Merge branch 'main' into kimi/issue-490

Co-authored-by: Kimi Agent <kimi@timmy.local>
Co-committed-by: Kimi Agent <kimi@timmy.local>

config/providers.yaml

@@ -54,19 +54,6 @@ providers:
         context_window: 2048
         capabilities: [text, vision, streaming]
     
-  # Secondary: Local AirLLM (if installed)
-  - name: airllm-local
-    type: airllm
-    enabled: false  # Enable if pip install airllm
-    priority: 2
-    models:
-      - name: 70b
-        default: true
-        capabilities: [text, tools, json, streaming]
-      - name: 8b
-        capabilities: [text, tools, json, streaming]
-      - name: 405b
-        capabilities: [text, tools, json, streaming]
     
   # Tertiary: OpenAI (if API key available)
   - name: openai-backup

src/config.py

@@ -64,17 +64,10 @@ class Settings(BaseSettings):
     # Seconds to wait for user confirmation before auto-rejecting.
     discord_confirm_timeout: int = 120
 
-    # ── AirLLM / backend selection ───────────────────────────────────────────
+    # ── Backend selection ────────────────────────────────────────────────────
     # "ollama"  — always use Ollama (default, safe everywhere)
-    # "airllm"  — always use AirLLM (requires pip install ".[bigbrain]")
-    # "auto"    — use AirLLM on Apple Silicon if airllm is installed,
-    #             fall back to Ollama otherwise
-    timmy_model_backend: Literal["ollama", "airllm", "grok", "claude", "auto"] = "ollama"
-
-    # AirLLM model size when backend is airllm or auto.
-    # Larger = smarter, but needs more RAM / disk.
-    # 8b  ~16 GB  |  70b  ~140 GB  |  405b  ~810 GB
-    airllm_model_size: Literal["8b", "70b", "405b"] = "70b"
+    # "auto"    — pick best available local backend, fall back to Ollama
+    timmy_model_backend: Literal["ollama", "grok", "claude", "auto"] = "ollama"
 
     # ── Grok (xAI) — opt-in premium cloud backend ────────────────────────
     # Grok is a premium augmentation layer — local-first ethos preserved.
@@ -138,7 +131,12 @@ class Settings(BaseSettings):
 
     # CORS allowed origins for the web chat interface (Gitea Pages, etc.)
     # Set CORS_ORIGINS as a comma-separated list, e.g. "http://localhost:3000,https://example.com"
-    cors_origins: list[str] = ["*"]
+    cors_origins: list[str] = [
+        "http://localhost:3000",
+        "http://localhost:8000",
+        "http://127.0.0.1:3000",
+        "http://127.0.0.1:8000",
+    ]
 
     # Trusted hosts for the Host header check (TrustedHostMiddleware).
     # Set TRUSTED_HOSTS as a comma-separated list. Wildcards supported (e.g. "*.ts.net").
@@ -471,8 +469,19 @@ def validate_startup(*, force: bool = False) -> None:
                 ", ".join(_missing),
             )
             sys.exit(1)
+        if "*" in settings.cors_origins:
+            _startup_logger.error(
+                "PRODUCTION SECURITY ERROR: CORS wildcard '*' is not allowed "
+                "in production. Set CORS_ORIGINS to explicit origins."
+            )
+            sys.exit(1)
         _startup_logger.info("Production mode: security secrets validated ✓")
     else:
+        if "*" in settings.cors_origins:
+            _startup_logger.warning(
+                "SEC: CORS_ORIGINS contains wildcard '*' — "
+                "restrict to explicit origins before deploying to production."
+            )
         if not settings.l402_hmac_secret:
             _startup_logger.warning(
                 "SEC: L402_HMAC_SECRET is not set — "

src/dashboard/app.py

@@ -484,15 +484,14 @@ app = FastAPI(
 
 
 def _get_cors_origins() -> list[str]:
-    """Get CORS origins from settings, with sensible defaults."""
+    """Get CORS origins from settings, rejecting wildcards in production."""
     origins = settings.cors_origins
-    if settings.debug and origins == ["*"]:
-        return [
-            "http://localhost:3000",
-            "http://localhost:8000",
-            "http://127.0.0.1:3000",
-            "http://127.0.0.1:8000",
-        ]
+    if "*" in origins and not settings.debug:
+        logger.warning(
+            "Wildcard '*' in CORS_ORIGINS stripped in production — "
+            "set explicit origins via CORS_ORIGINS env var"
+        )
+        origins = [o for o in origins if o != "*"]
     return origins
 
