Skip to main content

Updated Oct 31, 2026Verified against the product Oct 31, 2026

Product docs model

The product docs model brings your product documentation in from a data connection — a data warehouse or database — one row per document. It's the warehouse alternative to the Documentation site and Uploaded documents connections: each row becomes a searchable document that FunnelStory's AI can read and cite. Reach for it when your documentation already lives in a table you can query.

However the docs arrive — crawled, uploaded, or mapped in here — they land in the same place, so agents, flows, and MCP search them together. See Put Your Product Docs to Work for how they're used.

Field reference​

Model key​

PropertyTypeDescription
external_idstringA stable identifier for the document in your source; required. Rows with an empty external_id are skipped. FunnelStory namespaces it per model, so two models can map the same source IDs without colliding.

Default properties​

PropertyTypeDescription
titlestringThe document's title.
contentstringThe full document text, as Markdown — stored whole, with no chunking. A row with empty content is dropped on refresh, so in practice content is required.
urlstringWhere the document lives, so the AI can cite it.
product_idstringOptional. Ties the document to one product; leave it unset when the docs aren't product-specific (most aren't).

Any extra columns you map that aren't one of these are kept on the document as well.

Configure​

  1. Configuration → Data Models → Add New Model, then choose Product Docs.
  2. Choose a data connection — your warehouse or database.
  3. Provide a query (or pick a table) that returns one row per document.
  4. Map your columns to external_id (required), title, content, url, and optionally product_id.
  5. Set a refresh interval (default six hours; your workspace may differ), then Validate, Save, and Refresh.

Each refresh upserts the current rows and removes documents that have left the source, so the docs stay in step with your warehouse. (There's no AI-suggested setup for this model type — you map it yourself.)