althing.dev

{
  "name": "run_prompt",
  "kind": "act",
  "impl": "imperative",
  "description": "Send a single prompt to an LLM and get a response. No personas required. The simplest tool — ask a quick research question without constructing personas or running a full panel.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "prompt": {
        "type": "string",
        "description": "The question or prompt to send."
      },
      "model": {
        "type": "string",
        "description": "LLM model to use (e.g. 'sonnet', 'haiku', 'gpt-4o', 'gemini-2.5-flash'). Defaults to haiku."
      },
      "temperature": {
        "type": "number",
        "description": "Sampling temperature (0.0-1.0). Controls randomness."
      },
      "top_p": {
        "type": "number",
        "description": "Nucleus sampling threshold (0.0-1.0). Alternative to temperature."
      },
      "use_sampling": {
        "type": "boolean",
        "description": "Explicit mode override. true forces host sampling, false forces BYOK."
      }
    },
    "required": [
      "prompt"
    ]
  },
  "page": "/"
}

{
  "name": "run_panel",
  "kind": "act",
  "impl": "imperative",
  "description": "Run a full synthetic focus group panel. Each persona answers all questions independently in parallel; a synthesis step then aggregates findings into themes, agreements, disagreements, and recommendations. Supports v1/v2 flat instruments and v3 branching instruments.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "questions": {
        "type": "array",
        "description": "Flat list of question dicts (v1-equivalent). Each should have a 'text' key.",
        "items": {
          "type": "object"
        }
      },
      "personas": {
        "type": "array",
        "description": "Inline persona definitions. Each persona object accepts 'name' (required), plus optional 'age', 'occupation', 'background', 'personality_traits'.",
        "items": {
          "type": "object"
        }
      },
      "pack_id": {
        "type": "string",
        "description": "ID of a saved persona pack. Merged with inline personas (inline first)."
      },
      "instrument": {
        "type": "object",
        "description": "Raw instrument body (the value under the top-level 'instrument:' key in YAML). Takes precedence over 'questions'."
      },
      "instrument_pack": {
        "type": "string",
        "description": "Name of an installed instrument pack. Takes precedence over 'instrument' and 'questions'."
      },
      "model": {
        "type": "string",
        "description": "LLM model to use. Defaults to haiku."
      },
      "response_schema": {
        "type": "object",
        "description": "Optional JSON Schema for structured output per panelist response."
      },
      "synthesis": {
        "type": "boolean",
        "description": "Whether to run synthesis after collecting responses. Defaults to true."
      },
      "synthesis_model": {
        "type": "string",
        "description": "Model to use for synthesis. Defaults to panelist model."
      },
      "synthesis_prompt": {
        "type": "string",
        "description": "Custom synthesis prompt. Replaces the default."
      },
      "temperature": {
        "type": "number"
      },
      "top_p": {
        "type": "number"
      },
      "persona_models": {
        "type": "object",
        "description": "Per-persona model overrides: maps persona name to model alias."
      },
      "extract_schema": {
        "description": "Schema for post-hoc structured extraction. Built-in name ('sentiment', 'themes', 'rating') or inline JSON Schema."
      },
      "models": {
        "type": "array",
        "description": "List of model names for multi-model ensemble (length >= 2 enables ensemble mode).",
        "items": {
          "type": "string"
        }
      },
      "synthesis_temperature": {
        "type": "number"
      },
      "variants": {
        "type": "integer",
        "description": "Number of variant runs for sampling-based aggregation."
      },
      "use_sampling": {
        "type": "boolean"
      },
      "decision_being_informed": {
        "type": "string",
        "description": "v1.0.0 contract field — the decision this panel will inform (12-280 chars, single line)."
      }
    }
  },
  "page": "/"
}

{
  "name": "run_quick_poll",
  "kind": "act",
  "impl": "imperative",
  "description": "Quick single-question poll across personas. A simplified version of run_panel for quick feedback on one question. Includes synthesis by default. Defaults to a built-in diverse persona pack when none is provided.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "question": {
        "type": "string",
        "description": "The question to ask all personas."
      },
      "personas": {
        "type": "array",
        "description": "List of persona definitions. Optional — when omitted, a built-in pack of diverse personas is used.",
        "items": {
          "type": "object"
        }
      },
      "model": {
        "type": "string",
        "description": "LLM model to use. Defaults to haiku."
      },
      "response_schema": {
        "type": "object"
      },
      "synthesis": {
        "type": "boolean"
      },
      "synthesis_model": {
        "type": "string"
      },
      "synthesis_prompt": {
        "type": "string"
      },
      "temperature": {
        "type": "number"
      },
      "top_p": {
        "type": "number"
      },
      "use_sampling": {
        "type": "boolean"
      },
      "decision_being_informed": {
        "type": "string",
        "description": "v1.0.0 contract field — the decision this poll will inform (12-280 chars, single line)."
      }
    },
    "required": [
      "question"
    ]
  },
  "page": "/"
}

{
  "name": "extend_panel",
  "kind": "act",
  "impl": "imperative",
  "description": "Append one ad-hoc round to a saved panel result. Reuses each panelist's saved session so the follow-up sees full conversational context. NOT a re-entry into the v3 DAG — for adaptive branching, run a fresh run_panel against a v3 instrument instead.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "result_id": {
        "type": "string",
        "description": "ID of a previously saved panel result."
      },
      "questions": {
        "type": "array",
        "description": "One or more questions for the ad-hoc round. They run as a single round, in order, against the same personas as the original run.",
        "items": {
          "type": "object"
        }
      },
      "model": {
        "type": "string",
        "description": "LLM model to use for the new round. Defaults to haiku."
      },
      "synthesis": {
        "type": "boolean"
      },
      "synthesis_model": {
        "type": "string"
      },
      "synthesis_prompt": {
        "type": "string"
      },
      "decision_being_informed": {
        "type": "string",
        "description": "v1.0.0 contract field — the decision this extension informs (12-280 chars, single line)."
      }
    },
    "required": [
      "result_id",
      "questions"
    ]
  },
  "page": "/"
}