memory

Knowledge sources

Distinguish saved memory from retrieved sources, and understand how OpenbirdAI decides which evidence a question actually needs.

· OpenbirdAI documentation

Memory is context you chose to keep. A knowledge source is evidence OpenbirdAI fetched for the current run. Mixing the two up is why people sometimes assume OpenFeather knows something it has simply never been shown.

This guide describes a capability that depends on the connected provider, the permissions granted to it, and your current plan. If the provider cannot supply the required data, OpenbirdAI reports it as unavailable rather than estimating.

Memory versus source

  • Memory — pre-existing context you or your team stored. It shapes how an answer is framed.
  • Source — retrieved evidence for this specific request, cited and re-verified each run.
  • Generated interpretation — OpenFeather's explanation built from one or both of the above.

A statement can be shaped by memory and still need a source to be credible. When they conflict, the source wins for factual claims and memory wins for framing and preference.

How source selection works

The request planner decides which sources are required, retrieval runs only for those, and results are merged and ranked before synthesis. Keyword matching stays important for exact names, error codes, setting names and integration names, so a precise configuration match is not overridden by semantic similarity.

The documentation base is one of these sources. A question about how OpenbirdAI works is answered from OpenbirdAI's own documentation rather than the public web.

Keeping sources trustworthy

  • Prefer primary sources: your own records, the provider's own documentation, official announcements.
  • Check freshness before relying on a dated figure.
  • Treat a source you cannot open as unverified, even if the answer quoted it.
  • Report documentation that contradicts current behaviour so it can be corrected.

When nothing relevant exists

If no relevant source exists, OpenFeather states that it could not verify the claim from available sources. That is the correct behaviour. Ask for the missing connection or the missing source rather than pressing for a confident guess.

OpenbirdAI documentation as a knowledge source

Product questions are answered from OpenbirdAI's own documentation rather than from a web search. The same articles you are reading here are what OpenFeather retrieves, which is why an answer about how to connect a service should never contradict this site.

If an answer about the product does contradict the published guide, treat the verified current behaviour as correct and report the documentation gap so it can be fixed.

Documentation retrieval is preferred over web search for product questions because it is versioned with the product and cannot be outdated by a third party.

Common questions

How do I make OpenbirdAI use a specific document?

Store it in a connected service and reference it by name in the question, so the matching service and record are targeted.

Does OpenFeather search the whole documentation set every time?

No. It retrieves only the sections relevant to the request, within a bounded context budget.

Was this page helpful?