Using the Chat Editor
The Chat Editor is the recommended way to create and modify the prompts (Instructions, Rubric, Directions, Definitions) for your Analysis Field Drafts. It provides an interactive, AI-assisted environment to define and iterate on your analysis logic.Accessing the Chat Editor
- Create a Draft for the field you want to edit.
- When prompted, choose the “Edit Field with Chat” option.
- Alternatively, navigate to an existing “In Progress” draft’s page (
/settings/analysis/drafts/[id]) and click the “Chat Editor” button.
The Chat Editor Interface
The Chat Editor typically presents a split view:- Prompt Components (Left/Top): Displays the current draft content for Purpose, Instructions/Rubric, Directions, Definitions, etc., often using rich text editors.
- AI Chat Interface (Right/Bottom): A conversational interface where you interact with the AI assistant.
Interacting with the AI
Think of the AI as your knowledgeable co-author. You can ask it to:- Generate Initial Prompts: “Based on the Purpose I provided for this field, can you generate draft Instructions and Directions?”
- Refine Existing Prompts: “Let’s improve the Rubric for the 6/10 score. Add a criterion about mentioning budget, even if vague.”
- Incorporate Examples: “Here’s an example of good output: […]. Update the Instructions to reflect this.” or “Here’s a transcript snippet where the rep handled the objection well […]. Add this as a positive example in the Directions.”
- Format Output: “Please ensure the Instructions specify the output should be a Markdown numbered list.”
- Suggest Improvements: “Review the current Instructions. Are they clear? Can you suggest any improvements for clarity or to better achieve the field’s Purpose?”
- Address Test Failures: “I ran a test, and the draft failed to extract the competitor ‘Acme Corp’ from this call [provide context if needed, or add the call as a source]. How should we update the Instructions or Directions?”
Adding Context for the AI
To help the AI understand your specific business context and test cases, you can add sources directly within the Chat Editor:- Look for the “Add Sources” button near the chat input area.
- A drawer appears with tabs for:
- Calls: Search and select relevant call transcripts.
- Documents: Search and select relevant internal documents (like CRM notes, if applicable).
- Knowledge Documents: Select relevant playbooks, product docs, etc., from your Fabius knowledge base.
- Selected items appear as badges showing what’s included in your analysis context.
- Reference these sources in your chat: “Compare the draft’s output on the selected call ‘Discovery - Acme Q3’ to the live field’s output.” or “Based on the linked ‘Objection Handling Playbook’ knowledge document, refine the Directions for this score.”
The AI uses the content of these selected sources to understand examples,
test cases, and your internal standards when generating or refining prompts.
Reviewing AI Suggestions (Diff View)
When the AI proposes changes to your prompts:- Updated content appears in the relevant editor with a Diff View showing exact changes (additions in green, deletions in red).
- Review Options:
- Individual Review: Accept or reject changes for each component separately (Instructions, Directions, etc.).
- Bulk Review: Use the “Accept All Changes” button to approve all proposed changes at once.
- Accept: Applies the AI’s suggested changes to the editor.
- Reject: Discards the AI’s suggestions, keeping the previous version.
Applying Changes and Triggering Tests
- Changes you Accept are saved to the current draft version.
- When you have tests configured and accept all changes (either individually or using “Accept All Changes”), tests automatically re-run.
- Test results flow directly into your chat conversation, enabling immediate discussion with the AI about any issues.