<!-- Generated from the canonical Teloring documentation source. Do not edit this file. -->

Canonical page: https://docs.teloring.com/api/list-knowledge-base-chunks
Last modified: 2026-08-20T20:52:05.000Z

# List indexed chunks

import MethodEndpoint from "@theme/ApiExplorer/MethodEndpoint";
import ParamsDetails from "@theme/ParamsDetails";
import RequestSchema from "@theme/RequestSchema";
import StatusCodes from "@theme/StatusCodes";
import OperationTabs from "@theme/OperationTabs";
import TabItem from "@theme/TabItem";
import Heading from "@theme/Heading";
import Translate from "@docusaurus/Translate";

<Heading
  as={"h1"}
  className={"openapi__heading"}
  children={"List indexed chunks"}
>
</Heading>

<MethodEndpoint
  method={"get"}
  path={"/knowledge-bases/{kb_id}/chunks"}
  context={"endpoint"}
>
  
</MethodEndpoint>



The indexed passages in a knowledge base — literally what the AI can see.

This is the endpoint that lets you audit a knowledge base: if the assistant is
answering oddly, the answer is usually visible here. Filter to one source with
`source_id`, or pass `include_content=false` for metadata only, which is much
cheaper when walking a large base.

Embedding vectors are never returned: a thousand floats per chunk, meaningless
outside the model that produced them, and they would dwarf every other field.

<Heading
  id={"request"}
  as={"h2"}
  className={"openapi-tabs__heading"}
>
  <Translate id="theme.openapi.request.title">Request</Translate>
</Heading>

<ParamsDetails
  {...require("./list-knowledge-base-chunks.ParamsDetails.json")}
>
  
</ParamsDetails>

<RequestSchema
  {...require("./list-knowledge-base-chunks.RequestSchema.json")}
>
  
</RequestSchema>

<StatusCodes
  {...require("./list-knowledge-base-chunks.StatusCodes.json")}
>
  
</StatusCodes>
