{
  "_comment": "blackhole unified configuration — ~/.pi/agent/pi-blackhole-config.json",
  "_all_settings_in_one_file": true,

  "overrideDefaultCompaction": false,
  "noAutoCompact": false,
  "debug": false,

  "model": {
    "_comment": "Base model (last fallback before session model).",
    "provider": "openrouter",
    "id": "google/gemma-4-31b-it:free",
    "thinking": "low",
    "cooldownHours": 6
  },

  "observerModel": {
    "_comment": "Observer primary: Qwen3 Next 80B via OpenRouter free (strong extraction with reasoning).",
    "provider": "openrouter",
    "id": "qwen/qwen3-next-80b-a3b-instruct:free",
    "thinking": "low",
    "cooldownHours": 12
  },
  "observerFallbackModels": [
    {
      "_comment": "Observer fallback 1: Gemma 4 31B via Ollama Cloud.",
      "provider": "ollama",
      "id": "gemma4:31b-cloud",
      "thinking": "off",
      "cooldownHours": 6
    },
    {
      "_comment": "Observer fallback 2: Gemma 4 31B via OpenRouter free tier.",
      "provider": "openrouter",
      "id": "google/gemma-4-31b-it:free",
      "thinking": "off",
      "cooldownHours": 6
    }
  ],

  "reflectorModel": {
    "_comment": "Reflector primary: GPT OSS 120B via Cerebras (strong reasoning for synthesis).",
    "provider": "cerebras",
    "id": "gpt-oss-120b",
    "thinking": "low",
    "cooldownHours": 12
  },
  "reflectorFallbackModels": [
    {
      "_comment": "Reflector fallback 1: GLM-4.7 via z.ai with thinking enabled.",
      "provider": "zai",
      "id": "glm-4.7",
      "thinking": "low",
      "cooldownHours": 6
    },
    {
      "_comment": "Reflector fallback 2: GPT OSS 120B via OpenRouter free tier.",
      "provider": "openrouter",
      "id": "openai/gpt-oss-120b:free",
      "thinking": "low",
      "cooldownHours": 6
    }
  ],

  "dropperModel": {
    "_comment": "Dropper primary: GPT OSS 120B via Cerebras (judgment/prioritization).",
    "provider": "cerebras",
    "id": "gpt-oss-120b",
    "thinking": "off",
    "cooldownHours": 12
  },
  "dropperFallbackModels": [
    {
      "_comment": "Dropper fallback 1: GLM-4.7 via z.ai.",
      "provider": "zai",
      "id": "glm-4.7",
      "thinking": "low",
      "cooldownHours": 6
    },
    {
      "_comment": "Dropper fallback 2: GPT OSS 120B via OpenRouter free tier.",
      "provider": "openrouter",
      "id": "openai/gpt-oss-120b:free",
      "thinking": "low",
      "cooldownHours": 6
    }
  ],

  "observeAfterTokens": 10000,
  "reflectAfterTokens": 20000,
  "compactAfterTokens": 81000,
  "observationsPoolMaxTokens": 20000,
  "reflectorInputMaxTokens": 80000,
  "dropperInputMaxTokens": 80000,
  "observerChunkMaxTokens": 40000,
  "agentMaxTurns": 16,

  "passive": false,
  "noAutoCompact": false,
  "debugLog": false,

  "_notes": [
    "Model fallback chains per worker:",
    "  Observer: qwen3-next-80b → gemma4:31b-cloud → google/gemma-4-31b-it:free → base model → session",
    "  Reflector: gpt-oss-120b (cerebras) → glm-4.7 (z.ai) → openai/gpt-oss-120b:free → base model → session",
    "  Dropper:  gpt-oss-120b (cerebras) → glm-4.7 (z.ai) → openai/gpt-oss-120b:free → base model → session",
    "",
    "On retryable error (429, 5xx, timeout), the model is cooled down for cooldownHours.",
    "Cooled-down models are skipped on subsequent consolidation attempts until cooldown expires.",
    "After all candidates exhausted, the stage aborts. 30s retry gate prevents spam.",
    "Cooldowns persist in ~/.pi/agent/pi-blackhole/pi-blackhole-cooldown.json (survives pi restarts).",
    "Valid thinking values: off, minimal, low, medium, high, xhigh.",
    "Env override for passive mode: PI_BLACKHOLE_PASSIVE=true."
  ]
}