Skip to content

LOCAL_BACKEND_SETUP.md

Latest commit

 

History

History
28 lines (20 loc) · 802 Bytes

LOCAL_BACKEND_SETUP.md

File metadata and controls

28 lines (20 loc) · 802 Bytes
Raw

Local Backend Setup

This package expects a local model backend that exposes an OpenAI-compatible chat completions endpoint.

Supported first-class path

  • llama.cpp server or another OpenAI-compatible local endpoint
  • llamafile if configured to expose a compatible endpoint

Example endpoint

  • http://127.0.0.1:8080/v1/chat/completions

Health check

Run:

python scripts/execution/check_local_backend.py --backend llamacpp_server

Benchmark run

Run:

python scripts/execution/run_full_candidate_gate.py   --adapter openai_compatible   --prompt-profile minimal_doctrine   --model-name local_candidate_v1

Notes

  • Keep temperature low for evaluation runs.
  • Point the adapter at a local endpoint, not a cloud service, if the goal is local cognition evaluation.