop that reads JSON from Tier 2, converts them into \Ticket\ objects, and pushes them onto the queue.
+  - Implement the Stub Resolver to enforce \contract_stubber\ dependent execution flow.
+- **Task 5.4: UI Component Update**
+  - Remove direct LLM calls from \gui_2.py\. Wire user inputs into \UserRequestEvents\ for the Orchestrator's queue.

conductor/archive/mma_implementation_20260224/plan.md

@@ -0,0 +1,50 @@
+# Implementation Plan: 4-Tier Architecture Implementation & Conductor Self-Improvement
+
+## Phase 1: `manual_slop` Migration Planning [checkpoint: e07e8e5]
+- [x] Task: Synthesize MMA Documentation [46b351e]
+    - [x] Read and analyze `./MMA_Support/Data_Pipelines_and_Config.md` and `./MMA_Support/OriginalDiscussion.md`
+    - [x] Read and analyze `./MMA_Support/Tier1_Orchestrator.md` through `./MMA_Support/Tier4_Utility.md`
+    - [x] Document key takeaways and constraints for the migration plan
+- [x] Task: Draft Track 1 - The Memory Foundations (AST Parser) [bdd935d]
+    - [x] Define TDD approach for `tree-sitter` integration in `file_cache.py`
+    - [x] Specify tasks for `ASTParser` extraction rules (Skeleton View, Curated View)
+    - [x] Define acceptance testing criteria for AST extraction
+- [x] Task: Draft Track 2 - State Machine & Data Structures [1198aee]
+    - [x] Define TDD approach for `models.py` (`Track`, `Ticket`, `WorkerContext`)
+    - [x] Specify tasks for state mutator methods
+    - [x] Define acceptance testing criteria for state transitions
+- [x] Task: Draft Track 3 - The Linear Orchestrator & Execution Clutch [aaeed92]
+    - [x] Define TDD approach for `multi_agent_conductor.py` (`run_worker_lifecycle`)
+    - [x] Specify tasks for context injection and HITL Clutch implementation
+    - [x] Define acceptance testing criteria for the linear orchestration loop
+- [x] Task: Draft Track 4 - Tier 4 QA Interception [584bff9]
+    - [x] Define TDD approach for `shell_runner.py` stderr interception
+    - [x] Specify tasks for routing errors to the cheap API model
+    - [x] Define acceptance testing criteria for the QA interception loop
+- [x] Task: Draft Track 5 - UI Decoupling & Tier 1/2 Routing (The Final Boss) [67734c9]
+    - [x] Define TDD approach for async queue in `multi_agent_conductor.py`
+    - [x] Specify tasks for Tier 1 & 2 system prompts and the Dispatcher async loop
+    - [x] Define acceptance testing criteria for UI decoupling and dynamic routing
+- [x] Task: Conductor - User Manual Verification '`manual_slop` Migration Planning' (Protocol in workflow.md) [e07e8e5]
+
+## Phase 2: Conductor Self-Reflection & Upgrade Strategy [checkpoint: 40339a1]
+- [x] Task: Research Optimal Proposal Format [0c5f8b9]
+    - [x] Search Gemini CLI documentation for extension guidelines
+    - [x] Search Conductor documentation for tuning and advice
+    - [x] Define the structure for `proposal.md` based on findings
+- [x] Task: Draft Proposal - Memory Siloing & Token Firewalling [59556d1]
+    - [x] Evaluate current `conductor` context management
+    - [x] Propose strategies to prevent token bloat during planning and execution
+    - [x] Write the corresponding section in `proposal.md`
+- [x] Task: Draft Proposal - Execution Clutch & Linear Debug Mode [baff5c1]
+    - [x] Evaluate current `conductor` execution workflows
+    - [x] Propose mechanisms for manual step-through and auto modes
+    - [x] Write the corresponding section in `proposal.md`
+- [x] Task: Draft Proposal - Multi-Model/Sub-Agent Delegation [f62bf31]
+    - [x] Evaluate current `conductor` single-model reliance
+    - [x] Propose a design for delegating tasks (e.g., summarization, syntax-fixing) to sub-agents
+    - [x] Write the corresponding section in `proposal.md`
+- [x] Task: Review and Finalize Proposal [f62bf31]
+    - [x] Ensure all three core areas are addressed with equal priority
+    - [x] Verify alignment with the overall 4-Tier Architecture philosophy
+- [x] Task: Conductor - User Manual Verification 'Conductor Self-Reflection & Upgrade Strategy' (Protocol in workflow.md) [40339a1]

