Multi-level knowledge tree
Entries organized by product line / topic, with multi-format import and hierarchy.
Platform Knowledge Hub
AI gets specs wrong when it reads mixed-up sources
An isolated knowledge base and context per project: product materials, corporate knowledge and authoritative evidence layered and injected precisely at generation time — no more parameter mix-ups.
What it solves
- AI attaches product A's specs to product B — the multi-product brand's nightmare.
- The root cause is context: every product shares one pool of material, so AI reads across the lines and answers wrong.
- The fix is simpler than it sounds: separate the material, and AI says only what it was given.
How it works
- Multi-level knowledge tree: organize entries by product line / topic, import in many formats.
- Retrieval rework: global recall at generation time, injecting only relevant knowledge.
- Corporate knowledge and evidence: credentials, cases and data sources on record.
- Automatic llms.txt: declare authoritative on-site content in an AI-crawler-friendly way.
- Public knowledge portal (optional): password-shared knowledge pages and Q&A.
Capabilities
Six capabilities of the knowledge hub
Hybrid retrieval (RAG)
Vector + keyword + title recall with reranking — generation receives only what is truly relevant.
Evidence & provenance
Every entry traces back to its source and revisions; credentials, cases and data sources are layered.
Knowledge review flow
In strict mode knowledge changes require approval before generation can use them (Enterprise and above).
Evaluation sets & expiry detection
Fixed Q&A pairs test retrieval quality; stale content is flagged so AI stops quoting it.
llms.txt & public portal
llms.txt generated automatically to declare authoritative content; optional password-protected public knowledge portals with Q&A.