 

src/dashboard/middleware/csrf.py

@@ -100,7 +100,7 @@ class CSRFMiddleware(BaseHTTPMiddleware):
             ...
 
     Usage:
-        app.add_middleware(CSRFMiddleware, secret="your-secret-key")
+        app.add_middleware(CSRFMiddleware, secret=settings.csrf_secret)
 
     Attributes:
         secret: Secret key for token signing (optional, for future use).

src/infrastructure/router/api.py

@@ -183,6 +183,22 @@ async def run_health_check(
     }
 
 
+@router.post("/reload")
+async def reload_config(
+    cascade: Annotated[CascadeRouter, Depends(get_cascade_router)],
+) -> dict[str, Any]:
+    """Hot-reload providers.yaml without restart.
+
+    Preserves circuit breaker state and metrics for existing providers.
+    """
+    try:
+        result = cascade.reload_config()
+        return {"status": "ok", **result}
+    except Exception as exc:
+        logger.error("Config reload failed: %s", exc)
+        raise HTTPException(status_code=500, detail=f"Reload failed: {exc}") from exc
+
+
 @router.get("/config")
 async def get_config(
     cascade: Annotated[CascadeRouter, Depends(get_cascade_router)],

src/infrastructure/router/cascade.py

@@ -18,6 +18,8 @@ from enum import Enum
 from pathlib import Path
 from typing import Any
 
+from config import settings
+
 try:
     import yaml
 except ImportError:
@@ -100,7 +102,7 @@ class Provider:
     """LLM provider configuration and state."""
 
     name: str
-    type: str  # ollama, openai, anthropic, airllm
+    type: str  # ollama, openai, anthropic
     enabled: bool
     priority: int
     url: str | None = None
@@ -301,22 +303,13 @@ class CascadeRouter:
                 # Can't check without requests, assume available
                 return True
             try:
-                url = provider.url or "http://localhost:11434"
+                url = provider.url or settings.ollama_url
                 response = requests.get(f"{url}/api/tags", timeout=5)
                 return response.status_code == 200
             except Exception as exc:
                 logger.debug("Ollama provider check error: %s", exc)
                 return False
 
-        elif provider.type == "airllm":
-            # Check if airllm is installed
-            try:
-                import importlib.util
-
-                return importlib.util.find_spec("airllm") is not None
-            except (ImportError, ModuleNotFoundError):
-                return False
-
         elif provider.type in ("openai", "anthropic", "grok"):
             # Check if API key is set
             return provider.api_key is not None and provider.api_key != ""
@@ -815,6 +808,66 @@ class CascadeRouter:
         provider.status = ProviderStatus.HEALTHY
         logger.info("Circuit breaker CLOSED for %s", provider.name)
 