conductor/archive/mma_implementation_20260224/proposal.md

@@ -0,0 +1,40 @@
+# Conductor Self-Reflection & Upgrade Strategy Proposal
+
+## 1. Executive Summary
+This proposal outlines a strategic path for upgrading the Gemini CLI `conductor` extension to fully embrace the 4-Tier Hierarchical Multi-Model Architecture principles. By migrating from a monolithic, context-heavy single-agent loop to a compartmentalized, multi-model delegation system, Conductor can drastically reduce token burn, mitigate hallucination loops, and grant developers surgical Human-In-The-Loop (HITL) control over execution tasks.
+
+## 2. Memory Siloing & Token Firewalling
+
+### Current Evaluation
+Currently, the `conductor` extension relies heavily on reading index files and full markdown texts recursively through the project structure. This injects entire tracks, plans, guidelines, and specifications into the LLM context continuously. While beneficial for ensuring alignment with user instructions, this linear scaling creates immense token bloat during repetitive planning and execution loops. 
+
+### Proposed Upgrade Strategy
+To align with the 4-Tier Architecture, the Conductor extension must implement **Token Firewalling**:
+1. **Curated Manifests & Viewports:** Implement an extension tool or AST parser hook to generate "Skeleton Views" or restricted tree maps instead of fully loading index files into the prompt.
+2. **Stateless Sub-Agent Invocations:** Delegate localized tasks (like writing documentation updates to a single file) to a background sub-agent (via `run_shell_command` leveraging a separate stateless invocation, or by utilizing Gemini CLI's sub-agent framework). This prevents the main conductor thread from storing the trial-and-error generation in its history.
+3. **Amnesiac Context Management:** Incorporate lifecycle hooks (`before_tool_call`, `after_tool_call`) to clean up unnecessary tool outputs from the active memory array, only keeping the 50-token summaries of execution outcomes.
+
+## 3. Execution Clutch & Linear Debug Mode
+
+### Current Evaluation
+Conductor currently employs an iterative, fire-and-forget `execute_tasks` workflow where each `replace`, `write_file`, and `run_shell_command` is done sequentially via its prompt instructions. While autonomous, the user's only control mechanism during rapid tool-calling is the standard CLI prompt interruption, which may leave tracked artifacts in an inconsistent state or execute runaway hallucinated loops.
+
+### Proposed Upgrade Strategy
+To enforce precise developer control, Conductor should natively embed a **Human-In-The-Loop Execution Clutch**:
+1. **Interactive Checkpoints (Trust Levels):** Use extension hooks like `before_tool_call` to intercept payload executions based on heuristic models. Tools like `replace` might trigger an interactive payload editor (`vim` / CLI editor plugin) before applying the JSON parameters, ensuring full developer review.
+2. **Global Linear Mode Flag:** Implement a `gemini conductor:implement --step` flag. This configures the engine to pause execution and prompt the user using `ask_user` natively after every major milestone, allowing validation of file diffs and tool payloads before resuming.
+3. **Rollback Mutators:** Provide quick access commands (e.g., via `after_tool_call`) to reject the change, auto-restoring the last known file state, and feeding the error/feedback directly back to the model without breaking the run loop.
+
+## 4. Multi-Model/Sub-Agent Delegation
+
+### Current Evaluation
+Conductor heavily relies on the single primary LLM instantiated by the Gemini CLI session. When acting as a PM, Tech Lead, and Worker simultaneously, the model experiences extreme context exhaustion. Furthermore, handling minor formatting, syntax repairs, or summaries with expensive high-tier reasoning models results in suboptimal cost-efficiency.
+
+### Proposed Upgrade Strategy
+Conductor should leverage the native **Sub-Agent & Skill Routing capabilities**:
+1. **Dynamic Tier Routing:** Utilize specific Sub-agents (like `codebase_investigator` for planning/AST generation) and custom Skills for discrete tasks.
+2. **Stateless Utility Agents (Tier 4):** Hook into test runner commands via `after_tool_call`. If `pytest` fails with massive `stderr`, immediately invoke a cheap background utility sub-agent to parse the log and return a condensed 20-word summary back to the main Orchestrator, rather than feeding the main Orchestrator raw traceback tokens.
+3. **Contract Stubbers:** Embed `contract_stubber` skills that explicitly limit a sub-agent's action strictly to writing `class` or `def` definitions, ensuring cross-module dependency generation without full implementation drift.
+
+## 5. Implementation Strategy
+These upgrades can be realized by augmenting the `gemini-extension.json` manifest with designated MCP hooks, adding new custom Skills to `~/.gemini/skills/`, and overriding default CLI execution flows with `before_tool_call` and `after_tool_call` interception logic tailored explicitly for Token Firewalling and Execution Checkpoints.

