vibin

.gitignore

@@ -2,3 +2,4 @@
 __pycache__
 uv.lock
 colorforth_bootslop_002.md
+md_gen

MainContext.md

@@ -1,4 +1,4 @@
-**manual_slop** is a local GUI tool for manually curating and sending context to AI APIs. It aggregates files, screenshots, and discussion history into a structured markdown file and sends it to a chosen AI provider with a user-written message.
+**manual_slop** is a local GUI tool for manually curating and sending context to AI APIs. It aggregates files, screenshots, and discussion history into a structured markdown file and sends it to a chosen AI provider with a user-written message. The AI can also execute PowerShell scripts within the project directory, with user confirmation required before each execution.
 
 **Stack:**
 - `dearpygui` - GUI with docking/floating/resizable panels
@@ -8,9 +8,10 @@
 - `uv` - package/env management
 
 **Files:**
-- `gui.py` - main GUI, `App` class, all panels, all callbacks
+- `gui.py` - main GUI, `App` class, all panels, all callbacks, confirmation dialog
-- `ai_client.py` - unified provider wrapper, model listing, session management, send
+- `ai_client.py` - unified provider wrapper, model listing, session management, send, tool/function-call loop
-- `aggregate.py` - reads config, collects files/screenshots/discussion, writes numbered `.md` files to `output_dir/md_gen/`
+- `aggregate.py` - reads config, collects files/screenshots/discussion, writes numbered `.md` files to `output_dir`
+- `shell_runner.py` - subprocess wrapper that runs PowerShell scripts sandboxed to `base_dir`, returns stdout/stderr/exit code as a string
 - `config.toml` - namespace, output_dir, files paths+base_dir, screenshots paths+base_dir, discussion history array, ai provider+model
 - `credentials.toml` - gemini api_key, anthropic api_key
 
@@ -22,21 +23,41 @@
 - **Provider** - provider combo (gemini/anthropic), model listbox populated from API, fetch models button, status line
 - **Message** - multiline input, Gen+Send button, MD Only button, Reset session button
 - **Response** - readonly multiline displaying last AI response
+- **Tool Calls** - scrollable log of every PowerShell tool call the AI made, showing script and result; Clear button
+
+**AI Tool Use (PowerShell):**
+- 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 = 5`) 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
+- 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
+- Rejections return `"USER REJECTED: command was not executed"` to the AI
+- All tool calls (script + result/rejection) are appended to `_tool_log` and displayed in the Tool Calls panel
 
 **Data flow:**
 1. GUI edits are held in `App` state lists (`self.files`, `self.screenshots`, `self.history`) and dpg widget values
 2. `_flush_to_config()` pulls all widget values into `self.config` dict
 3. `_do_generate()` calls `_flush_to_config()`, saves `config.toml`, calls `aggregate.run(config)` which writes the md and returns `(markdown_str, path)`
-4. `cb_generate_send()` calls `_do_generate()` then threads a call to `ai_client.send(md, message)`
+4. `cb_generate_send()` calls `_do_generate()` then threads a call to `ai_client.send(md, message, base_dir)`
 5. `ai_client.send()` prepends the md as a `<context>` block to the user message and sends via the active provider chat session
-6. Sessions are stateful within a run (chat history maintained), `Reset` clears them
+6. If the AI responds with tool calls, the loop handles them (with GUI confirmation) before returning the final text response
+7. Sessions are stateful within a run (chat history maintained), `Reset` clears them and the tool log
 
 **Config persistence:**
 - Every send and save writes `config.toml` with current state including selected provider and model under `[ai]`
 - Discussion history is stored as a TOML array of strings in `[discussion] history`
 - File and screenshot paths are stored as TOML arrays, support absolute paths, relative paths from base_dir, and `**/*` wildcards
 
