Trust
Build a Source-Backed AI Knowledge Base for Agents
Build a source-backed AI knowledge base by connecting trusted sources, atomic knowledge, maintained LLM-wiki pages, citations, review, and refresh state.
About this guide
Concepts
Teams building a local AI knowledge base that agents and people can inspect
5 min read
01
Trusted sources remain separate from the knowledge derived from them.
02
Atomic knowledge preserves precise evidence before it becomes a page.
03
Maintained pages carry citations, review, revisions, and refresh state.
01
Quick answer
To build a source-backed AI knowledge base, keep trusted sources, atomic knowledge, and maintained LLM-wiki pages separate. Connect them with citations, review, revisions, and refresh state so agents retrieve current answers and people can inspect why those answers are trusted.
Wenlan separates Sources, Memories, and Pages. Pages retain source links, revisions, stale reasons, and review state so you can inspect what changed after a rebuild. Those records do not establish that the model preserved every number or exposed every source conflict.
02
When this problem appears
Raw document dumps and one-off summaries both decay. A dump makes agents search too much context; a detached summary can become stale without showing which source supported it or what should refresh it.
03
Build the smallest maintainable loop
Prove one topic end to end before importing a large archive.
- Register or scope the trusted source that should support the topic.
- Capture one atomic fact with its source and why it matters.
- Distill only after a topic repeats or needs a reusable answer.
- Open the page and inspect the source IDs behind its claims.
- Run lint to find thin, stale, conflicting, or unsupported knowledge.
- Curate revisions and refresh the page when its evidence changes.
Wenlan knowledge-base loop
/brief <topic>
/capture <fact + source + why>
/distill <topic>
/pages <topic>
/lint
/curate04
What to check next
A maintained page is still a claim, not ground truth. Verify important conclusions against the cited source, and mark the page stale when the source or operating context changes.
Try the local memory loop
Install Wenlan, connect your AI client, and verify that capture, recall, and handoff work on your machine.
Inspect a citation and a proposed revision in Wenlan
These are real Wenlan app recordings using Tally, a demonstration invoicing project. Enlarge the screens to follow a source reference and inspect a proposed page revision.
1. Follow the SQLite claim back to its source
The wiki says Tally uses SQLite for a single-user app. The open citation shows the source titled ‘SQLite for Single-User App’, its summary, and an ‘Open memory’ action. Compare the sentence with that source before reusing the decision.
2. Read the proposed change before approving it
The Project notes revision removes the undecided database wording and proposes a SQLite decision. Added and removed text, an earlier version, and Skip / Approve controls are visible. Check the revised wording and its sources before choosing what to keep.
FAQ