refactor(ai_client): Add strict type hints to global variables

ai_client.py

@@ -40,7 +40,7 @@ _max_tokens: int = 8192
 _history_trunc_limit: int = 8000
 
 # Global event emitter for API lifecycle events
-events = EventEmitter()
+events: EventEmitter = EventEmitter()
 
 def set_model_params(temp: float, max_tok: int, trunc_limit: int = 8000) -> None:
  global _temperature, _max_tokens, _history_trunc_limit
@@ -55,52 +55,52 @@ def set_history_trunc_limit(val: int) -> None:
  global _history_trunc_limit
  _history_trunc_limit = val
 
-_gemini_client = None
+_gemini_client: genai.Client | None = None
-_gemini_chat = None
+_gemini_chat: Any = None
-_gemini_cache = None
+_gemini_cache: Any = None
 _gemini_cache_md_hash: int | None = None
 _gemini_cache_created_at: float | None = None
 
 # Gemini cache TTL in seconds. Caches are created with this TTL and
 # proactively rebuilt at 90% of this value to avoid stale-reference errors.
-_GEMINI_CACHE_TTL = 3600
+_GEMINI_CACHE_TTL: int = 3600
 
-_anthropic_client = None
+_anthropic_client: anthropic.Anthropic | None = None
 _anthropic_history: list[dict] = []
-_anthropic_history_lock = threading.Lock()
+_anthropic_history_lock: threading.Lock = threading.Lock()
 
-_deepseek_client = None
+_deepseek_client: Any = None
 _deepseek_history: list[dict] = []
-_deepseek_history_lock = threading.Lock()
+_deepseek_history_lock: threading.Lock = threading.Lock()
 
-_send_lock = threading.Lock()
+_send_lock: threading.Lock = threading.Lock()
 
-_gemini_cli_adapter = None
+_gemini_cli_adapter: GeminiCliAdapter | None = 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
+confirm_and_run_callback: Callable[[str, str], str | None] | None = None
 
 # Injected by gui.py - called whenever a comms entry is appended.
 # Signature: (entry: dict) -> None
-comms_log_callback = None
+comms_log_callback: Callable[[dict[str, Any]], None] | None = None
 
 # Injected by gui.py - called whenever a tool call completes.
 # Signature: (script: str, result: str) -> None
-tool_log_callback = None
+tool_log_callback: Callable[[str, str], None] | None = None
 
 # Increased to allow thorough code exploration before forcing a summary
-MAX_TOOL_ROUNDS = 10
+MAX_TOOL_ROUNDS: int = 10
 
 # Maximum cumulative bytes of tool output allowed per send() call.
 # Prevents unbounded memory growth during long tool-calling loops.
-_MAX_TOOL_OUTPUT_BYTES = 500_000
+_MAX_TOOL_OUTPUT_BYTES: int = 500_000
 
 # Maximum characters per text chunk sent to Anthropic.
 # Kept well under the ~200k token API limit.
-_ANTHROPIC_CHUNK_SIZE = 120_000
+_ANTHROPIC_CHUNK_SIZE: int = 120_000
 
-_SYSTEM_PROMPT = (
+_SYSTEM_PROMPT: str = (
  "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. "
@@ -130,7 +130,7 @@ def _get_combined_system_prompt() -> str:
 
 _comms_log: list[dict] = []
 
-COMMS_CLAMP_CHARS = 300
+COMMS_CLAMP_CHARS: int = 300
 
 def _append_comms(direction: str, kind: str, payload: dict[str, Any]) -> None:
  entry = {
@@ -379,7 +379,7 @@ def _list_deepseek_models(api_key: str) -> list[str]:
  return ["deepseek-chat", "deepseek-reasoner", "deepseek-v3", "deepseek-r1"]
  # ------------------------------------------------------------------ tool definition
 
-TOOL_NAME = "run_powershell"
+TOOL_NAME: str = "run_powershell"
 
 _agent_tools: dict = {}
 
@@ -427,9 +427,9 @@ def _build_anthropic_tools() -> list[dict]:
   tools_list[-1]["cache_control"] = {"type": "ephemeral"}
  return tools_list
 
-_ANTHROPIC_TOOLS = _build_anthropic_tools()
+_ANTHROPIC_TOOLS: list[dict[str, Any]] = _build_anthropic_tools()
 
-_CACHED_ANTHROPIC_TOOLS = None
+_CACHED_ANTHROPIC_TOOLS: list[dict[str, Any]] | None = None
 
 def _get_anthropic_tools() -> list[dict]:
  """Return the Anthropic tools list, rebuilding only once per session."""
@@ -550,7 +550,7 @@ def _build_file_context_text(file_items: list[dict]) -> str:
   parts.append(f"### `{path}`\n\n```{suffix}\n{content}\n```")
  return "\n\n---\n\n".join(parts)
 
-_DIFF_LINE_THRESHOLD = 200
+_DIFF_LINE_THRESHOLD: int = 200
 
 def _build_file_diff_text(changed_items: list[dict]) -> str:
  """
@@ -933,18 +933,18 @@ def _send_gemini_cli(md_content: str, user_message: str, base_dir: str,
   # ------------------------------------------------------------------ anthropic history management
   # Rough chars-per-token ratio. Anthropic tokeniser averages ~3.5-4 chars/token.
   # We use 3.5 to be conservative (overestimate token count = safer).
-_CHARS_PER_TOKEN = 3.5
+_CHARS_PER_TOKEN: float = 3.5
 
 # Maximum token budget for the entire prompt (system + tools + messages).
 # Anthropic's limit is 200k. We leave headroom for the response + tool schemas.
-_ANTHROPIC_MAX_PROMPT_TOKENS = 180_000
+_ANTHROPIC_MAX_PROMPT_TOKENS: int = 180_000
 
 # Gemini models have a 1M context window but we cap well below to leave headroom.
 # If the model reports input tokens exceeding this, we trim old history.
-_GEMINI_MAX_INPUT_TOKENS = 900_000
+_GEMINI_MAX_INPUT_TOKENS: int = 900_000
 
 # Marker prefix used to identify stale file-refresh injections in history
-_FILE_REFRESH_MARKER = "[FILES UPDATED"
+_FILE_REFRESH_MARKER: str = "[FILES UPDATED"
 
 def _estimate_message_tokens(msg: dict) -> int:
  """