Post-conversation
Resources
The resources array in an Exchange records the configuration used for the call, runtime settings, identifiers, and media references. Use it to inspect the call's settings even if the Assistant has since changed.
Resource structure
Each resource is a key-value pair with a type indicator:
| Field | Type | Description |
|---|---|---|
key | string | Identifier for the resource (supports dot notation for nested values) |
value | any | The resource value |
type | string | Data type, such as "string", "number", "boolean", "array", "object", or "url" |
Resource categories
Assistant configuration
These values record the initial Assistant's settings at the time of the call.
| Key | Type | Description |
|---|---|---|
assistant_id | string | UUID of the Assistant |
assistant_config.llm | string | The LLM model used (e.g., "mistral-medium-3-5") |
assistant_config.temperature | number | LLM temperature setting |
assistant_config.stt_model | string | Speech-to-text model used |
assistant_config.tts_model | string | Text-to-speech model used |
assistant_config.tts_voice | string | TTS voice identifier |
assistant_config.language | string | Language code (e.g., "en-US") |
assistant_config.timezone | string | Timezone (e.g., "Europe/Paris") |
Dynamic configuration
These values record runtime settings for the call. They may differ from the Assistant's defaults, depending on how the conversation started.
| Key | Type | Description |
|---|---|---|
dynamic_config.assistant_id | string | UUID of the Assistant (same as assistant_id) |
dynamic_config.connection_type | string | Connection method: "webrtc" or "sip" |
dynamic_config.timezone | string | Timezone used for this conversation |
dynamic_config.language | string | Language used for this conversation |
dynamic_config.first_message | string | The Assistant's greeting message |
Media resources
References to media files generated during the conversation.
| Key | Type | Description |
|---|---|---|
audioUrl | url | URL to the conversation audio recording |
Example: technical support conversation
This excerpt from the Post-conversation example records the Assistant's models and voice, runtime settings, and audio recording URL.
Code
Configuration breakdown
- LLM:
mistral-medium-3-5with temperature0.3 - Speech-to-Text:
voxtral-mini-transcribe-realtime-2602for transcribing user speech - Text-to-Speech:
chirp-3with voiceen-US-Chirp3-HD-Orus - Language: English (US) with Europe/Paris timezone
- Connection: WebRTC (browser-based), not a SIP phone call
- First Message: The greeting shown at the start of the Messages transcript
Usage notes
- The dot notation in keys (e.g.,
assistant_config.llm) represents logical grouping, not actual nested objects - Resources record the call's configuration and identifiers. Additional keys include
conversation_id,room_id,channel_id,conversation_ended_by, andassistant_id.1,assistant_id.2, etc. for the Assistants entered during the call. SIP calls can also includesip.CallID. - The
assistant_config.*values describe the initial Assistant. Runtime variables marked as secret are omitted fromdynamic_config.*resources. - Configuration values here may differ from the current Assistant configuration if the Assistant has been modified since the conversation
What's next
Last modified on