+    def reload_config(self) -> dict:
+        """Hot-reload providers.yaml, preserving runtime state.
+
+        Re-reads the config file, rebuilds the provider list, and
+        preserves circuit breaker state and metrics for providers
+        that still exist after reload.
+
+        Returns:
+            Summary dict with added/removed/preserved counts.
+        """
+        # Snapshot current runtime state keyed by provider name
+        old_state: dict[
+            str, tuple[ProviderMetrics, CircuitState, float | None, int, ProviderStatus]
+        ] = {}
+        for p in self.providers:
+            old_state[p.name] = (
+                p.metrics,
+                p.circuit_state,
+                p.circuit_opened_at,
+                p.half_open_calls,
+                p.status,
+            )
+
+        old_names = set(old_state.keys())
+
+        # Reload from disk
+        self.providers = []
+        self._load_config()
+
+        # Restore preserved state
+        new_names = {p.name for p in self.providers}
+        preserved = 0
+        for p in self.providers:
+            if p.name in old_state:
+                metrics, circuit, opened_at, half_open, status = old_state[p.name]
+                p.metrics = metrics
+                p.circuit_state = circuit
+                p.circuit_opened_at = opened_at
+                p.half_open_calls = half_open
+                p.status = status
+                preserved += 1
+
+        added = new_names - old_names
+        removed = old_names - new_names
+
+        logger.info(
+            "Config reloaded: %d providers (%d preserved, %d added, %d removed)",
+            len(self.providers),
+            preserved,
+            len(added),
+            len(removed),
+        )
+
+        return {
+            "total_providers": len(self.providers),
+            "preserved": preserved,
+            "added": sorted(added),
+            "removed": sorted(removed),
+        }
+
     def get_metrics(self) -> dict:
         """Get metrics for all providers."""
         return {

src/timmy/agent.py

@@ -220,7 +220,7 @@ def create_timmy(
     print_response(message, stream).
     """
     resolved = _resolve_backend(backend)
-    size = model_size or settings.airllm_model_size
+    size = model_size or "70b"
 
     if resolved == "claude":
         from timmy.backends import ClaudeBackend
@@ -302,7 +302,7 @@ def create_timmy(
                 memory_context = memory_context[:max_context] + "\n... [truncated]"
             full_prompt = (
                 f"{base_prompt}\n\n"
-                "GROUNDED CONTEXT (retrieved from memory — cite when used):\n\n"
+                f"## GROUNDED CONTEXT (verified sources — cite when using)\n\n"
                 f"{memory_context}"
             )
         else:

src/timmy/prompts.py

@@ -23,9 +23,9 @@ Rules:
 - Remember what the user tells you during the conversation.
 - If you don't know something, say so honestly — never fabricate facts.
 - If a request is ambiguous, ask a brief clarifying question before guessing.
-- Source distinction: when using information from your memory context, cite it
-  ("From my memory...", "You mentioned..."). When guessing, hedge ("I think...",
-  "My sense is..."). Never present inferred claims as grounded facts.
+- SOURCE DISTINCTION: When answering from memory or retrieved context, cite it.
+  When answering from your own training, use hedging: "I think", "I believe".
+  The user must be able to tell grounded claims from pattern-matching.
 - Use the user's name if you know it.
 - When you state a fact, commit to it.
 - NEVER attempt arithmetic in your head. If asked to compute anything, respond:
@@ -81,15 +81,17 @@ HONESTY:
 - Never fabricate tool output. Call the tool and wait.
 - If a tool errors, report the exact error.
 
-SOURCE DISTINCTION:
-- Every claim comes from one of two places: a grounded source you can point to,
-  or your own pattern-matching. The user must be able to tell which is which.
-- When your response uses information from GROUNDED CONTEXT (memory retrieval,
-  vault files, retrieved facts), say so: "From my memory..." or "You mentioned..."
-- When no grounded source exists, hedge appropriately: "I think...", "My sense
-  is...", or "I don't have a record of that, but..."
-- Never present inferred claims with the same confidence as grounded ones.
-- If asked where you got something, distinguish: retrieved vs. inferred.
+SOURCE DISTINCTION (SOUL requirement — non-negotiable):
+- Every claim you make comes from one of two places: a verified source you
+  can point to, or your own pattern-matching. The user must be able to tell
+  which is which.
+- When your response uses information from GROUNDED CONTEXT (memory, retrieved
+  documents, tool output), cite it: "From memory:", "According to [source]:".
+- When you are generating from your training data alone, signal it naturally:
+  "I think", "My understanding is", "I believe" — never false certainty.
+- If the user asks a factual question and you have no grounded source, say so:
+  "I don't have a verified source for this — from my training I think..."
+- Prefer "I don't know" over a confident-sounding guess. Refusal over fabrication.
 
