> ## Documentation Index > Fetch the complete documentation index at: https://docs.trajectory.ai/llms.txt > Use this file to discover all available pages before exploring further. # Get training step summaries > Retrieve trajectory, group, and termination summaries for each recorded training step. ## OpenAPI ````yaml /openapi.json get /api/v1/train/{run_id}/step-summaries openapi: 3.1.0 info: title: Trajectory API version: 0.1.0 x-trajectory-api-version: '2026-10-14' description: >- This contract is API version 2026-10-14. The optional Trajectory-API-Version header pins it, and tmp-contract is accepted as an alias. Requests without the header get this version. Version 2026-10-07 and the pre-launch legacy response formats are retired. servers: - url: https://api.trajectory.ai security: [] tags: - name: agents x-group: Agents - name: benchmarks x-group: Specs (benchmarks) - name: diagnostics x-group: Diagnostics - name: benchmark-ingestion x-group: Spec (benchmark) submissions - name: eval x-group: Evaluations - name: training x-group: Training - name: deployments x-group: Deployments - name: model-endpoints x-group: Model endpoints - name: inference x-group: Inference - name: trajectories x-group: Trajectories - name: artifacts x-group: Artifacts - name: telemetry x-group: Telemetry - name: supported-models x-group: Supported models - name: organizations x-group: Organizations - name: secrets x-group: Secrets paths: /api/v1/train/{run_id}/step-summaries: get: tags: - training summary: Get training step summaries description: >- Retrieve trajectory, group, and termination summaries for each recorded training step. operationId: getTrainingStepSummaries parameters: - name: run_id in: path required: true schema: type: string title: Run Id - name: cursor in: query required: false schema: anyOf: - type: string - type: 'null' title: Cursor - name: limit in: query required: false schema: type: integer maximum: 200 minimum: 1 default: 25 title: Limit responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/TrainingStepSummariesPage' '400': description: >- The cursor is malformed, unsupported, or belongs to a different collection scope. headers: X-Request-ID: $ref: '#/components/headers/RequestId' Trajectory-Should-Retry: $ref: '#/components/headers/TrajectoryShouldRetry' Retry-After: $ref: '#/components/headers/RetryAfter' content: application/json: schema: $ref: '#/components/schemas/InvalidCursorErrorResponse' default: $ref: '#/components/responses/ApiErrorResponse' security: - ApiKeyAuth: [] components: schemas: TrainingStepSummariesPage: properties: items: items: $ref: '#/components/schemas/TrainingStepSummary' type: array title: Items next_cursor: anyOf: - type: string - type: 'null' title: Next Cursor has_more: type: boolean title: Has More training_run_id: type: string title: Training Run Id description: Training run whose per-step summaries are returned. type: object required: - items - has_more - training_run_id title: TrainingStepSummariesPage InvalidCursorErrorResponse: properties: error: $ref: '#/components/schemas/InvalidCursorErrorBody' type: object required: - error title: InvalidCursorErrorResponse TrainingStepSummary: properties: step: type: integer title: Step description: >- Training step these rollout, group, and termination aggregates cover. trajectory_count: type: integer title: Trajectory Count description: Rollouts generated at this training step. completed_count: type: integer title: Completed Count description: Finished rollouts at this training step. avg_turns: anyOf: - type: number - type: 'null' title: Avg Turns description: >- Mean agent turns (num_steps) per finished rollout; a turn is an agent step inside a trajectory, distinct from a training step. min_turns: anyOf: - type: integer - type: 'null' title: Min Turns description: Fewest agent turns among finished rollouts at this training step. max_turns: anyOf: - type: integer - type: 'null' title: Max Turns description: Most agent turns among finished rollouts at this training step. avg_tokens: anyOf: - type: number - type: 'null' title: Avg Tokens description: Mean generated tokens per finished rollout. min_tokens: anyOf: - type: integer - type: 'null' title: Min Tokens description: >- Fewest generated tokens among finished rollouts at this training step. max_tokens: anyOf: - type: integer - type: 'null' title: Max Tokens description: Most generated tokens among finished rollouts at this training step. tool_failure_count: type: integer title: Tool Failure Count description: >- Finished rollouts with at least one tool failure (recoverable; not folded into error_count). tool_failure_exec_count: type: integer title: Tool Failure Exec Count description: >- Finished rollouts with at least one tool call that raised while executing. tool_failure_parse_count: type: integer title: Tool Failure Parse Count description: >- Finished rollouts with at least one tool call the runtime could not parse. tool_failure_unk_count: type: integer title: Tool