# Imported from ~/.pi/agent/models.json (pi agent settings)
# - apiKey "EMPTY" -> auth: none (keyless local endpoint)
# - thinkingLevelMap {max: xhigh} -> compat.reasoningEffortMap
providers:
  "llamacpp (dankserver)":
    baseUrl: http://192.168.0.235:7770/v1
    api: openai-completions
    auth: none
    models:
      - id: frank-small
        name: "frank-small (dankserver)"
        contextWindow: 262144
        maxTokens: 262144
        reasoning: true
        input: [text]
        compat:
          reasoningEffortMap:
            max: xhigh

  "llamacpp (dankclaw)":
    baseUrl: http://192.168.0.143:7770/v1
    api: openai-completions
    auth: none
    models:
      - id: frank-large
        name: "frank-large (dankclaw-gpu)"
        contextWindow: 200000
        # Halogen caps max_tokens at HALOGEN_MAX_TOKENS_CAP=65536 (a request
        # above it is a 400). Admission reserves prompt + max_tokens against
        # the KV pool (401,408 positions since 2026-09-13): 200,000 + 65,536
        # = 265,536, comfortably inside. Overshooting the pool hangs for the
        # 3600 s queue timeout before a 503.
        maxTokens: 65536
        reasoning: true
        input: [text, image]
        compat:
          reasoningEffortMap:
            max: xhigh
          # Qwen3.8-Flash-Next recommended sampling; halogen has no server-side
          # sampler defaults, so without this the endpoint decodes greedy.
          # extraBody is merged last (after compat policy) and wins.
          extraBody:
            temperature: 1.0
            top_p: 0.95
            top_k: 20
            min_p: 0.0
            presence_penalty: 0.0
            repetition_penalty: 1.0
            # Keep prior thinking blocks in the rendered prompt across turns;
            # halogen honours it top-level (verified: 275-token prompt delta vs false).
            preserve_thinking: true