Concepts

Content

Content is text you push into a knowledge base. Once ingested, Kognita takes care of everything needed to make it searchable. You just query.

What Kognita manages for you

After you create a content item, Kognita handles the entire processing pipeline asynchronously in the background:

Splitting

Your text is broken into optimally-sized windows. The split strategy is tuned per knowledge base type so you don't have to think about token budgets.

Embedding

Each window is converted into a vector using the embedding model configured on the knowledge base. Kognita calls the model, pays for the tokens, and stores the result.

Indexing

Vectors are written to a dedicated vector database. A full-text index is built at the same time, enabling both keyword and semantic retrieval from a single API.

Re-embedding on update

When you update a content item via PUT, Kognita automatically regenerates all embeddings. There is no manual re-index step.

Cleanup on delete

Deleting a content item removes its vectors, full-text index entries, and stored text. Nothing is left behind.

Ingestion methods

You can push content as raw text or as a file upload:

Method	Endpoint	Use when
Text	`POST /contents`	You already have the text (FAQs, articles, product descriptions, structured data).
File upload	`POST /contents/file`	You have a document (PDF, etc.). Kognita extracts the text before processing.

Metadata

Every content item can carry arbitrary key/value metadata. Metadata is stored with the content, returned in search results, and can be used to filter searches so only relevant content is considered.

Ingest with metadata

curl -X POST https://api.kognita.ai/api/v1/knowledge-bases/{kbId}/contents \
  -H "x-kognita-api-key: YOUR_API_KEY" \
  -H "x-kognita-organization-id: YOUR_ORG_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Kognita provides hybrid search over your data.",
    "metadata": { "source": "docs", "version": "2", "language": "en" }
  }'

Filter by metadata at search time

curl -X POST https://api.kognita.ai/api/v1/knowledge-bases/{kbId}/search/hybrid \
  -H "x-kognita-api-key: YOUR_API_KEY" \
  -H "x-kognita-organization-id: YOUR_ORG_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "how does search work?",
    "limit": 10,
    "metadataFilter": { "source": "docs", "language": "en" }
  }'

Tips

Keep items focused. One article, FAQ entry, or product description per content item gives search more signal to work with. Giant blobs reduce precision; tiny fragments lack context.
Use metadata to segment your content. If your KB holds content for multiple products, languages, or categories, tag each item at ingest time. You can then filter searches to a specific subset without creating separate knowledge bases.
Updates re-embed automatically. Sending a PUT replaces the text and triggers a full re-index. There is no separate re-embed step.

→ Search types → API reference: Content