Skip to main content

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.

​
Overview

Call-Start Data Sourcing fetches live data from an external system at the moment a call starts and injects it into your voice assistant’s context window. This ensures the assistant always has current information — not data that may have gone stale since the call was scheduled. This works for both inbound and outbound calls. The same mechanism runs regardless of how the call was initiated.

​
Why Fetch at Call Start?

  • Inbound calls have no schedule time. The phone rings and the assistant picks up. There’s no advance window to pre-fetch data.
  • Outbound calls risk stale data. A call scheduled for 2:00 PM might reference CRM data that changed at 1:55 PM.
  • One flow for everything. A single data-fetching mechanism works for all call types, eliminating divergent code paths.

​
How It Works

Call starts (inbound or outbound)
  -> Voxworks looks up the contact and their linked objects
  -> Finds the automation with call-start config for this Script
  -> Fetches data from the configured source (CRM API or HTTP endpoint)
  -> Maps the response fields to context keys
  -> Validates required fields
  -> Injects the data into the assistant's context window
  -> Call proceeds
The entire process completes within a configurable timeout (default: 5 seconds). The assistant waits for the data before beginning the conversation.

​
Configuring Call-Start Data

Call-Start Data Sourcing is configured on the Call-Start Data tab of the automation editor.

​
Source

Choose how to fetch the data: Integration Fetch — Use a connected integration (e.g., Monday.com) to fetch data. The system automatically resolves the CRM identifier from the contact’s linked objects. HTTP — Fetch from any API endpoint. The URL, method, headers, and body support {{variable}} interpolation using the contact’s data and linked objects.

​
Result Mapping

Map fields from the API response to context keys that the assistant will receive:
Source FieldContext Key
source.notescustomer_notes
source.deal_stagedeal_stage
source.prioritypriority
These keys become available in the assistant’s context window during the call.

​
Required Fields

Optionally mark fields as required. If a required field is missing or malformed after the data fetch, the call is blocked:
  • Inbound calls — The call rings out (does not connect)
  • Outbound calls — The call fails with reason precall_validation_failed
Required field validation supports format checks: email, e164 (phone number), uuid, or a custom regex pattern. This is useful when your assistant cannot function without certain CRM data — for example, a customer account ID that’s needed to look up order history.

​
Timeout

The default timeout is 5 seconds. You can adjust this per automation. If the source fetch fails or times out, the call proceeds without the enrichment data (best-effort), unless required fields are configured.

​
One Automation Per Script

Only one automation can provide call-start data sourcing per Script. This is enforced by the system — if you try to configure call-start data on a second automation for the same Script, you’ll receive an error. If a call was created by an automation (e.g., via a Create Call step), that automation’s call-start config takes priority over any other.

​
Failure Behavior

ScenarioBehavior
No call-start config for this ScriptCall proceeds normally, no data fetched
API fetch fails or times outCall proceeds without enrichment (best-effort)
Required field validation failsCall is blocked — inbound rings out, outbound fails
Internal errorCall proceeds without enrichment
The assistant can always handle a call without enrichment data. Call-start sourcing adds context but should not be the only way the assistant can function — design your Scripts to gracefully handle missing data fields.