+**Threading model:**
+- DPG render loop runs on the main thread
+- AI sends and model fetches run on daemon background threads
+- `_pending_dialog` (guarded by a `threading.Lock`) is set by the background thread and consumed by the render loop each frame, calling `dialog.show()` on the main thread
+- `dialog.wait()` blocks the background thread on a `threading.Event` until the user acts
+
 **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`
 - System prompt support could be added as a field in `config.toml` and passed in `ai_client.send()`
 - Discussion history excerpts could be individually toggleable for inclusion in the generated md
+- `MAX_TOOL_ROUNDS` in `ai_client.py` caps agentic loops at 5 rounds; adjustable

ai_client.py

@@ -11,6 +11,13 @@ _gemini_chat = None
 _anthropic_client = None
 _anthropic_history: list[dict] = []
 
+# Injected by gui.py - called when AI wants to run a command.
+# Signature: (script: str) -> str | None
+# Returns the output string if approved, None if rejected.
+confirm_and_run_callback = None
+
+MAX_TOOL_ROUNDS = 5
+
 def _load_credentials() -> dict:
     with open("credentials.toml", "rb") as f:
         return tomllib.load(f)
@@ -61,21 +68,139 @@ def _list_anthropic_models() -> list[str]:
         models.append(m.id)
     return sorted(models)
 
+
+# --------------------------------------------------------- tool definition
+
+TOOL_NAME = "run_powershell"
+
+_ANTHROPIC_TOOLS = [
+    {
+        "name": TOOL_NAME,
+        "description": (
+            "Run a PowerShell script within the project base_dir. "
+            "Use this to create, edit, rename, or delete files and directories. "
+            "The working directory is set to base_dir automatically. "
+            "Always prefer targeted edits over full rewrites where possible. "
+            "stdout and stderr are returned to you as the result."
+        ),
+        "input_schema": {
+            "type": "object",
+            "properties": {
+                "script": {
+                    "type": "string",
+                    "description": "The PowerShell script to execute."
+                }
+            },
+            "required": ["script"]
+        }
+    }
+]
+
+def _gemini_tool_declaration():
+    from google.genai import types
+    return types.Tool(
+        function_declarations=[
+            types.FunctionDeclaration(
+                name=TOOL_NAME,
+                description=(
+                    "Run a PowerShell script within the project base_dir. "
+                    "Use this to create, edit, rename, or delete files and directories. "
+                    "The working directory is set to base_dir automatically. "
+                    "stdout and stderr are returned to you as the result."
+                ),
+                parameters=types.Schema(
+                    type=types.Type.OBJECT,
+                    properties={
+                        "script": types.Schema(
+                            type=types.Type.STRING,
+                            description="The PowerShell script to execute."
+                        )
+                    },
+                    required=["script"]
+                )
+            )
+        ]
+    )
+
+def _run_script(script: str, base_dir: str) -> str:
+    """
+    Delegate to the GUI confirmation callback.
+    Returns result string (stdout/stderr) or a rejection message.
+    """
+    if confirm_and_run_callback is None:
+        return "ERROR: no confirmation handler registered"
+    result = confirm_and_run_callback(script, base_dir)
+    if result is None:
+        return "USER REJECTED: command was not executed"
+    return result
+
 # ------------------------------------------------------------------ gemini
 
-def _ensure_gemini_chat():
+def _ensure_gemini_client():
-    global _gemini_client, _gemini_chat
+    global _gemini_client
-    if _gemini_chat is None:
+    if _gemini_client is None:
         from google import genai
         creds = _load_credentials()
         _gemini_client = genai.Client(api_key=creds["gemini"]["api_key"])
-        _gemini_chat = _gemini_client.chats.create(model=_model)
 
