update requirements

README.md

@@ -8,3 +8,27 @@ pinned: false
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+## API
+
+- GET `/health` → `{ status, uptime_sec }`
+- GET `/fact-check?claim=...`
+- POST `/fact-check` with JSON `{ "claim": "..." }`
+
+Ingestion from Firebase runs in the background on startup. If you get `INITIALIZING`, retry once `ingestion_finished` becomes true.
+
+## Environment variables (Settings → Variables & secrets)
+
+- Optional LLM keys:
+	- `OPENROUTER_API_KEY` or `OPENAI_API_KEY`
+- Vector DB persistence (recommended with Space Persistent Storage enabled):
+	- `VECTOR_PERSIST_DIR=/data/vector_db`
+- Faster cold start on small CPU instances:
+	- `USE_DUMMY_EMBEDDINGS=true`
+- Firebase overrides (only if you need to change defaults):
+	- `FIREBASE_API_KEY`, `FIREBASE_PROJECT_ID`, `FIREBASE_AUTH_DOMAIN`, `FIREBASE_STORAGE_BUCKET`, `FIREBASE_MESSAGING_SENDER_ID`, `FIREBASE_APP_ID`
+
+## Notes
+
+- The server runs on port 7860 via gunicorn.
+- Without an LLM key, the system uses a heuristic fallback for verdicts.

requirements.txt

@@ -25,4 +25,4 @@ pandas
 scikit-learn
 
 # Production WSGI server
-gunicorn==21.2.0
+gunicorn