> ## Documentation Index > Fetch the complete documentation index at: https://aspect.inc/docs/llms.txt > Use this file to discover all available pages before exploring further. # Run Generation > Start an AI generation run for the attribute. `mode` says what the run does: `fill` generates values only for assets not generated yet (or whose last run failed); `overwrite` regenerates every asset's value, replacing the AI's answers. Values entered by hand are kept unless `replace_manual` is true, and `count` leaves them out accordingly; an `overwrite` dry run also reports how many there are (`manually_entered`). `target` says which assets it covers: `watch` = every asset in the attribute's watch folders; `directories` = assets in the given directories (`recursive: true` includes their subdirectories); `assets` = only the listed assets, wherever they are. Requires edit-metadata access on the project for `watch`, on every listed directory or asset otherwise. An attribute without generation attached has nothing to run and is rejected with 409 once the target is validated. `POST /metadata/attributes/{attribute_id}/generation/run` is the deprecated form of this route. ## OpenAPI ````yaml /api-reference/openapi.json post /metadata/attributes/{attribute_id}/run openapi: 3.1.0 info: title: Aspect API description: Manage and search media in Aspect. version: 1.0.0 servers: - url: https://api.aspect.inc security: - ApiKeyAuth: [] tags: - name: Users - name: API Keys - name: Workspaces - name: Projects - name: Directories - name: Assets - name: Attachments - name: Filesystem Items - name: Jobs - name: Collections - name: Comments - name: Search - name: Transcription - name: Metadata - name: Shares - name: Share Templates - name: Watermarks paths: /metadata/attributes/{attribute_id}/run: post: tags: - Metadata summary: Run Generation description: >- Start an AI generation run for the attribute. `mode` says what the run does: `fill` generates values only for assets not generated yet (or whose last run failed); `overwrite` regenerates every asset's value, replacing the AI's answers. Values entered by hand are kept unless `replace_manual` is true, and `count` leaves them out accordingly; an `overwrite` dry run also reports how many there are (`manually_entered`). `target` says which assets it covers: `watch` = every asset in the attribute's watch folders; `directories` = assets in the given directories (`recursive: true` includes their subdirectories); `assets` = only the listed assets, wherever they are. Requires edit-metadata access on the project for `watch`, on every listed directory or asset otherwise. An attribute without generation attached has nothing to run and is rejected with 409 once the target is validated. `POST /metadata/attributes/{attribute_id}/generation/run` is the deprecated form of this route. operationId: run_generation parameters: - name: attribute_id in: path required: true schema: type: string format: uuid description: Attribute ID title: Attribute Id description: Attribute ID - name: dry_run in: query required: false schema: type: boolean description: >- Return the counts without starting the run; not supported when `target.kind` is `assets` default: false title: Dry Run description: >- Return the counts without starting the run; not supported when `target.kind` is `assets` requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/CellRunRequest' responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/CellRunResponse' '422': description: Validation Error content: application/json: schema: $ref: '#/components/schemas/HTTPValidationError' components: schemas: CellRunRequest: properties: mode: $ref: '#/components/schemas/CellMode' title: Run Mode description: >- `fill` runs only the assets not run yet (or whose last run failed); `overwrite` re-runs every asset, replacing the AI's answers; answers entered by hand are kept unless `replace_manual` is true replace_manual: type: boolean title: Replace Manual Answers description: >- With `overwrite`, also replace answers entered by hand. Ignored by `fill` default: false target: oneOf: - $ref: '#/components/schemas/WatchRunTarget' - $ref: '#/components/schemas/DirectoriesRunTarget' - $ref: '#/components/schemas/AssetsRunTarget' title: Run Target description: >- Which assets the run covers: `watch` = every asset in the watch folders, rejected when there are none; `directories` = assets in the given directories (`recursive: true` includes their subdirectories); `assets` = only the listed assets, wherever they are discriminator: propertyName: kind mapping: assets: '#/components/schemas/AssetsRunTarget' directories: '#/components/schemas/DirectoriesRunTarget' watch: '#/components/schemas/WatchRunTarget' type: object required: - mode - target title: CellRunRequest description: 'API request schema for a run: what it does and which assets it covers.' CellRunResponse: properties: count: type: integer title: Count description: >- Number of assets the run covers (with `dry_run`, the number it would cover) manually_entered: anyOf: - type: integer - type: 'null' title: Manually Entered description: >- How many assets of the target currently hold an answer entered by hand. Only returned for `overwrite` dry runs; `count` leaves them out unless `replace_manual` is true type: object required: - count title: CellRunResponse description: API response schema for an explicit run's counts. HTTPValidationError: properties: detail: items: $ref: '#/components/schemas/ValidationError' type: array title: Detail type: object title: HTTPValidationError CellMode: type: string enum: - fill - overwrite title: CellMode description: |- What a run does to cells that already hold a value. ``fill`` stands down when the cell already has a value; ``overwrite`` always runs and replaces whatever is there (values entered by hand only when the run asks to replace them). A missing payload ``mode`` reads as ``fill``, and modes are only ever promoted fill -> overwrite, never demoted. WatchRunTarget: properties: kind: type: string const: watch title: Kind type: object required: - kind title: WatchRunTarget description: Every asset the column watches; rejected when it watches nothing. DirectoriesRunTarget: properties: kind: type: string const: directories title: Kind directory_ids: items: type: string format: uuid type: array minItems: 1 title: Directory IDs description: >- Directories in the column's project; the trash directory is not allowed recursive: type: boolean title: Recursive description: >- `true` includes each directory's subdirectories; `false` covers only the assets directly inside it type: object required: - kind - directory_ids - recursive title: DirectoriesRunTarget description: |- Assets in the given directories (`recursive: true` includes their subdirectories), of every type the column runs on: the watch and its asset-type limit do not apply. AssetsRunTarget: properties: kind: type: string const: assets title: Kind asset_ids: items: type: string format: uuid type: array minItems: 1 title: Asset IDs description: Assets to run on type: object required: - kind - asset_ids title: AssetsRunTarget description: Only the listed assets, wherever they are. ValidationError: properties: loc: items: anyOf: - type: string - type: integer type: array title: Location msg: type: string title: Message type: type: string title: Error Type type: object required: - loc - msg - type title: ValidationError securitySchemes: ApiKeyAuth: type: apiKey in: header name: Authorization description: Aspect API key (sk_...) in Authorization header. 'Bearer ' optional. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.