Knowledge Base

The Knowledge Base is the information your AI bot draws on when answering customer questions. When a customer asks something, the bot searches the knowledge base for relevant content, then generates a response grounded in what it finds.

How it works

1. You add a knowledge source (a URL, a document, or structured content)
2. Hydra processes the content — scraping, chunking, and embedding it into a searchable vector index
3. When a customer asks a question, the bot finds the most relevant chunks and uses them to answer

The bot only answers based on what's in the knowledge base plus your system prompt. If information isn't there, it will say so rather than guess.

Source types

URL

Provide a web page URL. Hydra uses Firecrawl to scrape and extract the page content, then processes the text.

Use this for: documentation pages, product pages, FAQ pages, blog posts, help articles.

Plain text

Paste text directly. Useful for content that isn't publicly accessible — internal policies, custom FAQ answers, product descriptions.

JSON Schema

Paste a JSON Schema document describing your data model. Hydra converts the schema structure into readable text that the bot can use to answer questions about your platform's objects and fields.

OpenAPI Spec

Paste an OpenAPI (Swagger) specification. Hydra converts it into readable documentation covering your API's endpoints, parameters, and responses.

Adding a source

Go to Knowledge and click Add Source. Select the type, fill in the details, and click Add.

The source enters a Pending state. Once processing begins it moves to Processing, and when complete it becomes Ready. If an error occurs, the status shows Error with a message explaining what went wrong.

Source statuses

Chunk count

Each ready source shows a chunk count — the number of text segments the content was split into for the vector index. More chunks means more detailed coverage, but the overall content quality and relevance matter more than quantity.

Re-ingesting a source

If you update the content at a source URL, delete the existing source and add it again. Hydra will re-scrape and re-embed the latest content.

Best practices

• Be specific — Detailed, accurate content produces better bot answers than vague or generic content
• Cover FAQs explicitly — If you know what customers ask, write those answers directly
• Avoid duplication — Duplicate content can confuse the relevance ranking
• Update regularly — Outdated content leads to outdated answers; keep sources current when your product changes