adaptorch_run
Launches orchestrated task-DAG runs by submitting payloads to the AdaptOrch control plane and optionally waits for terminal results.
Instructions
Submit a task payload to AdaptOrch control-plane and optionally wait.
Tool call template
This tool declares no required parameters. Add optional parameters from the input schema as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {},
"name": "adaptorch_run"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| model | No | Model ID, or 'auto' on control-plane backends to resolve this tenant's provider default or confirmed run history. Provider credentials are still required. | |
| prompt | No | ||
| context | No | ||
| payload | No | Task DAG. subtasks is required and non-empty; dependencies may be omitted. | |
| budget_policy | No | ||
| connector_name | No | mcp | |
| synthesis_mode | No | Synthesis strategy executed by the installed engine. Supported modes: paper, robust, robust_lite, stable_hybrid. Deprecated aliases resolve to a supported mode: fourier_aggressive -> stable_hybrid. The serving-layer value 'auto' asks the control plane to pick a supported engine mode per run (control-plane backends only; the choice is reported via synthesis_mode_used). Hosted callers that omit the field receive robust. | robust |
| timeout_seconds | No | ||
| ensemble_members | No | BYOK request-scoped ensemble fan-out: 2-5 'provider:model' members riding the process provider credential (ADAPTORCH_MCP_PROVIDER, ADAPTORCH_MCP_PROVIDER_MODEL, ADAPTORCH_MCP_PROVIDER_API_KEY). Quota: one run debits one unit regardless of member count; model token cost bills to the caller's own provider account. | |
| output_extractor | No | Ensemble output extraction mode hint. Engine extractors: final_answer, multiple_choice_letter. Use 'none' to request no extractor; it is not forwarded to the engine. | |
| wait_for_terminal | No | ||
| poll_interval_seconds | No | ||
| prefer_ensemble_singleton | No | Hint the router to prefer multi_model_ensemble for single-node DAGs. true = explicit on, false = explicit off (the ensemble auto preference will not re-enable it), null/omitted = auto policy. |