 MEMORY (three tiers):
 - Tier 1: MEMORY.md (hot, always loaded)

src/timmy_serve/app.py

@@ -75,6 +75,8 @@ def create_timmy_serve_app() -> FastAPI:
     @asynccontextmanager
     async def lifespan(app: FastAPI):
         logger.info("Timmy Serve starting")
+        app.state.timmy = create_timmy()
+        logger.info("Timmy agent cached in app state")
         yield
         logger.info("Timmy Serve shutting down")
 
@@ -101,7 +103,7 @@ def create_timmy_serve_app() -> FastAPI:
     async def serve_chat(request: Request, body: ChatRequest):
         """Process a chat request."""
         try:
-            timmy = create_timmy()
+            timmy = request.app.state.timmy
             result = timmy.run(body.message, stream=False)
             response_text = result.content if hasattr(result, "content") else str(result)
 

tests/infrastructure/test_router_cascade.py

@@ -2,7 +2,7 @@
 
 import time
 from pathlib import Path
-from unittest.mock import AsyncMock, MagicMock, patch
+from unittest.mock import AsyncMock, patch
 
 import pytest
 import yaml
@@ -489,30 +489,182 @@ class TestProviderAvailabilityCheck:
 
         assert router._check_provider_available(provider) is False
 
-    def test_check_airllm_installed(self):
-        """Test AirLLM when installed."""
-        router = CascadeRouter(config_path=Path("/nonexistent"))
 
-        provider = Provider(
-            name="airllm",
-            type="airllm",
-            enabled=True,
-            priority=1,
+class TestCascadeRouterReload:
+    """Test hot-reload of providers.yaml."""
+
+    def test_reload_preserves_metrics(self, tmp_path):
+        """Test that reload preserves metrics for existing providers."""
+        config = {
+            "providers": [
+                {
+                    "name": "test-openai",
+                    "type": "openai",
+                    "enabled": True,
+                    "priority": 1,
+                    "api_key": "sk-test",
+                }
+            ],
+        }
+        config_path = tmp_path / "providers.yaml"
+        config_path.write_text(yaml.dump(config))
+
+        router = CascadeRouter(config_path=config_path)
+        assert len(router.providers) == 1
+
+        # Simulate some traffic
+        router._record_success(router.providers[0], 150.0)
+        router._record_success(router.providers[0], 250.0)
+        assert router.providers[0].metrics.total_requests == 2
+
+        # Reload
+        result = router.reload_config()
+
+        assert result["total_providers"] == 1
+        assert result["preserved"] == 1
+        assert result["added"] == []
+        assert result["removed"] == []
+        # Metrics survived
+        assert router.providers[0].metrics.total_requests == 2
+        assert router.providers[0].metrics.total_latency_ms == 400.0
+
+    def test_reload_preserves_circuit_breaker(self, tmp_path):
+        """Test that reload preserves circuit breaker state."""
+        config = {
+            "cascade": {"circuit_breaker": {"failure_threshold": 2}},
+            "providers": [
+                {
+                    "name": "test-openai",
+                    "type": "openai",
+                    "enabled": True,
+                    "priority": 1,
+                    "api_key": "sk-test",
+                }
+            ],
+        }
+        config_path = tmp_path / "providers.yaml"
+        config_path.write_text(yaml.dump(config))
+
+        router = CascadeRouter(config_path=config_path)
+
+        # Open circuit breaker
+        for _ in range(2):
+            router._record_failure(router.providers[0])
+        assert router.providers[0].circuit_state == CircuitState.OPEN
+
+        # Reload
+        router.reload_config()
+
+        # Circuit breaker state preserved
+        assert router.providers[0].circuit_state == CircuitState.OPEN
+        assert router.providers[0].status == ProviderStatus.UNHEALTHY
+
+    def test_reload_detects_added_provider(self, tmp_path):
+        """Test that reload detects newly added providers."""
+        config = {
+            "providers": [
+                {
+                    "name": "openai-1",
+                    "type": "openai",
+                    "enabled": True,
+                    "priority": 1,
+                    "api_key": "sk-test",
+                }
+            ],
+        }
+        config_path = tmp_path / "providers.yaml"
+        config_path.write_text(yaml.dump(config))
+
+        router = CascadeRouter(config_path=config_path)
+        assert len(router.providers) == 1
+
+        # Add a second provider to config
+        config["providers"].append(
+            {
+                "name": "anthropic-1",
+                "type": "anthropic",
+                "enabled": True,
+                "priority": 2,
+                "api_key": "sk-ant-test",
+            }
         )
+        config_path.write_text(yaml.dump(config))
 
