# FGAO API configuration. Paths are relative to this file.
llm:
  provider: openai                         # any OpenAI-compatible endpoint
  base_url: https://router.huggingface.co/v1
  model: Qwen/Qwen2.5-7B-Instruct          # any chat model served by Hugging Face Inference Providers
  api_key_env: HF_TOKEN                    # Space secret
  temperature: 0
  max_tokens: 128
  timeout: 60
resilience: {attempts: 3, backoff: 0.5, timeout: 60, max_concurrency: 32}
kb: {type: bm25, path: data/kb.jsonl}      # default corpus; requests may send their own "documents"
budget: 16
on_error: record
server: {max_inflight: 32, max_queue: 256, request_timeout: 300, max_batch: 64, max_documents: 500}