Create a pipeline
POST/pipelines
Creates a pipeline with its first search indexes. embedding_model_id creates a vector index on VectorChord for similarity and maximal marginal relevance (MMR) search, and full_text_search creates a full-text index (fts) or a BM25 index (bm25) for full-text search, with the tokenizer in bm25_tokenizer. Each index becomes the default index of its kind. A pipeline created with neither field has no search index and supports only history search. The search indexes are not fixed at creation: the operations under /pipelines/{pipeline_id}/search-indexers add, switch and remove indexes later.
The request also names the default text splitter, the classifications that documents in the pipeline can carry and an optional metadata schema. A request with an unknown field, such as has_full_text_search of earlier releases, returns HTTP 422. The key that creates the pipeline receives all three permissions on the new pipeline, with the allow-list *. The response body is the new pipeline.
Permissions. Write on the pipelines object type, write on the text splitter named in default_text_splitter_id and, when the request names one, write on the embedding model.
See Pipelines.
Request
Responses
- 201
- 400
- 401
- 403
- 404
- 409
- 422
The pipeline was created. The body is the new pipeline, with its search indexes.
The metadata schema in the request is not a valid schema (Schema error, with the field in details.field), or the search settings are invalid (Invalid parameter). details.error holds the reason, such as bm25_tokenizer requires full_text_search to be bm25 or a tokenizer message listed in BM25 tokenizers. The same status reports a body that is not valid JSON.
The authentication headers are missing or invalid, or the API key is inactive or expired. Errors lists the messages.
The key lacks write permission on the pipelines object type (Unauthorized: write access to Pipeline required), the license is not valid (Invalid license: <reason>) or the license does not allow another pipeline (License limits exceeded for Pipeline).
The request names a text splitter or an embedding model that does not exist, or on which the key lacks write permission (Text Splitter not found or Embedding Model not found, with the identifier in details.id).
A pipeline with this name already exists (Conflict unique error).
A required field such as name, default_text_splitter_id or classifications is missing, a field has the wrong type, the body has an unknown field, full_text_search is neither fts nor bm25, or bm25_tokenizer has an unknown field or option value. The message names the field (Failed to deserialize the JSON body into the target type: ...).