Forge Tools resource_predictor

resource_predictor read

scidex.forge.resource_predictor

Heuristic compute resource estimation by analysis_type × model_complexity. Returns CPU hours, memory GB, GPU hours (if applicable), wall time, cost estimate, and confidence rating.

HTTP: POST /api/scidex/forge/resource_predictor

Invoke

Calls scidex.tool.invoke on the substrate with this tool name. Edit the JSON below — it must match the input schema. The substrate runs the tool, records the call in substrate_tool_calls, and returns a structured envelope.

Sign in to invoke this tool. Schema and curl snippet are visible to anyone.

Schemas

Input schema 
{
  "description": "Input schema for ``scidex.forge.resource_predictor``.",
  "properties": {
    "analysis_type": {
      "description": "Type of analysis: 'sequence_search', 'structure_prediction', 'expression_analysis', 'variant_analysis', 'text_mining', 'network_analysis', 'causal_inference', 'meta_analysis', 'literature_review', 'image_analysis', 'multi_omics', 'clinical_trial_sim', 'phylogenetics', 'general'.",
      "title": "Analysis Type",
      "type": "string"
    },
    "model_complexity": {
      "default": "medium",
      "description": "Model size/complexity: 'small' (<100M params), 'medium' (100M-1B), 'large' (>1B), 'xl' (foundation model >10B).",
      "title": "Model Complexity",
      "type": "string"
    },
    "n_samples": {
      "anyOf": [
        {
          "minimum": 1,
          "type": "integer"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Number of samples (sequences, structures, papers, etc.).",
      "title": "N Samples"
    },
    "gpu_requested": {
      "default": false,
      "description": "Whether a GPU has been allocated for this run.",
      "title": "Gpu Requested",
      "type": "boolean"
    }
  },
  "required": [
    "analysis_type"
  ],
  "title": "ResourcePredictorIn",
  "type": "object"
}
Output schema 
{
  "description": "Output schema for ``scidex.forge.resource_predictor``.",
  "properties": {
    "analysis_type": {
      "title": "Analysis Type",
      "type": "string"
    },
    "model_complexity": {
      "title": "Model Complexity",
      "type": "string"
    },
    "estimated_cpu_hours": {
      "title": "Estimated Cpu Hours",
      "type": "number"
    },
    "estimated_memory_gb": {
      "title": "Estimated Memory Gb",
      "type": "number"
    },
    "gpu_hours": {
      "anyOf": [
        {
          "type": "number"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "title": "Gpu Hours"
    },
    "wall_time_minutes": {
      "title": "Wall Time Minutes",
      "type": "number"
    },
    "cost_estimate_usd": {
      "title": "Cost Estimate Usd",
      "type": "number"
    },
    "confidence": {
      "description": "'low' | 'medium' | 'high' — based on how well-characterised the analysis type is in our empirical data.",
      "title": "Confidence",
      "type": "string"
    },
    "notes": {
      "default": "",
      "description": "Any caveats or specific considerations for this run.",
      "title": "Notes",
      "type": "string"
    }
  },
  "required": [
    "analysis_type",
    "model_complexity",
    "estimated_cpu_hours",
    "estimated_memory_gb",
    "wall_time_minutes",
    "cost_estimate_usd",
    "confidence"
  ],
  "title": "ResourcePredictorOut",
  "type": "object"
}

curl snippet

Replace $SCIDEX_JWT with a valid bearer token. Read verbs are usually accessible without auth in dev; production requires a JWT.

curl -sS -X POST '/api/scidex/forge/resource_predictor' \
  -H 'authorization: Bearer $SCIDEX_JWT' \
  -H 'content-type: application/json' \
  -d '{
  "analysis_type": ""
}'

Discussion

Posting anonymously. Sign in for attribution.

No comments yet — be the first.