-def _send_gemini(md_content: str, user_message: str) -> str:
+def _send_gemini(md_content: str, user_message: str, base_dir: str) -> str:
-    _ensure_gemini_chat()
+    global _gemini_chat
+    from google import genai
+    from google.genai import types
+
+    _ensure_gemini_client()
+
+    # Gemini chats don't support mutating tools after creation,
+    # so we recreate if None (reset_session clears it).
+    if _gemini_chat is None:
+        _gemini_chat = _gemini_client.chats.create(
+            model=_model,
+            config=types.GenerateContentConfig(
+                tools=[_gemini_tool_declaration()]
+            )
+        )
+
     full_message = f"<context>\n{md_content}\n</context>\n\n{user_message}"
+
     response = _gemini_chat.send_message(full_message)
-    return response.text
+
+    for _ in range(MAX_TOOL_ROUNDS):
+        # Collect all function calls in this response
+        tool_calls = [
+            part.function_call
+            for candidate in response.candidates
+            for part in candidate.content.parts
+            if part.function_call is not None
+        ]
+        if not tool_calls:
+            break
+
+        # Execute each tool call and collect results
+        function_responses = []
+        for fc in tool_calls:
+            if fc.name == TOOL_NAME:
+                script = fc.args.get("script", "")
+                output = _run_script(script, base_dir)
+                function_responses.append(
+                    types.Part.from_function_response(
+                        name=TOOL_NAME,
+                        response={"output": output}
+                    )
+                )
+
+        if not function_responses:
+            break
+
+        response = _gemini_chat.send_message(function_responses)
+
+    # Extract text from final response
+    text_parts = [
+        part.text
+        for candidate in response.candidates
+        for part in candidate.content.parts
+        if hasattr(part, "text") and part.text
+    ]
+    return "\n".join(text_parts)
 
 # ------------------------------------------------------------------ anthropic
 
@@ -86,25 +211,65 @@ def _ensure_anthropic_client():
         creds = _load_credentials()
         _anthropic_client = anthropic.Anthropic(api_key=creds["anthropic"]["api_key"])
 
-def _send_anthropic(md_content: str, user_message: str) -> str:
+def _send_anthropic(md_content: str, user_message: str, base_dir: str) -> str:
     global _anthropic_history
+    import anthropic
+
     _ensure_anthropic_client()
+
     full_message = f"<context>\n{md_content}\n</context>\n\n{user_message}"
     _anthropic_history.append({"role": "user", "content": full_message})
-    response = _anthropic_client.messages.create(
+
-        model=_model,
+    for _ in range(MAX_TOOL_ROUNDS):
-        max_tokens=8096,
+        response = _anthropic_client.messages.create(
-        messages=_anthropic_history
+            model=_model,
-    )
+            max_tokens=8096,
-    reply = response.content[0].text
+            tools=_ANTHROPIC_TOOLS,
-    _anthropic_history.append({"role": "assistant", "content": reply})
+            messages=_anthropic_history
-    return reply
+        )
+
+        # Always record the assistant turn
+        _anthropic_history.append({
+            "role": "assistant",
+            "content": response.content
+        })
+
+        if response.stop_reason != "tool_use":
+            break
+
+        # Process tool calls
+        tool_results = []
+        for block in response.content:
+            if block.type == "tool_use" and block.name == TOOL_NAME:
+                script = block.input.get("script", "")
+                output = _run_script(script, base_dir)
+                tool_results.append({
+                    "type": "tool_result",
+                    "tool_use_id": block.id,
+                    "content": output
+                })
+
+        if not tool_results:
+            break
+
+        _anthropic_history.append({
+            "role": "user",
+            "content": tool_results
+        })
+
+    # Extract final text
+    text_parts = [
+        block.text
+        for block in response.content
+        if hasattr(block, "text") and block.text
+    ]
+    return "\n".join(text_parts)
 
 # ------------------------------------------------------------------ unified send
 
-def send(md_content: str, user_message: str) -> str:
+def send(md_content: str, user_message: str, base_dir: str = ".") -> str:
     if _provider == "gemini":
-        return _send_gemini(md_content, user_message)
+        return _send_gemini(md_content, user_message, base_dir)
     elif _provider == "anthropic":
-        return _send_anthropic(md_content, user_message)
+        return _send_anthropic(md_content, user_message, base_dir)
-    raise ValueError(f"unknown provider: {_provider}")
+    raise ValueError(f"unknown provider: {_provider}")

config.toml

@@ -5,13 +5,13 @@ output_dir = "./md_gen"
 [files]
 base_dir = "C:/projects/manual_slop"
 paths = [
-	"config.toml",
+    "config.toml",
-	"ai_client.py",
+    "ai_client.py",
-	"aggregate.py",
+    "aggregate.py",
-	"gemini.py",
+    "gemini.py",
-	"gui.py",
+    "gui.py",
-	"pyproject.toml",
+    "pyproject.toml",
-	"MainContext.md"
+    "MainContext.md",
 ]
 
