Live Meeting Assistant (LMA) summarizes meeting transcripts once the meeting is over.
You can configure Transcript Summarization by choosing a value for the EndOfCallTranscriptSummary CloudFormation parameter when deploying or updating your LMA stack. Valid values are
BEDROCK (default) and LAMBDA.
If the BEDROCK option is chosen, select a supported model ID from the list (BedrockModelId parameter)
The BEDROCK option is enabled by default. You must request model access for the model selected in the BedrockModelId parameter. By default, the selected model is anthropic.claude-3-haiku-20240307-v1:0.
LMA can run multiple LLM inferences after the call is complete. The prompt templates used to generate the insights from the transcript are stored in a DynamoDB table. There are two items (records) in the table:
-
Default prompt templates: These come with the LMA release, and define the summaries you get if you do not create custom prompt templates. Default prompts may change in new versions of LMA. View the default prompts by opening the DynamoDB URL in the LMA Stack output
LLMDefaultPromptSummaryTemplate.
-
Custom prompt templates: Initially after deploying LMA, there are no custom prompts defined, but you can add your own. Create custom prompt templates to override or disable default summary prompts, or to add new ones. Custom prompt templates are not overwritten when you update your LMA stack to a new version. View and edit the custom prompts by opening the DynamoDB URL in the LMA Stack output
LLMCustomPromptSummaryTemplate.
The attribute named **Information** is ignored by LMA - it is informational only.
All other attributes define the summary prompts that LMA executes when the meeting is over. Each has an attribute name used as the heading that shows up in the LMA summary, and an attribute value which defines the prompt template used to invoke the selected LLM.
Attribute Name: The attribute name must be formatted using a sequence number N, a # symbol, and the heading you want to use for the summary in the UI. For example, 1#Summary defines a heading value of Summary, that will always be displayed above other headings with a higher sequence number. LMA removes the sequence number before displaying the title.
Attribute Value: The attribute value is used as the prompt template. LMA replaces <br> tags with newlines. Use the template variable {transcript} to indicate where the meeting transcript will be placed in the prompt. LMA replaces {transcript} with the actual meeting transcript in the form:
<SpeakerName>: <transcription text>
<SpeakerName>: <transcription text>
...
LMA invokes the Bedrock model using your prompt, and renders the results in the Summary section of the Meeting detail page, under your heading.
Add a new custom summary: Edit the Custom prompt templated by opening the DynamoDB URL in the LMA Stack output LLMCustomPromptSummaryTemplate. Initially this item has no prompt templates defined, but you can choose Add new Attribute to define your own prompt. Use the attribute type String. Use the Attribute Name format described above, e.g. #4My Custom Meeting Insights, and use Attribute Value to define the prompt template as described above.
Customize default prompts: To override a default prompt value, provide an alternative prompt template for the same Attribute Name (heading) in the Custom prompt templates. LMA merges the default and custom prompts at runtime, and when both contain the same Attribute Name, the prompt template from the Custom prompt templates table is used instead of the default. Do not edit the Default prompt templates item directly, as it may be overwritten during future stack updates.
Remove default prompts: Create a custom prompt template attribute with the same name as the default prompt you want to disable, but leave the attribute value empty, or give it the value 'NONE'. When LMA merges the default and custom values, the empty (or 'NONE') valued prompts are skipped.
Use the LAMBDA option to provide your own summarization functions and/or machine learning models. This option allows you to experiment with different models and techniques to customize the summary as you need.
When you choose the LAMBDA option, you must provide the Arn of your custom Lambda function in the EndOfCallLambdaHookFunctionArn CloudFormation parameter. At the end of a call, the CallEventProcessor Lambda function will invoke the custom Lambda function and pass in the CallId of the call.
Your custom Lambda function must return the summary in the following JSON format:
{
"summary": "Summary of the call here."
}
The summary can optionally use Markdown syntax to include rich text, hyperlinks, images, media, etc.
If you would like to include more than one section in the summary, similar to the BEDROCK option, you may provide a JSON-encoded string that contains key value pairs as the summary's value. This will render each key as a section header, and the value as the body of the section. The following is an example:
{
"summary": "{\n\"Summary\":\"The weekly report was reviewed.\",\n\"Topic\":\"Weekly Report\",\n\"Follow-Up Actions\": \"John will set up a follow-up meeting.\"\n}"
}
This would be rendered as:
Here is a minimal example of a valid custom summarization Lambda function, written in Python.
import json
def lambda_handler(event, context):
print(json.dumps(event))
summary = "Placeholder for actual summary"
# call your own model, APIs, or summarization functions here...
return {
"summary": summary
}
This example function trivially returns a hardcoded string as the summary. Your function will be much smarter, and will implement your custom rules or models.
Use the provided FetchTranscript utility Lambda function in your custom summarization Lambda to retrieve the call transcript, and optionally truncated to the maximum input token limit imposed by your summarization model.
If your custom Lambda fails at runtime, or you do not want to return a summary for the call, return an empty string for the value of the summary field.