-        with patch("importlib.util.find_spec", return_value=MagicMock()):
-            assert router._check_provider_available(provider) is True
+        result = router.reload_config()
 
-    def test_check_airllm_not_installed(self):
-        """Test AirLLM when not installed."""
-        router = CascadeRouter(config_path=Path("/nonexistent"))
+        assert result["total_providers"] == 2
+        assert result["preserved"] == 1
+        assert result["added"] == ["anthropic-1"]
+        assert result["removed"] == []
 
-        provider = Provider(
-            name="airllm",
-            type="airllm",
-            enabled=True,
-            priority=1,
-        )
+    def test_reload_detects_removed_provider(self, tmp_path):
+        """Test that reload detects removed providers."""
+        config = {
+            "providers": [
+                {
+                    "name": "openai-1",
+                    "type": "openai",
+                    "enabled": True,
+                    "priority": 1,
+                    "api_key": "sk-test",
+                },
+                {
+                    "name": "anthropic-1",
+                    "type": "anthropic",
+                    "enabled": True,
+                    "priority": 2,
+                    "api_key": "sk-ant-test",
+                },
+            ],
+        }
+        config_path = tmp_path / "providers.yaml"
+        config_path.write_text(yaml.dump(config))
 
-        with patch("importlib.util.find_spec", return_value=None):
-            assert router._check_provider_available(provider) is False
+        router = CascadeRouter(config_path=config_path)
+        assert len(router.providers) == 2
+
+        # Remove anthropic
+        config["providers"] = [config["providers"][0]]
+        config_path.write_text(yaml.dump(config))
+
+        result = router.reload_config()
+
+        assert result["total_providers"] == 1
+        assert result["preserved"] == 1
+        assert result["removed"] == ["anthropic-1"]
+
+    def test_reload_re_sorts_by_priority(self, tmp_path):
+        """Test that providers are re-sorted by priority after reload."""
+        config = {
+            "providers": [
+                {
+                    "name": "low-priority",
+                    "type": "openai",
+                    "enabled": True,
+                    "priority": 10,
+                    "api_key": "sk-test",
+                },
+                {
+                    "name": "high-priority",
+                    "type": "openai",
+                    "enabled": True,
+                    "priority": 1,
+                    "api_key": "sk-test2",
+                },
+            ],
+        }
+        config_path = tmp_path / "providers.yaml"
+        config_path.write_text(yaml.dump(config))
+
+        router = CascadeRouter(config_path=config_path)
+        assert router.providers[0].name == "high-priority"
+
+        # Swap priorities
+        config["providers"][0]["priority"] = 1
+        config["providers"][1]["priority"] = 10
+        config_path.write_text(yaml.dump(config))
+
+        router.reload_config()
+
+        assert router.providers[0].name == "low-priority"
+        assert router.providers[1].name == "high-priority"

tests/test_lazy_init.py

@@ -49,6 +49,34 @@ class TestConfigLazyValidation:
             # Should not raise
             validate_startup(force=True)
 
