Knowledge Base

The knowledge base is the source of truth for AI deflection. When a question comes in, AnswerLoops searches the KB semantically to find the closest matching content, then uses that content to draft a reply. The better your KB, the more accurate and confident your AI answers.

Import methods

There are three ways to add content to the knowledge base.

1. URL crawl (Firecrawl)

Import a web page or an entire documentation site.

1. Go to Knowledge Base → Import
2. Paste a URL into the import field
3. Choose Single page or Entire site
   • Single page imports only that URL
   • Entire site crawls linked pages up to a maximum of 25 pages
4. Click Import — a progress spinner shows while Firecrawl runs (typically 15–60 seconds)
5. Imported pages appear in the Sources list when complete

Each page becomes a separate source entry. The crawler fetches the rendered content, splits it into chunks, embeds each chunk as a vector, and writes everything to the database.

Re-import deduplication: if you import the same URL a second time, pages that were already ingested are skipped automatically. Only new pages are processed. This means you can safely re-run a site import after adding new docs without creating duplicates.

2. File upload

Drag and drop files directly into the KB.

Supported formats: PDF, DOCX, MD, TXT, CSV

Maximum file size: 50 MB per file

1. Go to Knowledge Base → Import
2. Drag files onto the upload area, or click to open the file picker
3. Files are processed immediately — each becomes one or more source entries depending on size

Large files are split into chunks automatically. Each chunk is embedded and indexed for semantic search.

3. GitHub repo sync

Connect a GitHub repository and sync its markdown files into the KB.

1. Go to Knowledge Base → GitHub Sync
2. Select the repository from the picker (requires GitHub App installed on your org)
3. Click Sync now to run an immediate import of all .md and .mdx files
4. Optionally enable Auto-sync on push — AnswerLoops receives a webhook from GitHub on every push to the default branch and re-syncs changed files automatically

The sync respects the repo's file tree. Each markdown file becomes its own source entry.

Sources list

The Sources tab on the Knowledge Base page lists every imported source. Each row shows:

To remove a source, click the Delete icon on its row. Deleting a source removes all its chunks from the vector index — any articles derived from it will no longer surface in search results.

Promoting tickets to the KB

When a ticket is resolved with a good answer, you can turn it into a KB article so the AI can reuse that answer for similar future questions.

1. Open a ticket with status resolved
2. Click Promote to KB in the ticket actions panel
3. The Q&A pair is saved as a new KB article

Promoted articles appear in the Sources list with type ticket.

Search

Use the search bar at the top of the Knowledge Base page to find articles.

Semantic search (default when an AI provider is configured): searches by meaning rather than exact words. Asking "how do I reset my password" will surface an article titled "Account recovery steps" even though the words don't match exactly.

Keyword fallback (when no AI key is configured): falls back to exact text matching. An amber warning banner appears at the top of the KB page to indicate that semantic search is unavailable.

To clear a search and return to the full article list, click the × inside the search field or the Clear button.

Data persistence

KB articles and source chunks are stored in the kb_articles and kb_sources tables in PostgreSQL. They persist across Docker restarts as long as you retain the postgres-data volume.