Hybrid keyword + vector search across every page you've indexed. No infrastructure to manage, no tuning required — just results in under 300ms.
START FOR FREE →Combines BM25 keyword matching with vector similarity — best of both worlds in a single query.
Content is embedded at index time with OpenAI text-embedding-3-small, so queries are instant.
Each API key has its own Typesense collection — no shared latency, no cold starts.
Deployed at the edge with sub-millisecond data access. Results are back before the user blinks.
Call /v1/extract with any URL. We scrape, clean, and chunk the content automatically.
Every chunk is embedded and stored in your private Typesense collection with zero config.
Call /v1/search with a natural language query. Hybrid ranking returns the most relevant chunks.
Pipe the results into your LLM context window. Grounded, relevant, always up-to-date.
Index any content with /v1/extract, then search with /v1/search. No vector database to provision, no embedding pipeline to build.
Every result includes the source URL, a relevance score, and the matched chunk — ready to inject into your LLM prompt.
INDEX CONTENT
// Extract & store any URL const res = await fetch("https://api.knowledgesdk.com/v1/extract", { method: "POST", headers: { "x-api-key": "knowledgesdk_live_..." }, body: JSON.stringify({ url: "https://docs.example.com", store: true, }) });
SEARCH IN PLAIN ENGLISH
// Semantic search — results in <300ms const { results } = await fetch( 'https://api.knowledgesdk.com/v1/search', { method: "POST", body: JSON.stringify({ query: "how does authentication work?", limit: 5, }) } ).then(r => r.json()); // results[0] → { url, chunk, score }