conductor/archive/mma_implementation_20260224/spec.md

@@ -0,0 +1,37 @@
+# Specification: 4-Tier Architecture Implementation & Conductor Self-Improvement
+
+## 1. Overview
+This track encompasses two major phases. Phase 1 focuses on designing a comprehensive, step-by-step implementation plan to refactor the `manual_slop` codebase from a single-agent linear chat into an asynchronous, 4-Tier Hierarchical Multi-Model Architecture. Phase 2 focuses on evaluating the Gemini CLI `conductor` extension itself and proposing architectural upgrades to enforce multi-tier, cost-saving, and context-preserving disciplines.
+
+## 2. Functional Requirements
+
+### Phase 1: `manual_slop` Implementation Planning
+- **Synthesis:** Read and synthesize all markdown files within the `./MMA_Support/` directory.
+- **Plan Generation:** Generate a detailed implementation plan (`plan.md`) for the `manual_slop` migration.
+    - The plan must break down the migration into actionable sub-tracks or tickets (Epics and detailed technical tasks).
+    - It must strictly follow the iterative safe-migration strategy outlined in `MMA_Support/Implementation_Tracks.md`.
+    - The sequence must be:
+        1. Tree-sitter AST parsing.
+        2. State Machines.
+        3. Linear Orchestrator.
+        4. Tier 4 QA Interception.
+        5. UI Decoupling.
+    - Every ticket/task must include explicit steps for testing and verifying the implementation.
+
+### Phase 2: Conductor Self-Reflection & Upgrade Strategy
+- **Evaluation:** Critically evaluate the `conductor` extension's architecture and workflows against the principles of the 4-Tier Architecture.
+- **Formal Proposal:** Deliver a formal proposal document within this track's directory (`proposal.md`).
+    - **Format Research:** Investigate the optimal format for the proposal based on Google's documentation for extending or tuning Conductor.
+    - **Content:** The proposal must address three core areas with equal priority:
+        1. **Strict Memory Siloing & Token Firewalling:** How to reduce token bloat during Conductor's planning and execution loops.
+        2. **Execution Clutch & Linear Debug Mode:** How to implement manual step-through or auto modes when managing complex tracks.
+        3. **Multi-Model/Sub-Agent Delegation:** Design a system for internally delegating tasks (e.g., summarization, syntax fixing) to cheaper, faster models.
+
+## 3. Acceptance Criteria
+- [ ] A fully populated `plan.md` exists within this track, detailing the `manual_slop` migration with Epics, detailed tasks, and testing steps.
+- [ ] A formal proposal document (`proposal.md`) exists within this track, addressing the three core areas for Conductor's self-improvement.
+- [ ] The proposal's format is justified based on official documentation or best practices for Conductor extensions.
+
+## 4. Out of Scope
+- Actual implementation of the `manual_slop` refactor (this track is purely for planning the implementation).
+- Actual modification of the `conductor` extension's core logic.

