Documentation Index
Fetch the complete documentation index at: https://docs.voxworks.ai/llms.txt
Use this file to discover all available pages before exploring further.
What is the Main Flow?
Every Script has exactly one Main Flow. When a call starts:
- The Script is loaded
- The Main Flow is identified via the Script’s
main_flow_id
- The assistant begins execution at step 0 of the Main Flow
- The conversation continues until an end condition is reached
Main Flow Structure
A Main Flow contains an ordered sequence of steps, each step can branch to other steps or sub-flows based on conditions.
Starting the Conversation
The first step of the Main Flow handles the opening message:
Outbound Calls:
- The assistant may wait briefly for the caller to speak
- Example: “Hi, this is Sarah from Voxworks.”
Inbound Calls:
- The assistant speaks first with a greeting
- Example: “Thank you for calling Voxworks. How can I help you today?”
Flow Navigation
From any voice step, conditions determine where to go next:
| Condition Type | Description | Example |
|---|
| Specific match | User response matches a defined scenario | ”User wants to book” → Booking step |
| Otherwise | Fallback when no specific condition matches | Clarification step |
Knowledge Base and Guidelines
When a user asks a question or raises a topic during the conversation, the assistant checks the Knowledge Base and Guidelines for relevant information.
If the assistant has information to respond:
- The assistant answers the question using available knowledge
- The conversation stays on the current step
- The assistant continues with the current step’s objective
This allows the assistant to handle ad-hoc questions without derailing the flow. For example:
Step: "Would you like to schedule a demo?"
User: "How much does it cost?"
→ Assistant answers from Knowledge Base: "We offer flexible pricing
on a cents per minute basis. Would you like to schedule a demo
to discuss your specific needs?"
→ Conversation remains on the same step, waiting for demo response
Ending the Conversation
The Main Flow ends when:
- End condition reached — A step’s condition has
end_flow: success or fail
- User hangs up — The participant disconnects
- Maximum duration — The configured time limit is reached
- Assistant hangup — The assistant determines the call should end
When the Main Flow ends, the call terminates.
Main Flow vs Sub-Flows
| Aspect | Main Flow | Sub-Flow |
|---|
| Entry | Automatic at call start | Invoked by a Flow step |
| Exit | Ends the call | Returns to parent flow |
| Required | One per Script | Optional, unlimited |
| Reusability | Specific to Script | Can be shared across Scripts |
Maximising Reusability
While the Main Flow is specific to each Script, you can maximise reusability by keeping it minimal:
- Opening message — Greet the caller
- Flow step — Route immediately to a reusable sub-flow
- Closing messages — Handle success and failure outcomes from the sub-flow
This pattern lets you reuse the core conversation logic across multiple Scripts, with only the opening and closing messages varying per Script.
Interrupts
During a conversation, users may interrupt the assistant while it is speaking. How interrupts are handled depends on the step type:
| Step Type | Interruptible | Behaviour |
|---|
| Voice | Yes (default) | User can interrupt; assistant stops speaking and processes the interruption |
| Tool | No | User cannot interrupt while the tool executes |
| Function | No | User cannot interrupt while the function executes |
interruptible: false setting when you need the assistant to complete critical information without being cut off.
For more on interrupt handling limitations, see Assistant Behaviour.
Best Practices
- Start with a clear opening — Set expectations immediately
- Keep the main path simple — Use sub-flows for complex branches
- Handle common objections — Include conditions for typical responses
- Plan your exit points — Define clear end conditions
- Test the full journey — Walk through every possible path
Next Steps
