Knowledge Hub
Upload your product documentation, PRDs, and help articles so ProductBridge's AI knows what you've already built — and can tell new feedback apart from existing features.
Why a Knowledge Hub?
ProductBridge's AI gets dramatically smarter the moment it knows what your product actually does. Without that context, every incoming piece of feedback looks novel — even if you shipped the feature six months ago. With it, the AI can:
- Recognize duplicates of existing features. "Can you add dark mode?" gets flagged against your existing dark-mode docs.
- Refine deduplication accuracy. Semantic dedup uses the Knowledge Hub as a baseline.
- Power Ask AI. Questions about your product return answers grounded in your own docs.
- Better categorize incoming feedback. Knowing which features exist sharpens the AI's mental map.
The Knowledge Hub is an admin-only, internal corpus — it's not customer-facing. Think of it as the briefing document you give the AI.
What to Upload
Anything that describes what your product does, how it works, or what's already shipped:
| Type | Examples |
|---|---|
| Product specs | PRDs, feature briefs, design docs |
| Help articles | Your customer-facing help center content |
| Release notes | Past changelog entries, internal launch notes |
| Onboarding flows | Step-by-step guides for new users |
| API documentation | Public and internal API references |
| Internal wikis | Engineering notes, architecture docs |
You don't need to upload everything at once. Start with the most important 5–10 documents — your top-level feature docs and your help center articles — and add more as you see gaps.
Supported Formats
- PDF (.pdf)
- Word (.docx)
- Plain text (.txt)
- CSV (.csv)
- Markdown (.md)
Each document is automatically parsed, split into searchable chunks, and embedded into ProductBridge's vector database.
Upload a Document
Open Knowledge Hub
Go to Settings → Knowledge Hub and click Upload Document.
Choose a file
Select a file from your computer or drag-and-drop it into the upload area.
Wait for processing
The document moves through these stages:
- Uploaded — file stored, waiting in queue
- Processing — being parsed and chunked
- Completed — fully indexed and queryable
- Failed — something went wrong (the error reason is shown)
Most documents finish in under a minute. Very large PDFs may take a few minutes.
Verify it's working
Open Ask AI and ask a question that should be answered by your document. If the answer references your content, the document is indexed correctly.
Document Status
| Status | Meaning | What to Do |
|---|---|---|
| Uploaded | Queued, not yet processed | Wait — usually quick |
| Processing | Being parsed and chunked | Wait |
| Completed | Indexed and queryable | Nothing — it's working |
| Failed | Couldn't be processed | Check the error; try reprocessing or re-uploading |
If a document fails, you can click Reprocess to try again — useful if it failed due to a transient issue.
How the AI Uses Your Knowledge Hub
When new feedback arrives, the AI runs it against your Knowledge Hub in two ways:
- Existing-feature check — Is this asking for something already documented? If yes, the post is flagged as a likely match and surfaced to admins.
- Categorization context — Knowing what your product does helps the AI tag the right category, suggest the right board, and find the right team owner.
When you query Ask AI, the system performs a hybrid search (semantic + keyword) across your Knowledge Hub and your feedback corpus, then composes an answer grounded in both.
Keep It Updated
The Knowledge Hub is most valuable when it reflects what your product actually does today. A few maintenance habits:
- After every release, upload the relevant release notes or update the existing doc.
- When you sunset a feature, delete or update the corresponding doc so the AI stops referencing it.
- Periodically audit — outdated docs that say "coming soon" for a shipped feature confuse the AI.
If you have a public help center, export and upload those articles as a starting point — they're already written for clarity and cover your top features.
Privacy & Scope
- Documents are scoped to your organization — no other workspace can query your content.
- Documents are not exposed to end users. This is an internal corpus only.
- The same vector store powers deduplication and Ask AI — you upload once, use everywhere.
Don't upload sensitive data you wouldn't want the AI to surface in admin Ask AI answers — internal financials, employee records, etc. Stick to product knowledge.
Last updated 4 days ago
Built with Documentation.AI