 [screenshots]

gui.py

@@ -1,4 +1,3 @@
-# gui.py
 import dearpygui.dearpygui as dpg
 import tomllib
 import tomli_w
@@ -7,30 +6,108 @@ from pathlib import Path
 from tkinter import filedialog, Tk
 import aggregate
 import ai_client
+import shell_runner
 
 CONFIG_PATH = Path("config.toml")
 PROVIDERS = ["gemini", "anthropic"]
 
+
 def load_config() -> dict:
     with open(CONFIG_PATH, "rb") as f:
         return tomllib.load(f)
 
+
 def save_config(config: dict):
     with open(CONFIG_PATH, "wb") as f:
         tomli_w.dump(config, f)
 
+
 def hide_tk_root() -> Tk:
     root = Tk()
     root.withdraw()
     root.wm_attributes("-topmost", True)
     return root
 
+
+class ConfirmDialog:
+    """
+    Modal confirmation window for a proposed PowerShell script.
+    Background thread calls wait(), which blocks on a threading.Event.
+    Main render loop detects _pending_dialog and calls show() on the next frame.
+    User clicks Approve or Reject, which sets the event and unblocks the thread.
+    """
+
+    _next_id = 0
+
+    def __init__(self, script: str, base_dir: str):
+        ConfirmDialog._next_id += 1
+        self._uid = ConfirmDialog._next_id
+        self._tag = f"confirm_dlg_{self._uid}"
+        self._script = script
+        self._base_dir = base_dir
+        self._event = threading.Event()
+        self._approved = False
+
+    def show(self):
+        """Called from main thread only."""
+        w, h = 700, 440
+        vp_w = dpg.get_viewport_width()
+        vp_h = dpg.get_viewport_height()
+        px = max(0, (vp_w - w) // 2)
+        py = max(0, (vp_h - h) // 2)
+
+        with dpg.window(
+            label=f"Approve PowerShell Command #{self._uid}",
+            tag=self._tag,
+            modal=True,
+            no_close=True,
+            pos=(px, py),
+            width=w,
+            height=h,
+        ):
+            dpg.add_text("The AI wants to run the following PowerShell script:")
+            dpg.add_text(f"base_dir: {self._base_dir}", color=(200, 200, 100))
+            dpg.add_separator()
+            dpg.add_input_text(
+                tag=f"{self._tag}_script",
+                default_value=self._script,
+                multiline=True,
+                width=-1,
+                height=-72,
+                readonly=False,
+            )
+            dpg.add_separator()
+            with dpg.group(horizontal=True):
+                dpg.add_button(label="Approve & Run", callback=self._cb_approve)
+                dpg.add_button(label="Reject", callback=self._cb_reject)
+
+    def _cb_approve(self):
+        self._script = dpg.get_value(f"{self._tag}_script")
+        self._approved = True
+        self._event.set()
+        dpg.delete_item(self._tag)
+
+    def _cb_reject(self):
+        self._approved = False
+        self._event.set()
+        dpg.delete_item(self._tag)
+
+    def wait(self) -> tuple[bool, str]:
+        """Called from background thread. Blocks until user acts."""
+        self._event.wait()
+        return self._approved, self._script
+
+
 class App:
     def __init__(self):
         self.config = load_config()
         self.files: list[str] = list(self.config["files"].get("paths", []))
-        self.screenshots: list[str] = list(self.config.get("screenshots", {}).get("paths", []))
+        self.screenshots: list[str] = list(
-        self.history: list[str] = list(self.config.get("discussion", {}).get("history", []))
+            self.config.get("screenshots", {}).get("paths", [])
+        )
+        self.history: list[str] = list(
+            self.config.get("discussion", {}).get("history", [])
+        )
 
         ai_cfg = self.config.get("ai", {})
         self.current_provider: str = ai_cfg.get("provider", "gemini")
@@ -44,9 +121,63 @@ class App:
         self.send_thread: threading.Thread | None = None
         self.models_thread: threading.Thread | None = None
 
-        ai_client.set_provider(self.current_provider, self.current_model)
+        self._pending_dialog: ConfirmDialog | None = None
+        self._pending_dialog_lock = threading.Lock()
 
