$ docker-agent eval fixtures/agents/files.yaml fixtures/evals --judge-model dmr/ai/qwen3:4b -c 1 --models-gateway http://model-runner.docker.internal/engines --output work/eval-results-gateway
Evaluation run: <run>
Loading evaluation sessions...
Validating judge model...


Total Cost: $0.000000
Total Time: <dur>

Sessions DB: $CAPTURE/work/eval-results-gateway/<run>.db
Sessions JSON: $CAPTURE/work/eval-results-gateway/<run>.json
Log: $CAPTURE/work/eval-results-gateway/<run>.log
Error: running evaluations: judge model validation failed: streaming judge response: HTTP 403: POST "http://model-runner.docker.internal/engines/v1/chat/completions": 403 Forbidden 
[exit 1]