nova-snowy-tau.vercel.app

{
  "name": "create_experiment",
  "kind": "act",
  "impl": "imperative",
  "description": "Create a NOVA LAB plant-growth experiment, persist it, and submit it for human approval. Status becomes AWAITING_APPROVAL with approval_required=true. NOVA supports synthetic plant-growth factorial experiments only—not universal science. Supported factors: temperature, water, light, CO2, nutrients (also accepted as TEMPERATURE, WATER, LIGHT, CO2, NUTRIENTS). Provide factor_levels and units for each factor, plus replicates. Does not approve or run the experiment; the server state machine blocks execution until a human approves. Emits experiment_created and approval_requested events. Returns experiment_id, status, estimated_observations, and approval_required. Use when starting a new investigation.",
  "inputSchema": {
    "type": "object",
    "additionalProperties": false,
    "required": [
      "question",
      "hypothesis",
      "factors",
      "factor_levels",
      "units",
      "replicates",
      "seed",
      "simulation_version",
      "provenance"
    ],
    "properties": {
      "question": {
        "type": "string",
        "minLength": 1,
        "description": "Scientific question the experiment asks."
      },
      "hypothesis": {
        "type": "string",
        "minLength": 1,
        "description": "Testable hypothesis for this design."
      },
      "factors": {
        "type": "array",
        "minItems": 1,
        "uniqueItems": true,
        "items": {
          "type": "string",
          "enum": [
            "temperature",
            "water",
            "light",
            "CO2",
            "nutrients",
            "TEMPERATURE",
            "WATER",
            "LIGHT",
            "NUTRIENTS"
          ]
        },
        "description": "Plant-growth factors only: temperature, water, light, CO2, nutrients."
      },
      "factor_levels": {
        "type": "object",
        "description": "Map of factor → numeric levels (keys may be lowercase or TEMPERATURE-style). Every listed factor needs a non-empty array.",
        "additionalProperties": {
          "type": "array",
          "minItems": 1,
          "items": {
            "type": "number"
          }
        }
      },
      "units": {
        "type": "object",
        "description": "Map of factor → unit string (e.g. celsius, relative). Required for every factor.",
        "additionalProperties": {
          "type": "string",
          "minLength": 1
        }
      },
      "replicates": {
        "type": "integer",
        "minimum": 1,
        "description": "Number of replicates per factor combination."
      },
      "seed": {
        "type": "integer",
        "description": "Integer RNG seed for the deterministic synthetic simulation."
      },
      "simulation_version": {
        "type": "string",
        "minLength": 1,
        "description": "Simulation engine version string (e.g. nova-sim-v1)."
      },
      "provenance": {
        "type": "object",
        "additionalProperties": false,
        "required": [
          "source",
          "actor"
        ],
        "properties": {
          "source": {
            "type": "string",
            "minLength": 1
          },
          "actor": {
            "type": "string",
            "minLength": 1
          },
          "notes": {
            "type": "string"
          },
          "parent_experiment_id": {
            "type": "string",
            "description": "Optional parent experiment id for lineage (normally omit; prefer replicate_experiment)."
          }
        }
      }
    }
  },
  "page": "/"
}

{
  "name": "run_experiment",
  "kind": "act",
  "impl": "imperative",
  "description": "Execute an already APPROVED experiment on the NOVA server. Human approval is required beforehand; if status is AWAITING_APPROVAL (or otherwise not APPROVED), the server rejects the run with HUMAN_APPROVAL_REQUIRED—the tool does not decide approval itself. On success: synthetic simulation runs, observations persist, status becomes COMPLETED, and lifecycle events are recorded. Returns experiment_id, status, observation_count, simulation_version, and seed. Does not expose hidden simulator world parameters or equations. Use only after a human has approved the design.",
  "inputSchema": {
    "type": "object",
    "additionalProperties": false,
    "required": [
      "experiment_id"
    ],
    "properties": {
      "experiment_id": {
        "type": "string",
        "minLength": 1,
        "description": "Id of an experiment that is currently APPROVED."
      }
    }
  },
  "page": "/"
}