-    # ------------------------------------------------------------------ helpers
+        self._tool_log: list[tuple[str, str]] = []
+
+        ai_client.set_provider(self.current_provider, self.current_model)
+        ai_client.confirm_and_run_callback = self._confirm_and_run
+
+    # ---------------------------------------------------------------- tool execution
+
+    def _confirm_and_run(self, script: str, base_dir: str) -> str | None:
+        dialog = ConfirmDialog(script, base_dir)
+
+        with self._pending_dialog_lock:
+            self._pending_dialog = dialog
+
+        approved, final_script = dialog.wait()
+
+        if not approved:
+            self._append_tool_log(final_script, "REJECTED by user")
+            return None
+
+        self._update_status("running powershell...")
+        output = shell_runner.run_powershell(final_script, base_dir)
+        self._append_tool_log(final_script, output)
+        self._update_status("powershell done, awaiting AI...")
+        return output
+
+    def _append_tool_log(self, script: str, result: str):
+        self._tool_log.append((script, result))
+        self._rebuild_tool_log()
+
+    def _rebuild_tool_log(self):
+        if not dpg.does_item_exist("tool_log_scroll"):
+            return
+        dpg.delete_item("tool_log_scroll", children_only=True)
+        for i, (script, result) in enumerate(self._tool_log, 1):
+            with dpg.group(parent="tool_log_scroll"):
+                dpg.add_text(f"Call #{i}", color=(140, 200, 255))
+                dpg.add_input_text(
+                    default_value=script,
+                    multiline=True,
+                    readonly=True,
+                    width=-1,
+                    height=72,
+                )
+                dpg.add_text("Result:", color=(180, 255, 180))
+                dpg.add_input_text(
+                    default_value=result,
+                    multiline=True,
+                    readonly=True,
+                    width=-1,
+                    height=72,
+                )
+                dpg.add_separator()
+
+    # ---------------------------------------------------------------- helpers
 
     def _flush_to_config(self):
         self.config["output"]["namespace"] = dpg.get_value("namespace")
@@ -62,7 +193,7 @@ class App:
         self.config["discussion"] = {"history": self.history}
         self.config["ai"] = {
             "provider": self.current_provider,
-            "model": self.current_model
+            "model": self.current_model,
         }
 
     def _do_generate(self) -> tuple[str, Path]:
@@ -87,9 +218,7 @@ class App:
         for i, f in enumerate(self.files):
             with dpg.group(horizontal=True, parent="files_scroll"):
                 dpg.add_button(
-                    label="x",
+                    label="x", width=24, callback=self._make_remove_file_cb(i)
-                    width=24,
-                    callback=self._make_remove_file_cb(i)
                 )
                 dpg.add_text(f)
 
@@ -100,9 +229,7 @@ class App:
         for i, s in enumerate(self.screenshots):
             with dpg.group(horizontal=True, parent="shots_scroll"):
                 dpg.add_button(
-                    label="x",
+                    label="x", width=24, callback=self._make_remove_shot_cb(i)
-                    width=24,
-                    callback=self._make_remove_shot_cb(i)
                 )
                 dpg.add_text(s)
 
@@ -133,6 +260,7 @@ class App:
 
     def _fetch_models(self, provider: str):
         self._update_status("fetching models...")
+
         def do_fetch():
             try:
                 models = ai_client.list_models(provider)
@@ -141,6 +269,7 @@ class App:
                 self._update_status(f"models loaded: {len(models)}")
             except Exception as e:
                 self._update_status(f"model fetch error: {e}")
+
         self.models_thread = threading.Thread(target=do_fetch, daemon=True)
         self.models_thread.start()
 
@@ -193,7 +322,10 @@ class App:
         root = hide_tk_root()
         paths = filedialog.askopenfilenames(
             title="Select Screenshots",
-            filetypes=[("Images", "*.png *.jpg *.jpeg *.gif *.bmp *.webp"), ("All", "*.*")]
+            filetypes=[
+                ("Images", "*.png *.jpg *.jpeg *.gif *.bmp *.webp"),
+                ("All", "*.*"),
+            ],
         )
         root.destroy()
         for p in paths:
@@ -224,6 +356,8 @@ class App:
 
     def cb_reset_session(self):
         ai_client.reset_session()
+        self._tool_log.clear()
+        self._rebuild_tool_log()
         self._update_status("session reset")
         self._update_response("")
 