conductor/archive/mma_implementation_20260224/synthesis.md

@@ -0,0 +1,28 @@
+# MMA Documentation Synthesis
+
+## Key Takeaways
+
+1. **Architecture Model**: 4-Tier Hierarchical Multi-Model Architecture mimicking a senior engineering department.
+   - **Tier 1 (Product Manager)**: High-reasoning models (Gemini 3.1 Pro/Claude 3.5 Sonnet) focusing on Epics and Tracks.
+   - **Tier 2 (Tech Lead)**: Mid-cost models (Gemini 3.0 Flash/2.5 Pro) for Track delegation, Ticket generation, and interface-driven development (Stub-and-Resolve).
+   - **Tier 3 (Contributors)**: Cheap/Fast models (DeepSeek V3/R1, Gemini 2.5 Flash) acting as amnesiac workers for heads-down coding.
+   - **Tier 4 (QA/Compiler)**: Ultra-cheap models (DeepSeek V3) for stateless translation of raw errors to human language.
+
+2. **Strict Context Management**: 
+   - Uses `tree-sitter` for deterministic AST extraction (`Skeleton View`, `Curated Implementation View`, `Directory Map`).
+   - "Context Amnesia" ensures worker threads start fresh and do not accumulate hallucination-inducing token bloat.
+
+3. **Data Pipelines & Formats**:
+   - Tiers 1 & 2 output **Godot ECS Flat Relational Lists** (e.g., INI-style flat lists with `depends_on` pointers) to build DAGs. This avoids JSON nesting nightmares.
+   - Tier 3 uses **XML tags** (`<file_path>`, `<file_content>`) to avoid string escaping friction.
+
+4. **Execution Flow**:
+   - The engine is decoupled from the UI using an `asyncio` event bus.
+   - A global **"Execution Clutch"** allows falling back from `async` parallel swarm mode to strict `linear` step mode for deterministic debugging and human-in-the-loop (HITL) overrides.
+
+## Constraints for Migration Plan
+
+- **Security**: `credentials.toml` must be strictly isolated and ignored in version control.
+- **Phased Rollout**: Migration cannot be a single rewrite. It must follow strict tracks: AST Parser -> State Machine -> Linear Orchestrator -> Tier 4 QA -> UI Decoupling.
+- **Tooling Constraints**: `tree-sitter` is mandatory for AST parsing.
+- **UI State**: The GUI must be fully decoupled ("dumb" renderer) responding to queue events instead of blocking on LLM calls.

conductor/archive/mma_orchestrator_integration_20260226/index.md

@@ -0,0 +1,8 @@
+# MMA Orchestrator Integration
+
+This track implements the full hierarchical orchestration loop, connecting Tier 1 (PM) strategic planning with Tier 2 (Tech Lead) tactical ticket generation.
+
+### Navigation
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Previous Track: MMA Core Engine Implementation (Archived)](../../archive/mma_core_engine_20260224/index.md)

conductor/archive/mma_orchestrator_integration_20260226/metadata.json

@@ -0,0 +1,6 @@
+{
+  "id": "mma_orchestrator_integration_20260226",
+  "title": "MMA Orchestrator Integration",
+  "status": "in-progress",
+  "created_at": "2026-02-26T22:04:00.000000"
+}

