Ingesting Knowledge

Ways to Add Artifacts

• Upload files (PDF, TXT, MD) directly to the project.
• Paste a public URL to ingest web pages or hosted documents.
• Paste a Google Docs URL to ingest a document from your connected Google account (see below).
• Use the Axon Capture Chrome extension to capture screenshots and markdown from any webpage.
• Every artifact is tied to the active project and indexed for summaries and findings.

Ingesting Google Docs

If you have connected a Google account in Settings, you can ingest any Google Doc that account has access to directly from the URL ingestion tab. Paste the document URL and Axon will fetch and analyze it using your Google credentials — no manual export needed.

• Metadata from Google: The document's last-modified date is used as the artifact date, and the document owner is used as the source — both pulled directly from Google Drive rather than extracted from the body text.
• Document tabs: For Google Docs with multiple tabs, all tabs are combined into a single markdown document. Each tab's title becomes a top-level heading (H1), and all content from all tabs is preserved with links and formatting maintained.
• Re-sync: Because the original URL is saved as the artifact's reference, you can use the Regenerate → Source row action to re-fetch the latest version of the document at any time.
• Access errors: If the connected Google account does not have access to the document, Axon will surface a clear error. Check that the document is shared with the account you connected in Settings.
• Circular dependency prevention: Axon will not allow you to ingest a Google Doc that was generated by Axon for the same project (such as a synced artifact, live doc, or project overview), as this would create a feedback loop in the analysis.

Processing Flow

1. Axon stores the raw file or URL reference and creates an artifact record.
2. An edge function extracts text, converts it to markdown, and prepares it for analysis.
3. The summary prompt generates a structured summary.
4. The findings prompt extracts structured findings with optional line references.
5. Results are saved back to the artifact and finding tables for browsing.

Content Requirements

To ensure meaningful AI analysis and avoid wasted processing costs, Axon requires a minimum content length for all ingested artifacts.

• Minimum length: 1,000 characters (approximately 150-200 words) of extracted text content.
• Applies to: All content types including web pages, Google Docs, and uploaded text files.
• Exception: PDFs are exempt from this check as text extraction happens during AI processing.
• Purpose: Documents below this threshold are unlikely to produce useful summaries or findings, and the validation helps catch extraction failures early.

Operational Features

• Processing status updates automatically and supports retry for stalled summary generation.
• Errors can be expanded and copied for faster debugging or support handoff.
• Optional table columns include Added, Updated, and Google Drive link.
• Artifacts and individual rows can be synced to Google Drive when Drive is connected.