{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "https://raw.githubusercontent.com/api-evangelist/tale-dev/main/json-schema/tale-dev-run-projection-schema.json",
  "title": "RunProjection",
  "description": "A run read with `?fields=`: exactly the keys of `Run` the query named, no others — none is required, since the caller chose them.",
  "x-generated": "2026-10-09",
  "x-method": "derived",
  "x-generator": "derive-json-schema.py",
  "x-source": "openapi/tale-dev-openapi.yml#/components/schemas/RunProjection",
  "type": "object",
  "properties": {
    "id": {
      "type": "string",
      "description": "The run id (`runId` at start)"
    },
    "organizationId": {
      "type": "string"
    },
    "name": {
      "type": "string"
    },
    "version": {
      "type": "integer"
    },
    "projectId": {
      "type": [
        "string",
        "null"
      ],
      "description": "URL project for a project run; null for an organization run"
    },
    "status": {
      "type": "string",
      "enum": [
        "queued",
        "running",
        "waiting",
        "quarantined",
        "success",
        "failed",
        "cancelled"
      ]
    },
    "mode": {
      "type": "string",
      "enum": [
        "mock",
        "live"
      ]
    },
    "startedBy": {
      "type": "string",
      "description": "What started the run — one of four forms: `api-key:<userId>` (a start through this API or the MCP endpoint), `user:<userId>` (a start from the product), `trigger:<triggerId>` (a schedule, webhook or event — the id the trigger read answers), or a bare user id on runs the in-product builder recorded before it prefixed them. Split on the first `:`; treat a value without one as a user id."
    },
    "startedVia": {
      "type": "string",
      "enum": [
        "schedule",
        "webhook",
        "event"
      ],
      "description": "Which kind of trigger started a `trigger:<id>` run — read off the run’s own input (`input.trigger`), so it stays true after the binding changes kind. Absent on a run a person or an API key started."
    },
    "input": {},
    "output": {},
    "checkpoints": {},
    "trace": {},
    "effects": {},
    "detail": {
      "type": [
        "string",
        "null"
      ],
      "description": "The failure or wait reason; null while the run has none — and null again once a cancel lands (the park it named is over). While `waiting` it names the park: `approval:<approvalId>`, `agent:<nodeId>`, `room:<nodeId>`, `repeat:<nodeId>` or `in_doubt:<nodeId>` — `waitingFor` is the field to branch on; when `failed`, the failure sentence, and `failureCode` the stable cause to branch on — the sentence is not contractual."
    },
    "failureCode": {
      "type": [
        "string",
        "null"
      ],
      "enum": [
        "node_error",
        "connector_error",
        "llm_output_invalid",
        "approval_rejected",
        "execution_limit",
        "automation_deleted",
        "engine_incompatible",
        "effect_in_doubt",
        "missing_api_key",
        "credit_exhausted",
        "model_not_entitled",
        "auth_error",
        "model_not_found",
        "unsupported_parameter",
        "output_cap_too_high",
        "token_limit",
        "rate_limited",
        "content_filter",
        "context_length",
        "provider_unreachable",
        "provider_error",
        "budget_exceeded",
        "harness_error",
        "turn_crashed",
        "session_gone",
        "start_failed",
        "harvest_failed",
        "resume_failed",
        "deadline",
        "ask_expired",
        null
      ],
      "description": "Why a `failed` run failed — present only with `status: \"failed\"`, null otherwise (and null on a failure recorded before this field existed). `node_error` — the node’s own code, template, `forEach` or `output` expression failed, an authoring refusal: the author’s problem; `connector_error` — a connector action refused or failed; `llm_output_invalid` — the model’s reply did not satisfy the node’s `outputSchema`; `approval_rejected`; `execution_limit` — the 100-execution guard; `automation_deleted` — the automation vanished mid-flight; `engine_incompatible` — the run’s saved progress could not be read by this version of Tale, so it was stopped instead of starting over (no step ran twice); `effect_in_doubt` — a person failed the run at a write that may already have reached its service when the run was interrupted. The provider codes the chat surface documents (`credit_exhausted`, `auth_error`, `rate_limited`, `provider_unreachable`, `provider_error`, `model_not_found`, …) — an `llm` node’s provider: the account or the provider, not the request. The agent codes (`harness_error`, `turn_crashed`, `session_gone`, `deadline`, `ask_expired`, `budget_exceeded`, …) — an `agent` node’s turn, after its in-node retries. `budget_exceeded` — a budget limit refused an `agent` node’s turn or an `llm` node’s call, or the turn used up its allowance. Retry on `provider_error`, `provider_unreachable`, `rate_limited`, `turn_crashed`, `session_gone`, `harvest_failed`; alert a person on the rest."
