Ask natural language questions about your contracts and get answers with citations.
The Q&A feature uses RAG (Retrieval-Augmented Generation) to answer questions about your contracts by:
- Searching vector embeddings for relevant content
- Retrieving relevant document chunks
- Generating answers using LLM with citations
- Providing page numbers and excerpts
graph LR
A[User Question] --> B[Vector Search]
B --> C[Retrieve Chunks]
C --> D[LLM Generation]
D --> E[Answer + Citations]
E --> F[Display Results]
- Navigate to Q&A page
- Click New Conversation
- Enter optional title (auto-generated if empty)
- Conversation is created and ready
All conversations are listed in the sidebar:
- Title: Conversation name
- Last Updated: Most recent activity
- Message Count: Number of Q&A pairs
| Question Type | Example |
|---|---|
| Clause-Specific | "What are the termination terms?" |
| Comparison | "Compare liability caps across documents" |
| Specific Details | "What is the payment schedule?" |
| Risk Assessment | "Are there any high-risk clauses?" |
| General | "Summarize the key terms" |
✅ Good Questions:
- "What are the termination conditions?"
- "What is the liability cap?"
- "Explain the payment terms"
- "What are the IP ownership rights?"
❌ Poor Questions:
- "Tell me everything" (too broad)
- "What's in the contract?" (too vague)
- "Is this good?" (needs context)
Answers include:
- Answer Text: Generated response with markdown formatting
- Citations: Source references with page numbers
- Confidence: Answer confidence score (0-1)
Each citation shows:
- Document Name: Source document
- Page Number: Where information appears
- Section: Document section
- Excerpt: Relevant text excerpt
- Similarity Score: Relevance score
When viewing answers:
- Click on a citation
- PDF viewer opens to the cited page
- Relevant text is highlighted (if coordinates available)
Conversations maintain:
- User Questions: All questions asked
- Assistant Answers: All generated answers
- Citations: All source references
- Timestamps: When each message was created
The RAG pipeline uses conversation history to:
- Provide context for follow-up questions
- Maintain conversation flow
- Reference previous answers
Example:
- Q1: "What is the liability cap?"
- Q2: "Is that standard?" (refers to Q1 answer)
When asking questions, you can:
- Select specific documents
- Search only within selected documents
- Get answers from filtered set
Use Cases:
- Compare specific contracts
- Focus on particular documents
- Exclude irrelevant documents
- Click on conversation title
- Edit title text
- Save changes
- Select conversation
- Click Delete
- Confirm deletion
Warning: Deleting a conversation removes all messages and history.
Generate PDF evidence packs containing:
- All questions and answers
- Supporting citations
- Document excerpts
- Page references
See Evidence Packs for details.
Possible Causes:
- No documents in workspace
- Documents not processed
- Question too vague
- No relevant content in documents
Solutions:
- Verify documents are processed
- Check workspace has documents
- Rephrase question more specifically
- Try different question wording
- Rephrase Question: Be more specific
- Check Citations: Verify source relevance
- Review Documents: Ensure documents contain relevant content
- Try Different Question: Different phrasing may help
- PDF Format: Highlighting works best with PDFs
- Coordinates: May not be available for all chunks
- Browser: Ensure PDF viewer is supported
| Practice | Recommendation |
|---|---|
| Be Specific | Ask focused questions |
| Use Context | Reference previous answers |
| Check Citations | Verify source accuracy |
| Filter Documents | Narrow search scope |
| Export Results | Save important conversations |
- Evidence Packs - Generate PDF evidence packs
- Document Management - Upload more documents
- Clause Extraction - Extract clauses for analysis