Failure Unk Count description: >- Finished rollouts with at least one tool call that returned neither a result nor an error. correct_count: type: integer title: Correct Count description: Finished rollouts with reward = 1. partial_count: type: integer title: Partial Count description: Finished rollouts with 0 < reward < 1. incorrect_count: type: integer title: Incorrect Count description: Finished rollouts with reward = 0. error_count: type: integer title: Error Count description: >- Finished rollouts whose termination_reason is ERROR (reward not meaningful). incomplete_count: type: integer title: Incomplete Count description: >- Finished rollouts stopped by TIMEOUT or TRUNCATION (real reward, task unfinished). termination_reasons: additionalProperties: type: integer type: object title: Termination Reasons description: Finished rollouts per termination reason. group_count: type: integer title: Group Count description: Rollout groups (distinct uid) at the training step. completed_groups: type: integer title: Completed Groups description: Groups whose rollouts are all finished with a reward. no_signal_groups: type: integer title: No Signal Groups description: >- Completed groups whose rewards are all identical. Group-relative algorithms give them zero advantage; other algorithms can train on them. all_success_groups: type: integer title: All Success Groups description: Subset of no_signal_groups with every reward >= 1. all_failed_groups: type: integer title: All Failed Groups description: Subset of no_signal_groups with every reward <= 0. used_groups: type: integer title: Used Groups description: >- Completed groups with non-identical rewards that the trainer consumed. This count need not equal the number of groups in the optimizer batch. discarded_groups: type: integer title: Discarded Groups description: >- Completed groups with reward contrast the trainer did not and will not consume: stale, or unconsumed once the step settled. pending_groups: type: integer title: Pending Groups description: >- Completed groups with reward contrast the trainer has not consumed yet but still could. stale_groups: type: integer title: Stale Groups description: Subset of discarded_groups with at least one STALE rollout. settled: type: boolean title: Settled description: >- True once used/discarded can no longer change: the run is terminal, or the trainer's latest step > step + max_steps_off_policy with a nonnegative staleness bound. type: object required: - step - trajectory_count - completed_count - avg_turns - min_turns - max_turns - avg_tokens - min_tokens - max_tokens - tool_failure_count - tool_failure_exec_count - tool_failure_parse_count - tool_failure_unk_count - correct_count - partial_count - incorrect_count - error_count - incomplete_count - termination_reasons - group_count - completed_groups - no_signal_groups - all_success_groups - all_failed_groups - used_groups - discarded_groups - pending_groups - stale_groups - settled title: TrainingStepSummary InvalidCursorErrorBody: properties: type: type: string const: invalid_request_error title: Type code: type: string const: invalid_cursor title: Code message: type: string title: Message param: anyOf: - type: string - type: 'null' title: Param doc_url: type: string title: Doc Url context: $ref: '#/components/schemas/EmptyErrorContext' type: object required: - type - code - message - doc_url - context title: InvalidCursorErrorBody ApiErrorEnvelope: properties: error: $ref: '#/components/schemas/ApiErrorBody' required: - error title: ApiErrorEnvelope type: object EmptyErrorContext: properties: {} additionalProperties: false type: object title: EmptyErrorContext ApiErrorBody: properties: type: $ref: '#/components/schemas/ErrorType' code: type: string title: Code message: type: string title: Message param: anyOf: - type: string - type: 'null' default: null title: Param doc_url: type: string title: Doc Url context: additionalProperties: true title: Context type: object required: - type - code - message - doc_url - context title: ApiErrorBody type: object ErrorType: enum: - invalid_request_error - authentication_error - permission_error - conflict_error - billing_error - rate_limit_error - dependency_error - api_error title: ErrorType type: string headers: RequestId: schema: type: string TrajectoryShouldRetry: schema: type: boolean RetryAfter: schema: type: string responses: ApiErrorResponse: description: >- Common API error. Possible codes: `authentication_required`, `permission_denied`, `validation_error`, `rate_limited`, `internal_error`, `service_unavailable`. headers: X-Request-ID: $ref: '#/components/headers/RequestId' Trajectory-Should-Retry: $ref: '#/components/headers/TrajectoryShouldRetry' Retry-After: $ref: '#/components/headers/RetryAfter' content: application/json: schema: $ref: '#/components/schemas/ApiErrorEnvelope' securitySchemes: ApiKeyAuth: type: apiKey in: header name: X-API-Key description: Organization API key ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.