AssemblyAI - Apply LLM to transcript action
The Apply LLM to transcript action calls AssemblyAI's LLM Gateway to run a chat completion, optionally injecting a transcript's text into the prompt. Use this action for summaries, action items, or other free-form LLM tasks over a transcript.
INJECT THE TRANSCRIPT WITH A PLACEHOLDER
Enter a Transcript ID and include the literal tag {{ transcript }} (exact spacing) in Prompt or Messages. AssemblyAI replaces the tag with the transcript's text before sending the request to the model.
Input
The following are a representative set of input fields. Refer to the AssemblyAI Chat completions documentation for more information on available input fields.
| Input field | Description |
|---|---|
| Model | Select an LLM Gateway model, or enter a custom model ID. |
| Transcript ID | Optional. Enter a transcript's UUID to make it available to the {{ transcript }} placeholder. |
| Prompt | Enter a simple text prompt. Use this or Messages. |
| Messages | Enter OpenAI-style messages, each with a Role and Content. Use this or Prompt. |
| Max tokens | Optional. Enter the maximum number of tokens to generate. Default 1000. |
| Temperature | Optional. Enter the sampling temperature. Valid values are 0-2. |
| Stream | Optional. Select Yes or No to stream SSE responses. Only for OpenAI models. Leave off for standard JSON responses in recipes. |
| Model region | Optional. Enter the model region to opt into global routing. The only accepted value is global. |
This action also supports function calling (Tools, Tool choice, Tool choice function), structured JSON output (Response format), model fallbacks (Fallbacks, Fallback config), and Post processing steps. Refer to the AssemblyAI Chat completions documentation for more information on these fields.
Output
The following are a representative set of output fields. Refer to the AssemblyAI Chat completions documentation for more information on available output fields.
| Output field | Description |
|---|---|
| Request ID | The LLM Gateway request's unique identifier. |
| Choices | The completion choices. |
| Message (Choices) | The message choices. |
| Role (Message) | The message author's role, for example assistant. |
| Content (Message) | The generated message text. |
| Finish reason (Choices) | The reason the model stopped generating, for example stop or length. |
| Http status code | The LLM Gateway HTTP response status code. |
| Response time | The request's round-trip time, in milliseconds. |
This action's response also includes an echo of the request (Request) and token usage (Usage). Refer to the AssemblyAI Chat completions documentation for more information on these fields.
Last updated: