> ## Documentation Index > Fetch the complete documentation index at: https://docs.coreweave.com/llms.txt > Use this file to discover all available pages before exploring further. # Add or replace dataset entries > Writes up to 5,000 OpenAI chat rows in one transaction. Send a JSON array, or the same rows as NDJSON (one row per line) with `Content-Type: application/x-ndjson`. Any invalid row rejects the request and is reported by its index; NDJSON line N is index N - 1. `mode=append` rejects a `row_id` that already exists (409 `entry_exists`); `mode=upsert` replaces it, keeping the stored `group_id`, `split`, `metadata`, and `provenance` for any of those the row omits. New entries default to `train` and their own group. A write that would place one group in more than one split returns 409 `group_split_conflict`. Writing moves the dataset to a new revision and marks completed relabels and evaluations stale. ## OpenAPI ````yaml openapi/model-distillation/management.openapi.yaml POST /projects/{alias}/datasets/{datasetId}/entries openapi: 3.1.0 info: title: CoreWeave Model Distillation Management API version: 1.0.0 description: | Manage providers, projects, routing versions, datasets, relabeling, fine-tunes, evaluations, and analytics. This is the same API used by Model Distillation. servers: - url: https://forge.coreweave.com/api/distillation/v1 description: Production security: - WandbCredential: [] tags: - name: Providers description: OpenAI-compatible endpoints, credentials, model catalogs, and pricing. - name: Projects description: Stable project identity and project-level information. - name: Routing description: Versioned model targets, weights, and request parameters. - name: Datasets description: >- Reproducible data snapshots, entries, relabeling, and reusable model outputs. - name: Fine-tunes description: Supervised fine-tuning jobs and hosted model artifacts. - name: Evaluations description: Head-to-head, exact-match, and categorization evaluations. - name: Analytics description: Project traffic, token, error, and estimated-cost read models. - name: UI preferences description: Per-project display preferences used by the UI. paths: /projects/{alias}/datasets/{datasetId}/entries: parameters: - $ref: '#/components/parameters/WandbEntity' - $ref: '#/components/parameters/ProjectAlias' - $ref: '#/components/parameters/DatasetId' post: tags: - Datasets summary: Add or replace dataset entries description: >- Writes up to 5,000 OpenAI chat rows in one transaction. Send a JSON array, or the same rows as NDJSON (one row per line) with `Content-Type: application/x-ndjson`. Any invalid row rejects the request and is reported by its index; NDJSON line N is index N - 1. `mode=append` rejects a `row_id` that already exists (409 `entry_exists`); `mode=upsert` replaces it, keeping the stored `group_id`, `split`, `metadata`, and `provenance` for any of those the row omits. New entries default to `train` and their own group. A write that would place one group in more than one split returns 409 `group_split_conflict`. Writing moves the dataset to a new revision and marks completed relabels and evaluations stale. operationId: writeDatasetEntries parameters: - name: mode in: query description: >- append rejects rows whose row_id already exists; upsert replaces them. required: false schema: default: append type: string enum: - append - upsert - name: dry_run in: query description: Validate and report the change, then roll it back. required: false schema: type: boolean default: false requestBody: required: true content: application/json: schema: minItems: 1 maxItems: 5000 type: array items: type: object properties: row_id: description: >- Stable identity within the dataset. Upserts match on it. Defaults to a digest of the messages and request fields. type: string minLength: 1 maxLength: 512 group_id: description: >- Rows that must share a split, such as the turns of one conversation. Defaults to row_id. type: string minLength: 1 maxLength: 512 split: description: >- train, or val to hold the row out of training for evaluation. New rows default to train. type: string enum: - train - val messages: minItems: 2 maxItems: 1000 type: array items: type: object propertyNames: type: string additionalProperties: {} tools: maxItems: 128 type: array items: type: object properties: type: type: string const: function function: type: object properties: name: type: string minLength: 1 maxLength: 256 description: type: string maxLength: 16384 parameters: type: object propertyNames: type: string additionalProperties: {} strict: type: boolean required: - name additionalProperties: false required: - type - function additionalProperties: false tool_choice: {} response_format: {} metadata: description: >- Filterable attributes, such as the user or customer segment. type: object propertyNames: type: string additionalProperties: {} provenance: description: >- Where the row came from, such as the source system and trace id. type: object propertyNames: type: string additionalProperties: {} required: - messages additionalProperties: false application/x-ndjson: schema: type: string description: >- One JSON row per line, in the same shape as the JSON array's items. responses: '200': description: Entries were written, or would be for a dry run. content: application/json: schema: $ref: '#/components/schemas/DatasetEntryWriteResult' default: $ref: '#/components/responses/Error' components: parameters: WandbEntity: name: Wandb-Entity in: header required: false description: >- Accessible personal or team entity. Omit to use the authenticated user's default entity. schema: type: string ProjectAlias: name: alias in: path required: true schema: type: string pattern: ^[a-z0-9-]{1,64}$ DatasetId: name: datasetId in: path required: true schema: type: string format: uuid schemas: DatasetEntryWriteResult: type: object required: - dry_run - created - updated - revision - entry_counts - group_conflicts properties: dry_run: type: boolean created: type: integer description: Rows that became new entries. updated: type: integer description: Existing entries the rows replaced. revision: type: integer description: The dataset revision after the write; unchanged by a dry run. entry_counts: $ref: '#/components/schemas/DatasetEntryCounts' group_conflicts: type: array items: type: string description: >- Groups spanning more than one split. A real write with conflicts is rejected. DatasetEntryCounts: type: object required: - total - train - val properties: total: type: integer train: type: integer val: type: integer ErrorResponse: type: object required: - error properties: error: type: object required: - message - type properties: message: type: string type: type: string example: error: message: Project 'missing' not found in entity 'your-team' type: not_found responses: Error: description: Request failed. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' securitySchemes: WandbCredential: type: http scheme: bearer bearerFormat: wandb API key or JWT description: >- A wandb API key or wandb JWT. Browser sessions are exchanged by the Model Distillation web app backend; job capabilities are internal only. ````