Finalize pr2 fixes review

.github/dependabot.yml

@@ -2,7 +2,7 @@ version: 2
 updates:
   # Docker image dependency updates
   - package-ecosystem: "docker"
-    directory: "/"
+    directory: "/router"
     schedule:
       interval: "daily"
       time: "18:22"
@@ -11,7 +11,6 @@ updates:
       separator: "/"
     commit-message:
       prefix: "chore(deps)"
-      prefix-major: "chore(deps)!"
       include: "scope"
     labels:
       - "dependencies"
@@ -20,11 +19,6 @@ updates:
       - "sheepdestroyer"
     allow:
       - dependency-type: "direct"
-    ignore:
-      # Ignore major version updates for critical services
-      # Uncomment and adjust as needed
-      # - dependency-name: "postgres"
-      #   update-types: ["version-update:semver-major"]
 
   # GitHub Actions dependency updates (optional - add if using GH Actions)
   - package-ecosystem: "github-actions"

.github/workflows/test.yml

@@ -0,0 +1,28 @@
+name: Run Tests
+
+on:
+  push:
+    branches: [ master ]
+  pull_request:
+    branches: [ master ]
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout Repository
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Run Circuit Breaker Tests
+        run: python3 test_circuit_breaker.py
+
+      - name: Run Breaker Verification
+        run: python3 verify_breaker.py
+
+      - name: Run Integration Verification
+        run: python3 test_a2_verify.py

.jules/bolt.md

@@ -0,0 +1,3 @@
+## 2024-06-16 - Synchronous I/O in Async API Handlers
+**Learning:** `save_persisted_stats()` was being called synchronously on every API request, cache hit, and tool usage log, triggering blocking disk I/O in the main event loop.
+**Action:** Always throttle or batch background telemetry writes in async Python applications to prevent blocking the event loop under load.

pod.yaml

@@ -127,6 +127,8 @@ spec:
     volumeMounts:
     - mountPath: /config/router_dir
       name: router-config
+    - mountPath: /app/data
+      name: dataset-data
     - mountPath: /config/gemini_auth
       name: gemini-secrets
     - mountPath: /root/.gemini
@@ -466,3 +468,7 @@ spec:
   - hostPath:
       path: /home/gpav/Vrac/LAB/AI/LLM-Routing
     name: env-file
+  - hostPath:
+      path: /home/gpav/Vrac/LAB/AI/LLM-Routing/data
+      type: DirectoryOrCreate
+    name: dataset-data

router/Containerfile

@@ -6,7 +6,8 @@ WORKDIR /app
 RUN pip install --no-cache-dir fastapi uvicorn httpx pyyaml python-multipart asyncpg langfuse
 
 # Copy all source in one layer — removes dead config COPY (volume-mounted at runtime)
-COPY main.py agy_proxy.py circuit_breaker.py aa_scores.json free_models_roster.json best_free_model.json /app/
+COPY main.py agy_proxy.py circuit_breaker.py aa_scores.json free_models_roster.json /app/
+COPY static/ /app/static/
 
 EXPOSE 5000
 

router/agy_proxy.py

@@ -253,7 +253,7 @@ async def try_agy_proxy(prompt: str, messages: list = None,
         # Determine which breaker to use for this tier
         # Tier 0 (idx 0): gemini-3.5-flash → google_breaker
         # Tier 1 (idx 1): claude-opus-4.6  → vendor_breaker
-        is_google_tier = (actual_tier_idx == 0)
+        is_google_tier = "gemini" in tier.get("model_name", "").lower()
         tier_breaker = google_breaker if is_google_tier else vendor_breaker
 
         if not tier_breaker.is_allowed():
@@ -323,8 +323,8 @@ async def try_agy_proxy(prompt: str, messages: list = None,
 
             # Success! Stream has started.
             tier_breaker.record_success()
-            # Define the async generator
             async def token_generator(stream_resp, httpx_client, initial_line, current_conv_id):
+                """Asynchronously yields tokens from the agy daemon stream and manages session state updates."""
                 # Yield the initial token if it was a token
                 init_data = json.loads(initial_line)
                 if init_data.get("type") == "token" and init_data.get("content"):

router/circuit_breaker.py

@@ -34,6 +34,7 @@ class PerModelBreaker:
     """Tracks quota exhaustion for a specific model family with exponential cooldown."""
 
     def __init__(self, name: str):
+        """Initialize the per-model circuit breaker with a name and default tier/cooldown states."""
         self.name = name
         self.tier: int = 0        # 0 = open (allowed), 1-3 = cooldown active
         self.cooldown_until: float = 0.0
@@ -84,8 +85,6 @@ def record_failure(self):
 
         if self.tier == 0:
             new_tier = 1
-        elif self.probe_granted:
-            new_tier = min(self.tier + 1, MAX_TIER)
         else:
             new_tier = min(self.tier + 1, MAX_TIER)
 
@@ -130,13 +129,15 @@ class DualCircuitBreaker:
     """
 
     def __init__(self):
+        """Initialize the dual circuit breaker with separate google and vendor sub-breakers."""
         self.google = PerModelBreaker("google")
         self.vendor = PerModelBreaker("vendor")
 
     # Backward-compat: old code calls get_breaker().is_allowed()
     # Default to checking BOTH — if either allows, return allowed.
     # This ensures old code without model awareness works correctly.
     def is_allowed(self) -> bool:
+        """Check if either the google or vendor breaker allows the request (backward-compat)."""
         return self.google.is_allowed() or self.vendor.is_allowed()
 
     def record_failure(self):
@@ -151,9 +152,11 @@ def record_success(self):
 
     @property
     def tier(self) -> int:
+        """Return the maximum cooldown tier across both google and vendor sub-breakers."""
         return max(self.google.tier, self.vendor.tier)
 
     def status(self) -> dict:
+        """Return the aggregated status dictionary of both sub-breakers for the dashboard."""
         return {
             "google": self.google.status(),
             "vendor": self.vendor.status(),