@@ -237,12 +371,14 @@ class App:
         except Exception as e:
             self._update_status(f"generate error: {e}")
             return
+
         self._update_status("sending...")
         user_msg = dpg.get_value("ai_input")
+        base_dir = dpg.get_value("files_base_dir")
 
         def do_send():
             try:
-                response = ai_client.send(self.last_md, user_msg)
+                response = ai_client.send(self.last_md, user_msg, base_dir)
                 self._update_response(response)
                 self._update_status("done")
             except Exception as e:
@@ -270,6 +406,10 @@ class App:
     def cb_fetch_models(self):
         self._fetch_models(self.current_provider)
 
+    def cb_clear_tool_log(self):
+        self._tool_log.clear()
+        self._rebuild_tool_log()
+
     # ---------------------------------------------------------------- build ui
 
     def _build_ui(self):
@@ -280,19 +420,19 @@ class App:
             pos=(8, 8),
             width=400,
             height=200,
-            no_close=True
+            no_close=True,
         ):
             dpg.add_text("Namespace")
             dpg.add_input_text(
                 tag="namespace",
                 default_value=self.config["output"]["namespace"],
-                width=-1
+                width=-1,
             )
             dpg.add_text("Output Dir")
             dpg.add_input_text(
                 tag="output_dir",
                 default_value=self.config["output"]["output_dir"],
-                width=-1
+                width=-1,
             )
             with dpg.group(horizontal=True):
                 dpg.add_button(label="Browse Output Dir", callback=self.cb_browse_output)
@@ -304,16 +444,18 @@ class App:
             pos=(8, 216),
             width=400,
             height=500,
-            no_close=True
+            no_close=True,
         ):
             dpg.add_text("Base Dir")
             with dpg.group(horizontal=True):
                 dpg.add_input_text(
                     tag="files_base_dir",
                     default_value=self.config["files"]["base_dir"],
-                    width=-220
+                    width=-220,
+                )
+                dpg.add_button(
+                    label="Browse##filesbase", callback=self.cb_browse_files_base
                 )
-                dpg.add_button(label="Browse##filesbase", callback=self.cb_browse_files_base)
             dpg.add_separator()
             dpg.add_text("Paths")
             with dpg.child_window(tag="files_scroll", height=-64, border=True):
@@ -330,16 +472,18 @@ class App:
             pos=(416, 8),
             width=400,
             height=500,
-            no_close=True
+            no_close=True,
         ):
             dpg.add_text("Base Dir")
             with dpg.group(horizontal=True):
                 dpg.add_input_text(
                     tag="shots_base_dir",
                     default_value=self.config.get("screenshots", {}).get("base_dir", "."),
-                    width=-220
+                    width=-220,
+                )
+                dpg.add_button(
+                    label="Browse##shotsbase", callback=self.cb_browse_shots_base
                 )
-                dpg.add_button(label="Browse##shotsbase", callback=self.cb_browse_shots_base)
             dpg.add_separator()
             dpg.add_text("Paths")
             with dpg.child_window(tag="shots_scroll", height=-48, border=True):
@@ -354,14 +498,14 @@ class App:
             pos=(824, 8),
             width=400,
             height=500,
-            no_close=True
+            no_close=True,
         ):
             dpg.add_input_text(
                 tag="discussion_box",
                 default_value="\n---\n".join(self.history),
                 multiline=True,
                 width=-1,
-                height=-64
+                height=-64,
             )
             dpg.add_separator()
             with dpg.group(horizontal=True):
@@ -375,7 +519,7 @@ class App:
             pos=(1232, 8),
             width=420,
             height=280,
-            no_close=True
+            no_close=True,
         ):
             dpg.add_text("Provider")
             dpg.add_combo(
@@ -383,7 +527,7 @@ class App:
                 items=PROVIDERS,
                 default_value=self.current_provider,
                 width=-1,
-                callback=self.cb_provider_changed
+                callback=self.cb_provider_changed,
             )
             dpg.add_separator()
             with dpg.group(horizontal=True):
@@ -395,7 +539,7 @@ class App:
                 default_value=self.current_model,
                 width=-1,
                 num_items=6,
-                callback=self.cb_model_changed
+                callback=self.cb_model_changed,
             )
             dpg.add_separator()
             dpg.add_text("Status: idle", tag="ai_status")
@@ -406,13 +550,13 @@ class App:
             pos=(1232, 296),
             width=420,
             height=280,
-            no_close=True
+            no_close=True,
         ):
             dpg.add_input_text(
                 tag="ai_input",
                 multiline=True,
                 width=-1,
-                height=-64
+                height=-64,
             )
             dpg.add_separator()
             with dpg.group(horizontal=True):
@@ -425,21 +569,36 @@ class App:
             tag="win_response",
             pos=(1232, 584),
             width=420,
-            height=400,
+            height=300,
-            no_close=True
+            no_close=True,
         ):
             dpg.add_input_text(
                 tag="ai_response",
                 multiline=True,
                 readonly=True,
                 width=-1,
-                height=-1
+                height=-1,
             )
 
