Sign in
Embeddings

Embed documents

Perform embedding on a Telnyx Storage Bucket using the a embedding model. The current supported file types are:

  • PDF
  • HTML
  • txt/unstructured text files
  • json
  • csv
  • audio / video (mp3, mp4, mpeg, mpga, m4a, wav, or webm ) - Max of 100mb file size.

Any files not matching the above types will be attempted to be embedded as unstructured text.

This process can be slow, so it runs in the background and the user can check the status of the task using the endpoint /ai/embeddings/{task_id}.

Important Note: When you update documents in a Telnyx Storage bucket, their associated embeddings are automatically kept up to date. If you add or update a file, it is automatically embedded. If you delete a file, the embeddings are deleted for that particular file.

You can also specify a custom loader param. Currently the only supported loader value is intercom which loads Intercom article jsons as specified by the Intercom article API This loader will split each article into paragraphs and save additional parameters relevant to Intercom docs, such as article_url and heading. These values will be returned by the /v2/ai/embeddings/similarity-search endpoint in the loader_metadata field.

post/ai/embeddings

Headers

Idempotency-Keystring

Optional opaque, unquoted key for safely retrying the same logical request. Keys must contain 1 to 255 letters, numbers, hyphens, or underscores. Generate a unique UUID v4 for each operation and reuse it only when retrying that operation with the same request. Invalid headers—including duplicate, empty, malformed, or overlong values—return 400 with error code 10015. A request already in progress with the same key returns 409; reusing the key with a different request returns 422. Only successful responses are replayed, for up to 24 hours. Do not include sensitive data in the key.

Request body

bucket_namestring required
document_chunk_sizeinteger
document_chunk_overlap_sizeinteger
embedding_model'thenlper/gte-large' | 'intfloat/multilingual-e5-large'

Supported models to vectorize and embed documents.

loader'default' | 'intercom'

Supported types of custom document loaders for embeddings.

Response

Successful Response

Changes