Skip to content
CallMissed
Book a demo

Explore CallMissed

Knowledge & retrievalYour knowledge. In the conversation.

Make your product information, policies and reference material available as searchable sources. The Knowledge API turns content into a retrieval layer for your application.

Open Knowledge & retrievalRead the guide
An ink-wash card-catalogue cabinet with index cards fanned out

Simulated example

  1. Add the source

    Example source / A product care guide.

  2. It is indexed

    The passage about cleaning the ceramic surface is retrieved.

  3. Ask the question

    The answer uses the retrieved passage as its context.

In the console

A web page becomes an answer

Point it at a page, let it be chunked and indexed, and an agent answers a customer from what that page says.

Demo: adding a web page as a knowledge source and an agent answering a customer from it

Inside Knowledge & retrieval

Ingest documents and URLs, then retrieve relevant content for an agent or application.

  • Text sources
  • URL ingestion
  • PDF sources
  • Semantic search
  • Bot-scoped retrieval
  • Source management

Add useful sources

Ingest supported text, URLs and PDFs into the vector Knowledge API. Associate each source with the bot it belongs to.

Search by meaning

Retrieve relevant chunks with semantic search rather than requiring an exact keyword match.

Keep context scoped

Use workspace-scoped retrieval and a bot filter when looking up sources for a specific agent.

Manage the source material

Review ingestion state and manage sources as information changes. Use retrieved passages as context rather than treating an upload as proof of a correct answer.

An ink-and-wash library shelf of reference books and index cards

Follow a source into an answer

Ingest a simulated source, then ask a question and inspect the passages the retrieval layer returns.

1 · Add a source

  1. Received
  2. Text extracted
  3. Indexed

2 · Ask a question

Ingest a source first — retrieval only searches indexed material.

Simulated example — a retrieved passage is context for the answer, not proof it is correct.

Three steps to a working setup

  1. 1

    Choose the bot and gather the documents or URLs it should reference.

  2. 2

    Ingest the sources through the Knowledge API and check their processing state.

  3. 3

    Test representative searches and configure the agent to use the relevant knowledge.

Read this before you start

Use the vector Knowledge API for retrieval. The older flat bot document store is separate and is not searched by RAG. Supported file formats and text extraction limits apply; a scanned document may need preprocessing.

Create an account, then open Knowledge & retrieval in the console.

Create a free accountSee pricing

Works alongside

Need a system connected first? Browse the integrations directory