+        with dpg.window(
+            label="Tool Calls",
+            tag="win_tool_log",
+            pos=(1232, 892),
+            width=420,
+            height=300,
+            no_close=True,
+        ):
+            with dpg.group(horizontal=True):
+                dpg.add_text("Tool call history")
+                dpg.add_button(label="Clear", callback=self.cb_clear_tool_log)
+            dpg.add_separator()
+            with dpg.child_window(tag="tool_log_scroll", height=-1, border=False):
+                pass
+
     def run(self):
         dpg.create_context()
         dpg.configure_app(docking=True, docking_space=True)
-        dpg.create_viewport(title="manual slop", width=1600, height=900)
+        dpg.create_viewport(title="manual slop", width=1680, height=1200)
         dpg.setup_dearpygui()
         dpg.show_viewport()
         dpg.maximize_viewport()
@@ -447,6 +606,13 @@ class App:
         self._fetch_models(self.current_provider)
 
         while dpg.is_dearpygui_running():
+            # Show any pending confirmation dialog on the main thread
+            with self._pending_dialog_lock:
+                dialog = self._pending_dialog
+                self._pending_dialog = None
+            if dialog is not None:
+                dialog.show()
+
             dpg.render_dearpygui_frame()
 
         dpg.destroy_context()
@@ -458,4 +624,4 @@ def main():
 
 
 if __name__ == "__main__":
-    main()
+    main()

shell_runner.py

@@ -0,0 +1,36 @@
+import subprocess
+import shlex
+from pathlib import Path
+
+TIMEOUT_SECONDS = 60
+
+def run_powershell(script: str, base_dir: str) -> str:
+    """
+    Run a PowerShell script with working directory set to base_dir.
+    Returns a string combining stdout, stderr, and exit code.
+    Raises nothing - all errors are captured into the return string.
+    """
+    # Prepend Set-Location so the AI doesn't need to worry about cwd
+    full_script = f"Set-Location -LiteralPath '{base_dir}'\n{script}"
+
+    try:
+        result = subprocess.run(
+            ["powershell", "-NoProfile", "-NonInteractive", "-Command", full_script],
+            capture_output=True,
+            text=True,
+            timeout=TIMEOUT_SECONDS,
+            cwd=base_dir
+        )
+        parts = []
+        if result.stdout.strip():
+            parts.append(f"STDOUT:\n{result.stdout.strip()}")
+        if result.stderr.strip():
+            parts.append(f"STDERR:\n{result.stderr.strip()}")
+        parts.append(f"EXIT CODE: {result.returncode}")
+        return "\n".join(parts) if parts else f"EXIT CODE: {result.returncode}"
+    except subprocess.TimeoutExpired:
+        return f"ERROR: command timed out after {TIMEOUT_SECONDS}s"
+    except FileNotFoundError:
+        return "ERROR: powershell executable not found"
+    except Exception as e:
+        return f"ERROR: {e}"