feat: enhance agent configuration and model management with per-agent overrides

app/core/agent_runner.py

@@ -43,10 +43,9 @@ from app.agents.note_agent import NOTE_TOOLS
 from app.agents.project_agent import PROJECT_TOOLS
 from app.agents.task_agent import TASK_TOOLS
 from app.agents.timeline_agent import TIMELINE_TOOLS
-from app.config.settings import settings
 from app.core.device_manager import DeviceConnectionManager
 from app.core.langfuse_client import compile_prompt, extract_usage, get_langfuse, get_prompt_or_fallback
-from app.core.llm import get_llm
+from app.core.llm import get_agent_llm, model_for_agent
 from app.core.preprocessors import detect_content_type, preprocess
 from app.core.ws_context import clear_client_executor, execute_on_client, set_client_executor
 from app.db import async_session
@@ -74,13 +73,13 @@ _MAX_PROCESSING_STEPS: int = 12
 _MAX_SCAN_DEPTH: int = 5
 
 # ── Data-type to tool mapping ─────────────────────────────────────────────
-# NOTE: "projects" is intentionally excluded — project creation/assignment is
-# handled in code by the runner, never delegated to the Step 2 LLM.
 
 _DATA_TYPE_TOOLS: dict[str, list[Any]] = {
     "tasks": TASK_TOOLS,
     "notes": NOTE_TOOLS,
     "timelines": TIMELINE_TOOLS,
+    "timelineEvents": TIMELINE_TOOLS,
+    "projects": PROJECT_TOOLS,
 }
 
 # ── V2: Unified processing prompt (hot-swappable via Langfuse "unified_processing") ──
@@ -238,7 +237,7 @@ async def _run_agent_with_tools(
     run is appended to it (used by the caller to count ``create_*`` calls).
     """
     lf = get_langfuse()
-    llm = get_llm()
+    llm = get_agent_llm(agent_name)
     llm_with_tools = llm.bind_tools(tools)
     messages: list[Any] = [
         SystemMessage(content=system_prompt),
@@ -264,7 +263,7 @@ async def _run_agent_with_tools(
                 lf.start_as_current_observation(
                     as_type="generation",
                     name=f"{agent_name}-llm",
-                    model=settings.LLM_MODEL,
+                    model=model_for_agent(agent_name),
                     prompt=langfuse_prompt,
                     input=messages,
                 )
@@ -696,6 +695,12 @@ async def run_local_agent(
                 )
                 items_created += file_created
 
+                # Refresh project list when a project was created so
+                # subsequent files see it in the prompt context.
+                if "create_project" in file_tool_calls:
+                    projects = await _fetch_projects()
+                    projects_block = _format_projects(projects)
+
                 logger.info(
                     "agent_runner: run=%s file=%r created=%d result=%s",
                     run_id, file_path, file_created, result_text[:200],