QAVE Studio Docs
QAVE Studio Docs
Sign inGet started — the Studio loop
Getting StartedFirst assurance runQuick assure
ConceptsLifecycleGlossaryHow QAVE Studio worksComing soon
Account & workspaceSignup and loginPeopleDemo creditsRolesSettings
Generate simulation suitesContext templatePII and field enhanceDimensionsAgents and plannerDraft review and Save
ExportExport formatsExport formats (detail)
Assurance SDKSDK overviewQAVE Studio suite fetch APIAgent and JSONL queueLangfuse bridge
Execute and tracesRecipe chatbot (reference SUT)TracingFetch tracing
AssureEvaluate APIRubrics and evidenceReading scores
History
WorkbenchGolden data
CoverageCoverage dashboardCoverage strategiesCoverage advisorTest case quality
RAG AssuranceRAG runsRAG datasetsRAG analytics
Project evalsError analysisUnderstanding scores
Studio guidesDimensionsTuplesConversation modesMulti-lingual testingWorkflowsHuman reviewBest practices
ReferenceAPI and packagesTroubleshooting
Studio guides

Conversation modes

Conversation modes

This guide explains how to configure and test inbound, outbound, and mixed conversation modes in QAVE Studio.

Note: This guide replaces the older Bot-Initiated Testing article.


Conversation Modes

QAVE Studio supports three conversation modes:

ModeWho speaks firstUse case
Inbound (user-first)UserStandard customer support chatbot, FAQ bot
Outbound (bot-first)BotAppointment follow-up, proactive outreach, collections
BothEitherSystems that handle both inbound and outbound flows

How to Set Conversation Mode

Conversation mode is configured through the Context Template and the Dimensions system.

1. Set the Bot Greeting Template (Context Template)

In the Context Template form, fill in the Bot Greeting Template field with the opening message your bot uses when it initiates a conversation.

Example:

Hello! I'm calling to follow up on your appointment on [date]. Is this still convenient for you?

Placeholders like [date] are used by the generator to create realistic variations.

2. The "Conversation Initiator" Dimension

QAVE Studio automatically manages a Conversation Initiator dimension with values user and bot. You do not need to create this manually.

  • If your system is always user-first: Approve only the user value, reject bot
  • If your system is always bot-first: Approve only the bot value, reject user
  • If your system handles both: Approve both values

When generating tuples, the Conversation Initiator dimension participates like any other dimension. Tuples with Conversation Initiator=bot will generate bot-first conversations.


How Generation Adapts

User-initiated conversations

User: I'd like to reschedule my appointment.
Bot: Of course! What date works best for you?
User: How about next Tuesday?

Bot-initiated conversations

Bot: Hello! I'm calling about your appointment on March 15th. Are you still planning to attend?
User: Actually, I need to reschedule.
Bot: No problem. What date works for you?

When a tuple has Conversation Initiator=bot, QAVE Studio uses the Bot Greeting Template from the context template as the first message. The generator then creates realistic user responses and subsequent bot handling.


Examples

Appointment Follow-Up Bot (outbound)

Context Template settings:

  • Bot Greeting Template: "Hi, this is a reminder call about your appointment on [date] at [time]. Do you need to reschedule or cancel?"

Dimension setup:

  • Conversation Initiator: approve bot only
  • Intent Type: values — rescheduling, cancellation, confirmation, no-show risk
  • Emotional Tone: values — cooperative, resistant, distracted, confused

This setup produces test cases where the bot always speaks first, covering different user response patterns to the greeting.

Outbound Call Center (collections)

Context Template settings:

  • Bot Greeting Template: "Hello, I'm calling from [company] regarding an outstanding balance on your account. Can I speak to [name]?"

Dimension setup:

  • Conversation Initiator: approve bot only
  • User Response Type: values — cooperative, denial, dispute, request callback

Support Chatbot (inbound)

Context Template settings:

  • Bot Greeting Template: (leave empty or set a welcome banner — not an initiator)

Dimension setup:

  • Conversation Initiator: approve user only
  • Intent Type: values — complaint, inquiry, technical issue, billing question

Multi-mode Systems

For systems that handle both inbound and outbound conversations:

  • Approve both user and bot values on the Conversation Initiator dimension
  • The tuple generator will create tuples covering both conversation styles
  • Review coverage in the Coverage Dashboard — you should see test cases for both initiator types

Tips

  • The Bot Greeting Template is critical for bot-initiated test quality. A specific, realistic opening message produces better test conversations than a generic fallback.
  • Multiple bot types in one project: Use the generation_focus field on the Generate page to guide each generation run (e.g., "Focus on appointment follow-up bot" vs. "Focus on complaint resolution bot").
  • Language + bot-initiated: For non-English languages, add translations of the bot greeting template to the dimension values. QAVE Studio will use translated versions when generating in those languages.

Tuples

Tuples

Multi-lingual testing

Multi-lingual testing

On this page

Conversation ModesHow to Set Conversation Mode1. Set the Bot Greeting Template (Context Template)2. The "Conversation Initiator" DimensionHow Generation AdaptsUser-initiated conversationsBot-initiated conversationsExamplesAppointment Follow-Up Bot (outbound)Outbound Call Center (collections)Support Chatbot (inbound)Multi-mode SystemsTips