> ## Documentation Index > Fetch the complete documentation index at: https://docs.kiteml.com/llms.txt > Use this file to discover all available pages before exploring further. # Approve Augmentation > **Beta.** Usable today, but this endpoint can change before it is stable, without a new `Kite-Version`. Changes are announced before they ship. **Beta.** Approve a background / world run whose preview passed (`awaiting_approval`): its episodes are charged and start. Not enough tokens, or the account's monthly budget hard limit reached, is a 402 and the run keeps waiting, so it can be approved after a top-up. ## OpenAPI ````yaml /api-reference/openapi.json post /v1/augmentations/{aug_id}/approve openapi: 3.1.0 info: title: Kite Platform API description: >- Public API for Kite. Authenticate with `Authorization: Bearer kite_...`, and send `Kite-Version: 2026-10-12` to pin the dated contract your code was written against (without it you get 2026-09-27). Every response carries the version that served it. version: '2026-10-12' servers: - url: https://api.kiteml.com description: Kite Platform API security: - bearerAuth: [] tags: - name: Augmentations description: Beta. Generate new episodes from a LeRobot dataset. - name: Training runs description: Fine-tune a policy on a dataset, one run per policy. - name: Twins description: >- Rebuild a robot episode as an interactable MuJoCo scene (beta, enabled per account). - name: RL runs description: >- Train a robot policy in simulation from a task spec; get the policy, its scene and a verdict. - name: Evaluations description: >- Evaluate a policy on a Kite SDK spec on Kite's machines: episodes, evaluators and a scorecard (beta). - name: Scenario runs description: >- Run a policy across built-in simulated scenarios; get its success rate, failure reasons and edge cases (beta). - name: Scenarios description: >- Beta. The scenarios a scenario run runs, and your own, composed from robots, tasks and objects. - name: Objects and environments description: Beta. Bring your own objects and environments to your scenarios. - name: Catalog description: Policies, hardware tiers, and dataset inspection. - name: Operations description: Poll any run by its id. - name: Webhooks description: Endpoints that receive events, and the event log. - name: API keys description: The key making the request. - name: Usage description: Token usage and credit balance. - name: Requests description: >- The request log: every call of the last 30 days, who made it and how it ended. - name: Status description: Whether Kite is up, part by part. Needs no key. paths: /v1/augmentations/{aug_id}/approve: post: tags: - Augmentations summary: Approve Augmentation description: >- **Beta.** Usable today, but this endpoint can change before it is stable, without a new `Kite-Version`. Changes are announced before they ship. **Beta.** Approve a background / world run whose preview passed (`awaiting_approval`): its episodes are charged and start. Not enough tokens, or the account's monthly budget hard limit reached, is a 402 and the run keeps waiting, so it can be approved after a top-up. operationId: approve_augmentation_v1_augmentations__aug_id__approve_post parameters: - name: aug_id in: path required: true schema: type: string title: Aug Id - $ref: '#/components/parameters/KiteVersion' responses: '200': description: The augmentation, running every episode content: application/json: schema: $ref: '#/components/schemas/Augmentation' example: awaiting_approval: false config: episode_count: 3 created_at: '2026-09-27T01:00:00Z' id: aug_01K62QZ8V4X9C3N7M5T2R8W6HB instructions: Make the table dark walnut and the light warm model: video object: augmentation output: files: - bytes: 3186 path: meta/info.json url: >- https://storage.googleapis.com/kite-augment/augment/meta/info.json?X-Goog-Signature=abc type: download plan: changes: - description: make the table dark walnut kind: in_front engine: video reason: The table's material changes, which needs new video. source_revision: b28c65d7b892cb8dd4ed61f45158484374330086 progress: 1 source: repo_id: lerobot/svla_so101_pickplace type: huggingface status: succeeded status_message: Output published tokens: charged: 32670 webhook_metadata: ref: run-42 '400': description: >- Invalid request (`parameter_invalid`, `unsupported_version`, ...); `param` names the field content: application/json: schema: $ref: '#/components/schemas/ApiErrorResponse' '401': description: Missing or invalid API key (`unauthenticated`) content: application/json: schema: $ref: '#/components/schemas/ApiErrorResponse' '402': content: application/json: schema: $ref: '#/components/schemas/ApiErrorResponse' description: >- Not enough credits for this run (`insufficient_tokens`), or the account's monthly budget is spent and its hard limit is on (`budget_exceeded`) '403': description: >- The key lacks a scope this route needs (`missing_scope`), or the feature is not enabled for the account content: application/json: schema: $ref: '#/components/schemas/ApiErrorResponse' '404': content: application/json: schema: $ref: '#/components/schemas/ApiErrorResponse' description: >- No such resource (`resource_not_found`); another account's ids answer the same '409': content: application/json: schema: $ref: '#/components/schemas/ApiErrorResponse' description: >- Conflicts with the resource's state or an earlier request (e.g. `not_cancelable`, `idempotency_key_reused`) '429': description: >- Rate or concurrency limited (`rate_limit_exceeded`, `concurrency_limit_exceeded`, `capacity_exceeded`); honor `Retry-After` content: application/json: schema: $ref: '#/components/schemas/ApiErrorResponse' components: parameters: KiteVersion: name: Kite-Version in: header required: false description: >- The dated contract your code was written against. Without it you get 2026-09-27; every response says which version served it. schema: type: string enum: - '2026-09-27' - '2026-10-12' default: '2026-09-27' schemas: Augmentation: properties: id: type: string title: Id description: '`aug_` + a time-ordered ULID' object: type: string enum: - augmentation const: augmentation title: Object default: augmentation status: type: string enum: - processing - succeeded - failed - canceled title: Status progress: type: number maximum: 1 minimum: 0 title: Progress description: >- Of the current phase (import, generate, publish): it starts again from 0 at each status_message: anyOf: - type: string - type: 'null' title: Status Message description: What it is doing now, or the last thing it did error: anyOf: - $ref: '#/components/schemas/AugmentationError' - type: 'null' description: Set when `failed` model: type: string title: Model description: >- `auto` until the planner decides, then `relight`, `video`, `background`, `world` or `mixed` source: $ref: '#/components/schemas/AugmentationSource' instructions: type: string title: Instructions config: $ref: '#/components/schemas/AugmentationConfig' environments: anyOf: - $ref: '#/components/schemas/AugmentationEnvironments' - type: 'null' must_keep: anyOf: - items: type: string type: array - type: 'null' title: Must Keep plan: anyOf: - $ref: '#/components/schemas/AugmentationPlan' - type: 'null' description: 'Set once the run is planned: what it does and why' awaiting_approval: type: boolean title: Awaiting Approval description: >- The preview passed and the run waits for POST /v1/augmentations/{id}/approve default: false gates: anyOf: - $ref: '#/components/schemas/AugmentationGates' - type: 'null' description: Background and world, once the episodes ran output: oneOf: - $ref: '#/components/schemas/AugmentationHuggingFaceOutput' - $ref: '#/components/schemas/AugmentationDownloadOutput' title: Output description: Where the episodes go, as given at create discriminator: propertyName: type mapping: download: '#/components/schemas/AugmentationDownloadOutput' huggingface: '#/components/schemas/AugmentationHuggingFaceOutput' tokens: $ref: '#/components/schemas/AugmentationTokens' webhook_metadata: anyOf: - type: object - type: 'null' title: Webhook Metadata created_at: anyOf: - type: string - type: 'null' title: Created At type: object required: - id - status - progress - model - source - instructions - config - output - tokens title: Augmentation description: >- One augmentation of a Hugging Face LeRobot dataset, keeping its motion: every episode relit, or new episodes generated by video augmentation. examples: - awaiting_approval: false config: episode_count: 3 created_at: '2026-09-27T01:00:00Z' id: aug_01K62QZ8V4X9C3N7M5T2R8W6HB instructions: Make the table dark walnut and the light warm model: video object: augmentation output: files: - bytes: 3186 path: meta/info.json url: >- https://storage.googleapis.com/kite-augment/augment/meta/info.json?X-Goog-Signature=abc type: download plan: changes: - description: make the table dark walnut kind: in_front engine: video reason: The table's material changes, which needs new video. source_revision: b28c65d7b892cb8dd4ed61f45158484374330086 progress: 1 source: repo_id: lerobot/svla_so101_pickplace type: huggingface status: succeeded status_message: Output published tokens: charged: 32670 webhook_metadata: ref: run-42 ApiErrorResponse: properties: error: $ref: '#/components/schemas/ApiErrorBody' type: object required: - error title: ApiErrorResponse description: Every /v1 error has this shape. AugmentationError: properties: code: type: string title: Code description: >- `system`, `generation`, `capacity`, `insufficient_tokens`, `dataset_unavailable`, `incompatible_dataset`, `hf_unauthorized`, `model_not_feasible`, `out_of_scope`, `environment_not_ready`, `environment_not_found`, `parameter_invalid`, `gates_failed`, or `pipeline_failed` when the run recorded no code message: anyOf: - type: string - type: 'null' title: Message description: What went wrong and what to do; safe to show a user type: object required: - code title: AugmentationError AugmentationSource: properties: type: type: string enum: - huggingface const: huggingface title: Type default: huggingface repo_id: type: string maxLength: 500 minLength: 3 title: Repo Id description: HF dataset id, e.g. 'lerobot/pusht' type: object required: - repo_id title: AugmentationSource AugmentationConfig: properties: episode_count: type: integer maximum: 500 minimum: 1 title: Episode Count description: >- Video augmentation: episodes to generate, from the start of the dataset (at most 50). Relighting covers every episode, and reports the dataset's count here; the background and world engines report the episodes they make (`episodes` x `variants`) default: 3 variation_prompts: anyOf: - items: type: string type: array maxItems: 500 - type: 'null' title: Variation Prompts description: One prompt per episode; overrides episode_count when provided relight: anyOf: - additionalProperties: $ref: '#/components/schemas/RelightParams' type: object - type: 'null' title: Relight description: >- Relight parameters per camera ("top" or "observation.images.top"), or "*" for every camera; a camera's own entry overrides "*" field by field. With `model` left out or `auto`, selects relighting and skips the planner. episodes: anyOf: - type: string enum: - all const: all - type: integer minimum: 1 - items: type: integer minimum: 0 type: array minItems: 1 - type: 'null' title: Episodes description: >- Background and world: the source episodes to augment, `all`, the first N, or a list of episode indices (in that order, each once) (default: `episode_count` when given or with model `auto`, else `all`). Not capped: a whole dataset runs on the augment workers preview_first: anyOf: - type: boolean - type: 'null' title: Preview First description: >- Background and world only; relight and video ignore it. Default true: augment windows at the start, middle and end of one episode per camera first and check their quality gates (a camera that fails moves to the next engine) before charging for the rest. From the console the run then waits for POST /v1/augmentations/{id}/approve; with an API key it goes on by itself once the gates pass variants: anyOf: - type: integer maximum: 10 minimum: 1 - type: 'null' title: Variants description: >- Background and world: environments per source episode (default 1); each is its own episode, with its own environment allow_video_fallback: anyOf: - type: boolean - type: 'null' title: Allow Video Fallback description: >- Model `auto`: when no scene engine can do the request on a camera (or a camera fails its preview on all of them), use generative video there instead of failing the run. Video costs many times more and keeps no pixel exactly; never used when the request asks to keep things exactly or names `must_keep` (default false: the run fails before any charge, with the measured reasons and the video price) labelling: anyOf: - type: string enum: - rig - episode - auto - type: 'null' title: Labelling description: >- Background and world: what the background segmenter learns from. `rig`: the rig profile's sites only; an episode filmed where no profiled site serves a camera (a new room or camera placement) is skipped, not charged. `episode`: such an episode labels that camera itself (a segmenter trained on detections of the episode's own frames and probed on them, which decides the camera), for a surcharge per labelled camera (its GPU time: a fixed part plus one per frame) shown in `plan.estimate.labelling_tokens`. `auto` (default): `episode` when `allow_episode_labelling` is true, else `rig` allow_episode_labelling: anyOf: - type: boolean - type: 'null' title: Allow Episode Labelling description: >- With `labelling` `auto`: let episodes the rig profile can't serve label themselves, for the surcharge (default false: they are skipped, and the plan prices this option where the rig profile measured it passing) allow_environment_substitutes: anyOf: - type: boolean - type: 'null' title: Allow Environment Substitutes description: >- Background and world: when the place the instructions name matches no ready environment ("a forest"), use environments matched to the rig's light instead, with a warning (default false: the run fails before any charge with `environment_not_found`) type: object title: AugmentationConfig AugmentationEnvironments: properties: ids: anyOf: - items: type: string type: array maxItems: 20 minItems: 1 - type: 'null' title: Ids description: Environment ids (`aenv_...`) or library ids, used in turn filter: anyOf: - $ref: '#/components/schemas/EnvironmentFilter' - type: 'null' prompt: anyOf: - type: string maxLength: 2000 minLength: 3 - type: 'null' title: Prompt description: >- Make new environments from a description (minutes; not enabled for every account). The run waits for them count: anyOf: - type: integer maximum: 20 minimum: 1 - type: 'null' title: Count description: How many to pick or make (default 3) type: object title: AugmentationEnvironments description: >- What the background and world engines put behind the work cell: exactly one of `ids` or `prompt`, or a `filter` and/or `count` to pick from your ready environments and the Kite library (fitted to the rig's light). AugmentationPlan: properties: engine: type: string enum: - relight - video - background - world - mixed title: Engine description: >- `relight`: colour and exposure of every episode, motion untouched. `video`: generative video augmentation. `background` / `world`: a new place behind the work cell (`mixed` when cameras differ, see `cameras`) reason: anyOf: - type: string - type: 'null' title: Reason description: Why, in a sentence relight: anyOf: - additionalProperties: $ref: '#/components/schemas/RelightParams' type: object - type: 'null' title: Relight description: >- For `relight`: parameters per camera, or `*` for every camera; a camera's own entry overrides `*` field by field. Cameras left out are copied unchanged video_instruction: anyOf: - type: string - type: 'null' title: Video Instruction description: 'For `video` chosen from reference photos: the edit derived from them' source_revision: anyOf: - type: string - type: 'null' title: Source Revision description: >- For `relight` and the scene engines: the commit of the source dataset the run reads from start to finish cameras: anyOf: - additionalProperties: $ref: '#/components/schemas/AugmentationCameraPlan' type: object - type: 'null' title: Cameras description: 'Background and world: each camera''s engines and why' changes: anyOf: - items: $ref: '#/components/schemas/AugmentationChange' type: array - type: 'null' title: Changes description: What the request changes, as classified must_keep: anyOf: - items: type: string type: array - type: 'null' title: Must Keep environments: anyOf: - items: $ref: '#/components/schemas/AugmentationPlanEnvironment' type: array - type: 'null' title: Environments description: Used in turn, one per episode and variant rig_profile: anyOf: - $ref: '#/components/schemas/AugmentationRigProfile' - type: 'null' estimate: anyOf: - $ref: '#/components/schemas/AugmentationPlanEstimate' - type: 'null' description: What the episodes will cost, charged when they start preview: anyOf: - $ref: '#/components/schemas/AugmentationPreview' - type: 'null' description: 'With `config.preview_first`: one episode per camera, in windows' feasibility: anyOf: - $ref: '#/components/schemas/AugmentationFeasibility' - type: 'null' labelling: anyOf: - type: string enum: - rig - episode - type: 'null' title: Labelling description: >- Background and world: `episode` when episodes the rig profile can't serve label themselves (`config.labelling`), else `rig` warnings: anyOf: - items: type: string type: array - type: 'null' title: Warnings description: >- What the plan could not do as asked (an option left unused, episodes skipped, a change not made) type: object required: - engine title: AugmentationPlan AugmentationGates: properties: passed: type: integer title: Passed description: Episodes that passed every gate and shipped failed: type: integer title: Failed description: Episodes that failed a gate or could not run; refunded failures: items: $ref: '#/components/schemas/AugmentationGateFailure' type: array title: Failures description: The first 20 gate failures type: object required: - passed - failed - failures title: AugmentationGates description: >- Background and world: every episode is measured before it ships, and one that fails a gate never does. AugmentationHuggingFaceOutput: properties: type: type: string enum: - huggingface const: huggingface title: Type default: huggingface repo_id: anyOf: - type: string - type: 'null' title: Repo Id description: The Hugging Face dataset the episodes are pushed to url: anyOf: - type: string - type: 'null' title: Url description: 'Once `succeeded`: the pushed dataset on huggingface.co' type: object title: AugmentationHuggingFaceOutput AugmentationDownloadOutput: properties: type: type: string enum: - download const: download title: Type default: download files: anyOf: - items: $ref: '#/components/schemas/AugmentationOutputFile' type: array - type: 'null' title: Files description: >- Only on GET /v1/augmentations/{id} once `succeeded`: every file of the LeRobot dataset. The key is absent everywhere else (lists, create, cancel, webhook events) type: object title: AugmentationDownloadOutput AugmentationTokens: properties: charged: type: integer title: Charged description: >- Charged up front when the work starts. Video augmentation refunds the episodes it never generates, and a relight that fails is refunded in full, when the run ends (see GET /v1/usage); refunds are not subtracted here type: object required: - charged title: AugmentationTokens ApiErrorBody: properties: type: type: string title: Type description: >- invalid_request_error, authentication_error, permission_error, not_found_error, conflict_error, rate_limit_error, or api_error code: type: string title: Code description: 'Machine-readable: branch on this, not on the message' message: type: string title: Message param: anyOf: - type: string - type: 'null' title: Param description: The request field the error is about, when there is one request_id: anyOf: - type: string - type: 'null' title: Request Id description: Also in the X-Request-Id header; quote it to support details: anyOf: - type: object - type: 'null' title: Details description: >- Extra data for some codes (e.g. the cameras a dataset does have, for camera_not_found) type: object required: - type - code - message title: ApiErrorBody RelightParams: properties: exposure: type: number maximum: 3 minimum: -5 title: Exposure description: Brightness in stops (-1 = half the light). default: 0 temperature: type: number maximum: 12000 minimum: 1500 title: Temperature description: >- Colour temperature of the simulated light in kelvin. 6500 = unchanged, ~3000 = warm tungsten, ~9000 = cool overcast. Never changes brightness. default: 6500 tint: type: number maximum: 1 minimum: -1 title: Tint description: Green (-) to magenta (+) shift. default: 0 contrast: type: number maximum: 2 minimum: 0.5 title: Contrast description: 'S-curve around mid-grey: >1 deepens shadows and lifts highlights.' default: 1 saturation: type: number maximum: 2 minimum: 0 title: Saturation description: Colour intensity; 1 = unchanged. default: 1 type: object title: RelightParams EnvironmentFilter: properties: category: anyOf: - type: string maxLength: 40 - type: 'null' title: Category description: >- workshop, warehouse, hangar, garage, office, factory, lab, hall, or your own lighting: anyOf: - type: string enum: - daylight - artificial - type: 'null' title: Lighting type: object title: EnvironmentFilter AugmentationCameraPlan: properties: engines: items: type: string type: array title: Engines description: >- Run in order on this camera: `background`, `world` or `video`, then `relight` when the light changes too; none when it sees nothing behind the work cell and passes through reason: type: string title: Reason episodes: anyOf: - additionalProperties: type: integer type: object - type: 'null' title: Episodes description: >- Of the selected source episodes, by the rig site each is on: how many this camera can run its scene engine on (`scene`), passes through on because it sees nothing behind the work cell there (`pass_through`), labels from the episode itself (`labelled`, `config.labelling`), or can't run on (`blocked`: the episode is skipped) type: object required: - engines - reason title: AugmentationCameraPlan AugmentationChange: properties: description: type: string title: Description kind: type: string enum: - photometric - behind_cell - in_front - dynamic - out_of_scope title: Kind description: >- `dynamic`: people or something that moves, which only generative video can animate type: object required: - description - kind title: AugmentationChange AugmentationPlanEnvironment: properties: id: type: string title: Id name: anyOf: - type: string - type: 'null' title: Name yaw_deg: anyOf: - type: number - type: 'null' title: Yaw Deg description: >- How far the panorama is turned, so its daylight falls where the rig's light comes from type: object required: - id title: AugmentationPlanEnvironment AugmentationRigProfile: properties: key: type: string title: Key description: The rig's profile, reused by later runs on the same rig robot_model: type: string title: Robot Model cameras: additionalProperties: $ref: '#/components/schemas/AugmentationRigCamera' type: object title: Cameras type: object required: - key - robot_model - cameras title: AugmentationRigProfile AugmentationPlanEstimate: properties: tokens: type: integer title: Tokens episodes: type: integer title: Episodes description: 'Episodes it makes: source episodes x environment variants' labelling_tokens: anyOf: - type: integer - type: 'null' title: Labelling Tokens description: >- Of `tokens`, the surcharge for episodes that label cameras themselves (`config.labelling`) type: object required: - tokens - episodes title: AugmentationPlanEstimate AugmentationPreview: properties: episode_index: type: integer title: Episode Index passed: type: boolean title: Passed description: Every camera passed (some perhaps on a fallback engine) cameras: additionalProperties: $ref: '#/components/schemas/AugmentationPreviewCamera' type: object title: Cameras type: object required: - episode_index - passed - cameras title: AugmentationPreview AugmentationFeasibility: properties: selected: type: integer title: Selected description: Source episodes selected by `config.episodes` runnable: type: integer title: Runnable description: Those every camera can run on; only these are charged and run skipped: items: $ref: '#/components/schemas/AugmentationEpisodeCount' type: array title: Skipped description: Why the others are skipped, most common first type: object required: - selected - runnable - skipped title: AugmentationFeasibility AugmentationGateFailure: properties: episode_index: type: integer title: Episode Index camera: anyOf: - type: string - type: 'null' title: Camera reasons: items: type: string type: array title: Reasons type: object required: - episode_index - reasons title: AugmentationGateFailure AugmentationOutputFile: properties: path: type: string title: Path description: Relative to the dataset root, e.g. `meta/info.json` bytes: type: integer title: Bytes url: type: string title: Url description: >- A signed storage link (no credentials; valid for an hour), or this API's own `/v1/augmentations/{id}/files/{path}` when signing is unavailable, which needs your Authorization header source_url: anyOf: - type: string - type: 'null' title: Source Url description: >- For a relight's videos: the same file of the source dataset on Hugging Face, at the commit the run read, with the same timestamps (the before to `url`'s after). Null for every other file type: object required: - path - bytes - url title: AugmentationOutputFile AugmentationRigCamera: properties: motion: type: string enum: - static - moving title: Motion located: type: boolean title: Located description: Its lens and pose were measured, not guessed segmenter: anyOf: - type: boolean - type: 'null' title: Segmenter description: >- The rig's background segmenter passed its probe here; null when not recorded travel_m: type: number title: Travel M description: How far it moves over episode 0, from the robot's kinematics type: object required: - motion - located - travel_m title: AugmentationRigCamera AugmentationPreviewCamera: properties: engines_used: items: type: string type: array title: Engines Used passed: anyOf: - type: boolean - type: 'null' title: Passed description: null when not previewed (video, or passing through) episode_index: anyOf: - type: integer - type: 'null' title: Episode Index description: The source episode previewed on this camera windows: anyOf: - items: type: object type: array - type: 'null' title: Windows description: 'Each previewed window: `start` frame, `frames`, `passed`, `metrics`' gates: type: object title: Gates description: >- The quality gates' `metrics` (the visual judge's as `judge`: `passed`, `failed` or `skipped`) and `failures` video_url: anyOf: - type: string - type: 'null' title: Video Url description: This API's path for the preview clip type: object required: - engines_used - gates title: AugmentationPreviewCamera AugmentationEpisodeCount: properties: reason: type: string title: Reason episodes: type: integer title: Episodes type: object required: - reason - episodes title: AugmentationEpisodeCount securitySchemes: bearerAuth: type: http scheme: bearer description: >- A Kite API key (`kite_...`), from **Platform API → Keys** in the dashboard. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.