Knowledge base
The co-pilot is designed to stay grounded in your content. The knowledge base is the indexed collection of documentation, files, and web pages that Pillar searches when a user asks a question.
When a user asks something, Pillar retrieves the most relevant sections from your indexed content and passes them to the co-pilot as context. The co-pilot then generates a response grounded in that retrieved content when relevant matches are available.
This is retrieval-augmented generation (RAG) — a standard pattern for grounding LLM responses in source material.
Content sources
| Source | Description |
|---|---|
| Website | Automatically crawl and index a docs site, help center, or marketing page |
| Files | Upload PDFs, Word documents, and text files directly |
| Cloud storage | Sync documents from an Amazon S3 or Google Cloud Storage bucket |
| Snippets | Write custom text for corrections, FAQ answers, or topic-specific AI instructions |
All sources are searched together at query time. The co-pilot finds the best matches regardless of source type.
What users see
Users don't interact with the knowledge base directly. They ask questions in the co-pilot and get answers. Behind the scenes, the co-pilot searches your knowledge base and uses the results to respond.
Progress indicators show when the co-pilot is searching, so users know their question is being processed.
Feedback loop
When users rate responses with thumbs up or down, that feedback is captured in your dashboard. Use it to identify gaps — if users keep asking about a topic and getting poor answers, you probably need to add content for it.
You can also create corrections from the dashboard. When the co-pilot gets something wrong, you flag the conversation and write the correct answer. The correction gets indexed and prevents the same mistake from repeating.
Next steps
See the Knowledge Base section for setup guides on each source type.