Ask A Question

Home Screen

After login, the default view shows:

• Playground (center): Query input box
• Agent selector (top): Dropdown to choose agent
• Navigation (left): Data, Agents, Inbox, Admin, Settings
• Query history (right sidebar): Recent questions

Query an Agent

Location: Playground (home screen or Playground menu item)

1. Select agent from dropdown (required)
2. Type question in input box
3. Press Enter or click Send

Query format: Natural language, 5-200 words

Examples:

• "How do I reset a user password?"
• "What are the pricing tiers for enterprise customers?"
• "Explain the API rate limits"

Expected result: Response appears in 2-5 seconds with citations

Read Response

Response components:

• Question (top): Your query, verbatim
• Answer (main area): Paragraph response, may include code blocks
• Citations (below answer): Numbered sources [1], [2], etc.
• Sources panel (right): Shows chunks retrieved with similarity scores

Citations Format

Answer text mentioning refund policy [1] and shipping [2].

Sources:
[1] Refund Policy (similarity: 0.89) - docs.example.com/refunds
[2] Shipping Guide (similarity: 0.85) - help.example.com/shipping

Citation actions:

• 🔗 Link icon: Opens source document in new tab (URL from ingestion)
• 👁️ Eye icon: Shows chunk text in modal (no navigation)

How to verify: Click link icon → check if citation matches claim in response

Response Actions

Available buttons (below response):

• 📋 Copy: Copies response text to clipboard (includes citations)
• 👍 Accept + Copy (Train role+): Copies + marks accurate (logged for analytics)
• 📝 Edit (Train role+): Opens edit panel (see below)

Role visibility:

• ReadOnly users: See Copy only
• Train/Configure/Admin: See all buttons

Edit Responses

Requires: Train, Configure, or Admin role

Click 📝 Edit button → Right panel opens

Edit Panel Features

Text editor (top):

• Edit response text directly
• Markdown supported (bold, italic, lists, code blocks)
• Click Copy to save edits to clipboard

Citations (middle):

• Checkboxes for each source
• Uncheck sources → Click Regenerate → New response without those sources
• Regeneration uses same agent config, different context

Style presets (buttons):

• Shorten: Reduce response to 1-2 paragraphs
• Lengthen: Expand with more detail
• Bulletize: Convert to bullet points
• Make Empathetic: Add empathetic language
• Make Casual: Conversational tone

Each button triggers AI rewrite (~2-3 seconds)

Custom rephrase (text input at bottom):

• Enter instruction: "Respond in a single sentence"
• Or: "Add code examples"
• Or: "Remove technical jargon"
• Click Rephrase

Expected result: Response regenerates with new style, same factual content

Switch Agents

Location: Agent selector dropdown (top of Playground)

Click dropdown → List of available agents appears

Agent differences:

• Different system prompts (Support agent: detailed steps, Sales agent: concise)
• Different data sources (Support: all docs, Sales: only pricing/features)
• Different RAG strategies (Support: Redwood for speed, Sales: Cypress for accuracy)

How to test differences:

1. Ask same question with Agent A
2. Note response length and tone
3. Switch to Agent B (dropdown)
4. Ask same question
5. Compare responses

Common Mistakes

Symptom: Response says "I don't have information about that"

Cause: No relevant chunks retrieved (all below similarity threshold 0.7)

Diagnostic steps:

1. Check Sources panel → if empty, retrieval failed
2. Verify question matches content in data sources
3. Try more specific query with exact terminology from docs

Fix: Rephrase query or add relevant data source

---

Symptom: Citations don't match response claims

Cause: LLM hallucinated despite context, or citations parsed incorrectly

Diagnostic steps:

1. Click 👁️ icon on each citation
2. Check if chunk text supports the claim
3. If mismatch, mark response inaccurate (👎 button if Train role)

Fix: Click Edit → Regenerate → Check citations again. If persists, contact support with interaction ID.

When This Doesn't Apply

This guide covers Playground UI. For API queries, see REST API Reference.

---

Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the ask query parameter:

GET /dev/getting-started/ask-a-question.md?ask=<question>

The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.