Trace
/api/v1/telemetry/tracesA trace/span record.
listgetSchema
| Field | Type | Notes |
|---|---|---|
| id* | uuid | Server-assigned ULID with type prefix (e.g. per_…). |
| tenant_id* | string | Tenant scope — auto-stamped from the caller's JWT. |
| app_id | string | App scope. Stamped ONLY when the caller's JWT was minted for a specific Application (integration API keys). Absent for human-user sessions. Filters reads when present. |
| created_at* | date-time | Server stamp. |
| updated_at* | date-time | Server stamp; updated on every patch. |
| created_by | uuid | Person id from the caller's JWT (sub). |
| updated_by | uuid | Person id from the last writer's JWT (sub). |
| source | string | Provenance tag — defaults to 'edm'. |
| source_type | enum | frontend | backend | server | system | apiWhere the write originated. Defaults to 'api'. |
| is_deleted | boolean | Soft-delete flag. Excluded from default list queries. |
| deleted_at | date-time | Stamped when soft-deleted; null otherwise. |
| deleted_by | uuid | Person id who soft-deleted; null otherwise. |
| schema_version | number | Document schema version. Bumped on incompatible writes. |
| parent_span_id | string | Distributed-tracing parent span ID, if this trace is a child span. |
| agent_run_id | string | Identifier of the agent run this trace belongs to, if part of a multi-step agent workflow. |
| user_id | string | Identifier of the user who initiated the LLM call. Null for system-initiated calls. |
| api_key_id | string | Identifier of the virtual API key used for this call. |
| model | string | Model identifier in 'provider/model-name' format. |
| provider | string | LLM provider name. |
| span_kind | string | Kind of LLM operation represented by this trace. |
| input_tokens | number | Number of prompt/input tokens consumed. |
| output_tokens | number | Number of completion/output tokens generated. |
| cache_read_tokens | number | Number of tokens served from the prompt cache. |
| cache_write_tokens | number | Number of tokens written into the prompt cache. |
| latency_ms | number | End-to-end latency of the LLM call in milliseconds. |
| cache_hit | boolean | Whether this response was served (fully or partially) from the prompt cache. |
| cost_usd | number | Estimated cost of this LLM call in USD. |
| finish_reason | string | Reason the model stopped generating. One of 'stop', 'length', 'tool_calls', or 'error'. |
| routing_rule | string | Identifier of the routing rule that selected the provider for this call. |
| policy_decisions | json | Policy decisions evaluated for this call by the coreiq policy engine. |
| security_violations | json | PII or DLP violations detected by the policy engine for this call. |
| request_hash | string | SHA-256 hash of the canonicalized request payload, used as the cache key. |
| episode_id | string | Linked evolution episode identifier, if this trace is part of a tracked episode. |
| ts | date-time | ISO 8601 UTC timestamp of when the LLM call occurred. |
API
Loading manifest…