Verify the installation

Run these checks in WSL Bash against the environment you started. They inspect different parts of the system.

Check the backend

curl --fail-with-body http://localhost:8000/health

Expected fields: status, model_loaded, and vector_store_size. The implementation counts stored knowledge documents and sets model_loaded to true unconditionally. Treat this as API/store reachability, not proof that a model can answer.

Check Ollama

curl --fail-with-body http://localhost:11434/api/tags
ollama list

Confirm the configured models are listed. If your OLLAMA_HOST points elsewhere, inspect that server instead.

Check authenticated inference

Follow Your first query and verify a nonempty response. Check model rather than assuming the router used the requested model.

Record your environment

For a reproducible issue report, record the source revision, Python and Ollama versions, model names, deployment method, and whether memory is enabled. Do not include your key, environment file, or private documents.

git rev-parse --short HEAD
.venv/bin/python --version
ollama --version

Next: Use the command line. If any check fails, diagnose the symptom.