v15

latestOpenAPI 3.0.0raw.githubusercontent.com2026-04-287323136.0 KB
tasks

Submit a task for execution

Full 8-stage pipeline: guardrail → credit → compute → verify → anchor → receipt

post/api/v1/tasks/submit

Request body

promptstring required

The prompt/instruction for the LLM

taskType'general' | 'code' | 'analysis' | 'creative' | 'trade_signal' | 'evm-action' required

Type of task. Use evm-action for deterministic EVM state verification.

agentIdstring required

Agent fingerprint ID (unique per agent)

metadataobject

Optional metadata for the task

modelstring

Model slug from the whitelist (e.g. "llama-3.3-70b"). Defaults to llama-3.3-70b if omitted. Ignored for evm-action type.

Example request

{
  "prompt": "Analyze the sentiment of the following text: \"Bitcoin is looking bullish today\"",
  "taskType": "analysis",
  "agentId": "agent-alpha-001",
  "model": "llama-3.3-70b"
}

Response

Task executed — provenance receipt returned

taskIdstring
status'COMPLETED' | 'UNMETERED' | 'GUARDRAIL_BLOCKED' | 'INSUFFICIENT_CREDITS' | 'COMPUTE_FAILED' | 'OUTPUT_BLOCKED'
agentIdstring
modelstring
providerstring
latencyMsnumber
outputstring
ipfsCidstring
stateRootstring
verificationStatusstring
verificationScorenumber
verificationProofstring nullable

TX hash from on-chain verification anchor

verificationBasescanUrlstring nullable
attestationUidstring nullable

EAS Omni-Stamp UID (bytes32 hex)

easscanUrlstring nullable

EASScan deep link