Troubleshooting

Installation & Startup

ModuleNotFoundError: No module named 'inferia'

Cause: The package is not installed in your current environment. Fix:

pip install -e .

Or ensure you are running commands with the correct python executable or Docker container.

Database connection failed

Cause: PostgreSQL is not running or credentials in .env are incorrect. Fix:

1. Check if Postgres is up: docker ps or pg_isready.
2. Verify POSTGRES_DSN in .env.
3. Ensure the database user/password matches what you set up.

Redis Connection Error

Cause: Redis is not reachable. Fix: Ensure Redis is running on port 6379 (default).

redis-cli ping
# Should return PONG

Runtime Issues

401 Unauthorized on Internal Endpoints

Cause: Missing or mismatched INTERNAL_API_KEY. Fix: Ensure all services (Gateways, Workers) share the exact same INTERNAL_API_KEY in their .env file.

Models Stuck in "Pending"

Cause: No compute nodes are available to take the job. Fix:

1. Check the Dashboard > Compute Pools.
2. Ensure you have active workers (or Nosana nodes) registered.
3. Check orchestrator.log for adapter errors.

"Model not found"

Cause: The deployment name requested does not match any active deployment. Fix:

• Verify the model parameter in your API call matches the Deployment Name (e.g., llama-3-8b), not just the base model name.