Filter most relevant chunks

This endpoint can be used to filter document chunks based on a given query. The query should be a single string. Chunk ids should be a list of strings. Each string should be the UUID of an existing document chunk. The model field should be the name of the model to use for filtering. An optional integer n can be provided, which can be used to return the first n chunks that pass the filter.

POST /api/v2/filter/chunks

Try it

Filter most relevant chunks

cURL

curl --request POST \
  --url https://paradigm.lighton.ai/api/v2/filter/chunks \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "query": "<string>",
  "chunk_ids": [
    "<string>"
  ],
  "model": "<string>",
  "n": 123
}
'

200

{
  "query": "<string>",
  "chunks": [
    {
      "uuid": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
      "text": "<string>",
      "metadata": {},
      "filter_score": 123
    }
  ]
}

Authorizations

Authorization · string · header · required

Bearer token authentication

Body

Request serializer for filtering document chunks based on a query.

query · string · required

The query string to filter document chunks against

chunk_ids · string[] · required

List of document chunk UUIDs to filter

model · string · required

Model to use for the filtering, must exist and be configured. If no model is given, will try to use the 'reranker' finetune.

n · integer | null

Maximum number of chunks to return (optional)

Response

200

application/json

Chunks retrieved

Response serializer for document chunk filtering results.

query · string · required

The original query used for filtering

chunks · object[] · required

List of filtered document chunks with their relevance scores

Show child attributes

chunks. uuid · string<uuid> · required

Unique identifier for the document chunk

chunks. text · string · required

Text content of the document chunk

chunks. metadata · object · required

Metadata associated with the document chunk

Show child attributes

chunks.metadata. {key} · any

chunks. filter_score · number<double> · required

Relevance score assigned by the filtering model