    },
    "waitingFor": {
      "type": "string",
      "enum": [
        "approval",
        "ask",
        "in_doubt",
        "agent",
        "room",
        "repeat"
      ],
      "description": "Present only while `status` is `waiting`: what the run is parked on. `approval` — a person’s decision on a gate; `ask` — a question a person has to answer; `in_doubt` — a write the run was making when its server stopped may or may not have reached its service, and a person must decide how to continue (in the app: run it again, skip it, or fail the run); `agent` — an agent turn still running, no one to page; `room` — an agent turn whose start waits for sandbox room, no one to page; `repeat` — a node polling until its `repeatUntil` condition holds, no one to page."
    },
    "legacyQuarantine": {
      "$ref": "#/$defs/LegacyRunQuarantine"
    },
    "resumeCount": {
      "type": "integer",
      "minimum": 0,
      "description": "How often the run moved to another server: a server that was being updated or restarted handed it on, or another one took it over after its own stopped responding. Steps it had finished never run again. `Run` always carries it; `RunSummary` omits it while it is 0."
    },
    "lastResume": {
      "type": "object",
      "required": [
        "reason",
        "at"
      ],
      "additionalProperties": false,
      "description": "Why and when the run last moved to another server; absent while it never did. `shutdown` — its server was being updated or restarted and handed it on; `lease_expired` — its server stopped responding and another one took it over.",
      "properties": {
        "reason": {
          "type": "string",
          "enum": [
            "shutdown",
            "lease_expired"
          ]
        },
        "at": {
          "type": "integer",
          "minimum": 0,
          "description": "Epoch milliseconds"
        }
      }
    },
    "stalled": {
      "type": "boolean",
      "description": "True while a `running` run waits for a server to take it over after its own stopped (the app reads it as “Interrupted — resuming”): nothing is working on it right now, and another server picks it up within about a minute and a half. `Run` always carries it; `RunSummary` carries it only while true."
    },
    "claimEpoch": {
      "type": "integer",
      "description": "The stepper’s claim fence: incremented on every claim — each turn, a takeover after a server stopped, or a queue retry; a worker holding an older epoch has its writes refused as stale. Diagnostic — above 1 means the run was claimed more than once."
    },
    "chainSeq": {
      "type": "integer",
      "description": "The poll-chain fence of a waiting run: incremented each time the run parks, so a wake-up scheduled for an earlier park is ignored. Diagnostic — how many times the run has parked."
    },
    "startedAt": {
      "type": "integer",
      "minimum": 0,
      "description": "When the start was accepted — the 202 of a start, a webhook delivery or a trigger occurrence; the run is `queued` from this moment. There is no separate pick-up stamp: `queued` → `running` is the stepper’s first claim, usually within a second, so `finishedAt - startedAt` is wall-clock including any wait for a worker."
    },
    "finishedAt": {
      "type": [
        "integer",
        "null"
      ],
      "minimum": 0,
      "description": "Epoch milliseconds"
    }
  },
  "additionalProperties": false,
  "$defs": {
    "LegacyRunQuarantine": {
      "type": "object",
      "additionalProperties": false,
      "description": "A legacy execution whose external effects or termination cannot be proven. Present on held runs. A recorded stop request does not resolve the hold or establish that work stopped.",
      "required": [
        "reason",
        "observedAt",
        "claimEpoch",
        "priorStatus",
        "resolution"
      ],
      "properties": {
        "reason": {
          "type": "string",
          "enum": [
            "legacy_execution_unproven"
          ]
        },
        "observedAt": {
          "type": "integer",
          "minimum": 0,
          "maximum": 9007199254740991
        },
        "claimEpoch": {
          "type": "integer",
          "minimum": 0,
          "maximum": 9007199254740991
        },
        "priorStatus": {
          "type": "string",
          "enum": [
            "queued",
            "running",
            "waiting"
          ]
        },
        "resolution": {
          "type": [
            "object",
            "null"
          ],
          "additionalProperties": false,
          "required": [
            "action",
            "actor",
            "at"
          ],
          "properties": {
            "action": {
              "type": "string",
              "enum": [
                "stop"
              ]
            },
            "actor": {
              "type": "string",
              "minLength": 1,
              "maxLength": 200
            },
            "at": {
              "type": "integer",
              "minimum": 0,
              "maximum": 9007199254740991
            }
          }
        }
      }
    }
  }
}