+    def test_validate_startup_exits_on_cors_wildcard_in_production(self):
+        """validate_startup() should exit in production when CORS has wildcard."""
+        from config import settings, validate_startup
+
+        with (
+            patch.object(settings, "timmy_env", "production"),
+            patch.object(settings, "l402_hmac_secret", "test-secret-hex-value-32"),
+            patch.object(settings, "l402_macaroon_secret", "test-macaroon-hex-value-32"),
+            patch.object(settings, "cors_origins", ["*"]),
+            pytest.raises(SystemExit),
+        ):
+            validate_startup(force=True)
+
+    def test_validate_startup_warns_cors_wildcard_in_dev(self):
+        """validate_startup() should warn in dev when CORS has wildcard."""
+        from config import settings, validate_startup
+
+        with (
+            patch.object(settings, "timmy_env", "development"),
+            patch.object(settings, "cors_origins", ["*"]),
+            patch("config._startup_logger") as mock_logger,
+        ):
+            validate_startup(force=True)
+            mock_logger.warning.assert_any_call(
+                "SEC: CORS_ORIGINS contains wildcard '*' — "
+                "restrict to explicit origins before deploying to production."
+            )
+
     def test_validate_startup_skips_in_test_mode(self):
         """validate_startup() should be a no-op in test mode."""
         from config import validate_startup

tests/timmy/test_prompts.py

@@ -77,17 +77,3 @@ def test_lite_prompt_brevity():
     prompt = get_system_prompt(tools_enabled=False).lower()
     assert "brief" in prompt
     assert "plain text" in prompt or "not markdown" in prompt
-
-
-def test_full_prompt_source_distinction():
-    """Full prompt must include source distinction instructions (SOUL.md)."""
-    prompt = get_system_prompt(tools_enabled=True)
-    assert "SOURCE DISTINCTION" in prompt
-    assert "grounded" in prompt.lower()
-    assert "inferred" in prompt.lower()
-
-
-def test_lite_prompt_source_distinction():
-    """Lite prompt must include source distinction instructions."""
-    prompt = get_system_prompt(tools_enabled=False).lower()
-    assert "source distinction" in prompt

tests/timmy/test_timmy_serve_app.py

@@ -8,11 +8,14 @@ from fastapi.testclient import TestClient
 
 @pytest.fixture
 def serve_client():
-    """Create a TestClient for the timmy-serve app."""
-    from timmy_serve.app import create_timmy_serve_app
+    """Create a TestClient for the timmy-serve app with mocked Timmy agent."""
+    with patch("timmy_serve.app.create_timmy") as mock_create:
+        mock_create.return_value = MagicMock()
+        from timmy_serve.app import create_timmy_serve_app
 
-    app = create_timmy_serve_app()
-    return TestClient(app)
+        app = create_timmy_serve_app()
+        with TestClient(app) as client:
+            yield client
 
 
 class TestHealthEndpoint:
@@ -34,18 +37,40 @@ class TestServeStatus:
 
 class TestServeChatEndpoint:
     @patch("timmy_serve.app.create_timmy")
-    def test_chat_returns_response(self, mock_create, serve_client):
+    def test_chat_returns_response(self, mock_create):
         mock_agent = MagicMock()
         mock_result = MagicMock()
         mock_result.content = "I am Timmy."
         mock_agent.run.return_value = mock_result
         mock_create.return_value = mock_agent
 
-        resp = serve_client.post(
-            "/serve/chat",
-            json={"message": "Who are you?"},
-        )
+        from timmy_serve.app import create_timmy_serve_app
+
+        app = create_timmy_serve_app()
+        with TestClient(app) as client:
+            resp = client.post(
+                "/serve/chat",
+                json={"message": "Who are you?"},
+            )
         assert resp.status_code == 200
         data = resp.json()
         assert data["response"] == "I am Timmy."
         mock_agent.run.assert_called_once_with("Who are you?", stream=False)
+
+    @patch("timmy_serve.app.create_timmy")
+    def test_agent_cached_at_startup(self, mock_create):
+        """Verify create_timmy is called once at startup, not per request."""
+        mock_agent = MagicMock()
+        mock_result = MagicMock()
+        mock_result.content = "reply"
+        mock_agent.run.return_value = mock_result
+        mock_create.return_value = mock_agent
+
+        from timmy_serve.app import create_timmy_serve_app
+
+        app = create_timmy_serve_app()
+        with TestClient(app) as client:
+            # Two requests — create_timmy should only be called once (at startup)
+            client.post("/serve/chat", json={"message": "hello"})
+            client.post("/serve/chat", json={"message": "world"})
+        mock_create.assert_called_once()