conductor/archive/mma_orchestrator_integration_20260226/plan.md

@@ -0,0 +1,42 @@
+# Implementation Plan: MMA Orchestrator Integration
+
+## Phase 1: Tier 1 Strategic PM Implementation
+- [x] Task: PM Planning Hook
+    - [x] Create `orchestrator_pm.py` to handle the Tier 1 Strategic prompt.
+    - [x] Implement the `generate_tracks(user_request, repo_map)` function.
+- [x] Task: Project History Aggregation
+    - [x] Summarize past track results to provide context for new epics.
+
+## Phase 2: Tier 2 Tactical Dispatcher Implementation
+- [x] Task: Tech Lead Dispatcher Hook
+    - [x] Create `conductor_tech_lead.py` to handle the Tier 2 Dispatcher prompt.
+    - [x] Implement the `generate_tickets(track_brief, module_skeletons)` function.
+- [x] Task: DAG Construction
+    - [x] Build the topological dependency graph from the Tech Lead's ticket list.
+
+## Phase 3: Guided Planning UX & Interaction
+- [x] Task: Strategic Planning View
+    - [x] Implement a "Track Proposal" modal in `gui_2.py` for reviewing Tier 1's plans.
+    - [x] Allow manual editing of track goals and acceptance criteria (Manual Curation).
+- [x] Task: Tactical Dispatcher View
+    - [x] Implement a "Ticket DAG" visualization or interactive list in the MMA Dashboard.
+    - [x] Allow manual "Skip", "Retry", or "Re-assign" actions on individual tickets.
+- [x] Task: The Orchestrator Main Loop
+    - [x] Implement the async state machine in `gui_2.py` that moves from Planning -> Dispatching -> Execution.
+- [x] Task: Project Metadata Serialization
+    - [x] Persist the active epic, tracks, and tickets to `manual_slop.toml`.
+
+## Phase 4: Product Alignment & Refinement
+- [x] Task: UX Differentiator Audit
+    - [x] Ensure the UX prioritizes "Expert Oversight" over "Full Autonomy" (Manual Slop vs. Gemini CLI).
+    - [x] Add detailed token metrics and Tier-specific latency indicators to the Dashboard.
+
+## Phase 5: Exhaustive Testing & Regression
+- [x] Task: Headless Engine Verification (d087a20)
+    - [x] Create `tests/test_orchestration_logic.py` to verify Tier 1 -> Tier 2 -> Tier 3 flow without a GUI.
+    - [x] Verify DAG resolution and error handling (e.g., blocked tickets).
+- [x] Task: Visual Verification Suite (d087a20)
+    - [x] Create `tests/visual_orchestration_verification.py` using `ApiHookClient`.
+    - [x] Simulate a full "Epic" lifecycle: User Prompt -> Track Review -> Ticket Generation -> Execution.
+- [x] Task: Core Regression Suite (d087a20)
+    - [x] Run all existing MMA, Conductor, and GUI tests to ensure no regressions.

conductor/archive/mma_orchestrator_integration_20260226/spec.md

@@ -0,0 +1,21 @@
+# Track Specification: MMA Orchestrator Integration
+
+## Overview
+Implement the full hierarchical orchestration loop, connecting Tier 1 (PM) strategic planning with Tier 2 (Tech Lead) tactical ticket generation. This track will enable the GUI to autonomously break down high-level user 'Epics' into actionable tracks and tickets, and manage their execution through the multi-agent system.
+
+## Goals
+1. **Tier 1 Epic Planning:** Implement the logic for the Orchestrator to analyze project state and user requests to generate Implementation Tracks.
+2. **Tier 2 Ticket Dispatcher:** Enable the Tech Lead to take a Track Brief and generate a Directed Acyclic Graph (DAG) of worker tickets.
+3. **GUI Orchestrator Loop:** Create the 'main thinking loop' in gui_2.py that manages the transition between planning and execution.
+4. **Project History Integration:** Ensure that completed tracks and tickets are properly summarized and stored in the project's persistent memory.
+
+## Technical Scope
+- **Prompt Engineering:** Finalize and integrate the Tier 1 and Tier 2 system prompts from mma_prompts.py.
+- **Backend Plumbing:** Extend multi_agent_conductor.py to handle track generation and ticket dispatching.
+- **UI Interaction:** Connect the 'New Project/Epic' button to the Tier 1 generator and the 'Start Track' button to the Tier 2 dispatcher.
+- **State Management:** Implement the transition logic between 'Strategic Planning', 'Tactical Dispatching', and 'Worker Execution' states.
+
+## Constraints
+- **Hierarchy Enforcement:** All planning must flow from Tier 1 to Tier 2, and all execution from Tier 2 to Tiers 3/4.
+- **Token Firewalling:** Use scripts/mma_exec.py for all tiered sub-agent calls.
+- **HITL Verification:** Maintain Step Mode capabilities for each transition (Planning -> Dispatching -> Execution).

conductor/archive/mma_utilization_refinement_20260226/index.md

@@ -0,0 +1,5 @@
+# Track mma_utilization_refinement_20260226 Context
+
+- [Specification](./spec.md)
+- [Implementation Plan](./plan.md)
+- [Metadata](./metadata.json)

conductor/archive/mma_utilization_refinement_20260226/metadata.json

@@ -0,0 +1,8 @@
+{
+  "track_id": "mma_utilization_refinement_20260226",
+  "type": "feature",
+  "status": "new",
+  "created_at": "2026-02-26T08:23:00Z",
+  "updated_at": "2026-02-26T08:23:00Z",
+  "description": "Refine MMA utilization by segregating tiers, enhancing sub-agent tooling with AST skeletons, and improving observability via dedicated logging."
+}