{
  "name": "get_results",
  "kind": "answer",
  "impl": "imperative",
  "description": "Return persisted synthetic observations for one experiment_id from the NOVA database. Does not regenerate data or re-run the simulator. Each observation includes factor_values, measured_outcome (biomass, growth_rate), replicate_index, and combination_index. Records a results_retrieved audit event. Returns an empty observations list if the experiment has not been run. Use after run_experiment (or whenever observations already exist).",
  "inputSchema": {
    "type": "object",
    "additionalProperties": false,
    "required": [
      "experiment_id"
    ],
    "properties": {
      "experiment_id": {
        "type": "string",
        "minLength": 1,
        "description": "Experiment whose persisted observations should be returned."
      }
    }
  },
  "page": "/"
}

{
  "name": "analyze_results",
  "kind": "act",
  "impl": "imperative",
  "description": "Analyze persisted observations for a COMPLETED experiment using NOVA’s statistical engine (growth ~ temperature + water + temperature:water). Retrieves stored observations, calculates effects (not fabricated or LLM-invented), persists an Analysis row, emits analysis_completed, and sets status to ANALYZED. Returns experiment_id, analysis_id, analysis_version, model specification, factor/interaction effects, confidence intervals, significance statistics (t, p, SE), simulation_version, and seed. Does not expose hidden simulator equations. Use after run_experiment.",
  "inputSchema": {
    "type": "object",
    "additionalProperties": false,
    "required": [
      "experiment_id"
    ],
    "properties": {
      "experiment_id": {
        "type": "string",
        "minLength": 1,
        "description": "COMPLETED experiment with persisted observations."
      }
    }
  },
  "page": "/"
}

{
  "name": "save_finding",
  "kind": "act",
  "impl": "imperative",
  "description": "Persist an interpretive finding written by the external agent, linked to experiment_id and analysis_id. The server validates that the experiment exists, the analysis exists, and the analysis belongs to that experiment—unrelated analysis references are rejected (PROVENANCE_MISMATCH). Does not auto-generate findings; the agent must supply finding_text and confidence. Maintains provenance Finding → Analysis → Experiment → Observations, records finding_saved, and returns finding_id, experiment_id, analysis_id, finding, confidence, and timestamp. Use after analyze_results.",
  "inputSchema": {
    "type": "object",
    "additionalProperties": false,
    "required": [
      "experiment_id",
      "analysis_id",
      "finding_text",
      "confidence"
    ],
    "properties": {
      "experiment_id": {
        "type": "string",
        "minLength": 1,
        "description": "Experiment that owns the analysis evidence."
      },
      "analysis_id": {
        "type": "string",
        "minLength": 1,
        "description": "Persisted analysis that evidences this finding (must belong to experiment_id)."
      },
      "finding_text": {
        "type": "string",
        "minLength": 1,
        "description": "Agent-authored interpretive claim; not generated by NOVA."
      },
      "confidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1,
        "description": "Reporter confidence in [0, 1]."
      },
      "replication_id": {
        "type": "string",
        "description": "Optional persisted replication record id."
      }
    }
  },
  "page": "/"
}

{
  "name": "replicate_experiment",
  "kind": "act",
  "impl": "imperative",
  "description": "Create an independent replication of original_experiment_id using the same experimental design and a NEW server-derived seed. Persists a replication experiment, generates new synthetic observations (not copies), independently calculates replication effects, and evaluates replication_success = same direction AND confidence interval overlap AND relative effect difference < 20%. Persists the Replication record and emits replication_completed. Returns original/replication ids, seeds, effects, confidence intervals, criterion flags, and replication_success. Does not hardcode success or expose hidden simulator internals. Original must already have persisted observations.",
  "inputSchema": {
    "type": "object",
    "additionalProperties": false,
    "required": [
      "original_experiment_id"
    ],
    "properties": {
      "original_experiment_id": {
        "type": "string",
        "minLength": 1,
        "description": "Experiment to replicate (must already have persisted observations)."
      }
    }
  },
  "page": "/"
}