When a Circuit agent responds to your question, the answer includes several elements beyond the main text. Understanding these helps you verify accuracy and get more value from each response.Documentation Index
Fetch the complete documentation index at: https://docs.circuit.ai/llms.txt
Use this file to discover all available pages before exploring further.
Text responses
The main body of the response is a natural-language answer to your question. The agent synthesizes information from multiple documents and presents it in a readable format: paragraphs, lists, tables, or step-by-step procedures depending on what fits best.Citations
Citations are references to the source documents the agent used to construct its answer. They typically appear as numbered references or inline links and include:- Document name: the file the information came from
- Page number or section: where in the document the relevant content appears
Tool usage indicators
As the agent works on your question, you may see indicators showing what it’s doing:- Searching: the agent is querying one or more indexes to find relevant documents
- Reading: the agent is reading through a specific document or section
- Fetching: the agent is retrieving content from a web source (if web tools are enabled)
Images and tables
Agents can include images from your documents (diagrams, charts, photos) and format data as tables when appropriate. If you need a visual from a document, ask directly:Can you show me the wiring diagram from the installation manual?
Multi-part responses
For complex questions, the agent may break its response into sections. If you only need one part, you can ask it to focus:That’s helpful. Can you expand on the third point about temperature limits?
When something looks wrong
If a response seems inaccurate or incomplete:- Check the citations: verify the source documents are current and relevant
- Ask for clarification: “Are you sure about the 500 PSI rating? I thought it was higher”
- Provide feedback: use the thumbs down button and explain what was wrong. This helps improve the system for everyone
- Try rephrasing: give the agent more context about what you expected