Change model and parameters

Change the model type

LeMUR features the following LLMs:

Claude 3.5 Sonnet
Claude 3 Opus
Claude 3 Haiku
Claude 3 Sonnet
Claude 2.1 (Legacy - sunsetting on 02/06/25)
Claude 2 (Legacy - sunsetting on 02/06/25)

You can switch the model by specifying the final_model parameter.

Model	SDK Parameter	Description
Claude 3.5 Sonnet	`aai.LemurModel.claude3_5_sonnet`	Claude 3.5 Sonnet is the most intelligent model to date, outperforming Claude 3 Opus on a wide range of evaluations, with the speed and cost of Claude 3 Sonnet. This uses Anthropic’s Claude 3.5 Sonnet model version`claude-3-5-sonnet-20240620`.
Claude 3.0 Opus	`aai.LemurModel.claude3_opus`	Claude 3 Opus is good at handling complex analysis, longer tasks with many steps, and higher-order math and coding tasks.
Claude 3.0 Haiku	`aai.LemurModel.claude3_haiku`	Claude 3 Haiku is the fastest model that can execute lightweight actions.
Claude 3.0 Sonnet	`aai.LemurModel.claude3_sonnet`	Claude 3 Sonnet is a legacy model with a balanced combination of performance and speed for efficient, high-throughput tasks.
Claude 2.1(Legacy - sunsetting on 02/06/25)	`aai.LemurModel.claude2_1`	Claude 2.1 is a legacy model similar to Claude 2. The key difference is that it minimizes model hallucination and system prompts, has a larger context window, and performs better in citations.
Claude 2(Legacy - sunsetting on 02/06/25)	`aai.LemurModel.claude2_0`	Claude 2 is a legacy model that has good complex reasoning. It offers more nuanced responses and improved contextual comprehension.

You can find more information on pricing for each model .

Change the maximum output size

You can change the maximum output size in tokens by specifying the max_output_size parameter. Up to 4000 tokens are allowed.

result = transcript.lemur.task(
    prompt,
    max_output_size=1000
)

const { response } = await client.lemur.task({
  transcript_ids: [transcript.id],
  prompt,
  max_output_size: 1000
})

var params aai.LeMURTaskParams
params.Prompt = aai.String(prompt)
params.TranscriptIDs = []string{aai.ToString(transcript.ID)}
params.MaxOutputSize = aai.Int64(2000)

result, _ := client.LeMUR.Task(ctx, params)

var params = LemurTaskParams.builder()
        .prompt(prompt)
        .transcriptIds(List.of(transcript.getId()))
        .maxOutputSize(1000)
        .build();

var lemurTaskParams = new LemurTaskParams
{
    Prompt = prompt,
    TranscriptIds = [transcript.Id],
    MaxOutputSize = 1000
};

response = client.lemur.task(
  prompt: prompt,
  transcript_ids: [transcript_id],
  max_output_size: 1000
)

Change the temperature

You can change the temperature by specifying the temperature parameter, ranging from 0.0 to 1.0. Higher values result in answers that are more creative, lower values are more conservative.

result = transcript.lemur.task(
    prompt,
    temperature=0.7
)

const { response } = await client.lemur.task({
  transcript_ids: [transcript.id],
  prompt,
  temperature: 0.7
})

var params aai.LeMURTaskParams
params.Prompt = aai.String(prompt)
params.TranscriptIDs = []string{aai.ToString(transcript.ID)}
params.Temperature = aai.Float64(0.7)

result, _ := client.LeMUR.Task(ctx, params)

var params = LemurTaskParams.builder()
        .prompt(prompt)
        .transcriptIds(List.of(transcript.getId()))
        .temperature(0.7)
        .build();

var params = LemurTaskParams.builder()
        .prompt(prompt)
        .transcriptIds(List.of(transcript.getId()))
        .temperature(0.7)
        .build();

response = client.lemur.task(
  prompt: prompt,
  transcript_ids: [transcript_id],
  temperature: 0.7
)

Send customized input

You can submit custom text inputs to LeMUR without transcript IDs. This allows you to customize the input, for example, you could include the speaker labels for the LLM. To submit custom text input, use the input_text parameter on aai.Lemur().task().

config = aai.TranscriptionConfig(
  speaker_labels=True,
)
transcript = transcriber.transcribe(audio_url, config=config)

text_with_speaker_labels = ""
for utt in transcript.utterances:
    text_with_speaker_labels += f"Speaker {utt.speaker}:\n{utt.text}\n"

result = aai.Lemur().task(
    prompt,
    input_text=text_with_speaker_labels
)

const params = {
  audio: audioUrl,
  speaker_labels: true
}
const transcript = await client.transcripts.transcribe(params)

const textWithSpeakerLabels = ''
for (let utterance of transcript.utterances!) {
  textWithSpeakerLabels += `Speaker ${utterance.speaker}:\n${utterance.text}\n`
}

const { response } = await client.lemur.task({
  prompt: prompt,
  input_text: textWithSpeakerLabels
})

transcript, _ := client.Transcripts.TranscribeFromURL(ctx, audioURL, &aai.TranscriptOptionalParams{
    SpeakerLabels: aai.Bool(true),
})

var textWithSpeakerLabels string

for _, utterance := range transcript.Utterances {
    textWithSpeakerLabels += fmt.Sprintf("Speaker %s:\n%s\n",
        aai.ToString(utterance.Speaker),
        aai.ToString(utterance.Text),
    )
}

var params aai.LeMURTaskParams
params.Prompt = aai.String(prompt)
params.InputText = aai.String(textWithSpeakerLabels)

result, _ := client.LeMUR.Task(ctx, params)

var params = TranscriptOptionalParams.builder()
        .speakerLabels(true)
        .build();

Transcript transcript = client.transcripts().transcribe(audioUrl, params);

String textWithSpeakerLabels = transcript.getUtterances()
        .map(utterances -> utterances.stream()
                .map(utterance -> "Speaker " + utterance.getSpeaker() + ":\n" + utterance.getText() + "\n")
                .collect(Collectors.joining()))
        .orElse("");

var response = client.lemur().task(LemurTaskParams.builder()
        .prompt(prompt)
        .inputText(textWithSpeakerLabels)
        .build());

var transcript = await client.Transcripts.TranscribeAsync(new TranscriptParams
{
    AudioUrl = "https://assembly.ai/sports_injuries.mp3",
    SpeakerLabels = true
});

var textWithSpeakerLabels = string.Join(
    "",
    transcript.Utterances!.Select(utterance => $"Speaker {utterance.Speaker}:\n{utterance.Text}\n")
);

var lemurTaskParams = new LemurTaskParams
{
    Prompt = prompt,
    InputText = textWithSpeakerLabels
};

var response = await client.Lemur.TaskAsync(lemurTaskParams);

transcript = client.transcripts.transcribe(
  audio_url: audio_url,
  speaker_labels: true
)

text_with_speaker_labels = (transcript.utterances.map do |utterance|
  sprintf(
    "Speaker %<speaker>s:\n%<text>s\n",
    speaker: utterance.speaker,
    text: utterance.text
  )
end).join("\n")

response = client.lemur.task(
  prompt: prompt,
  input_text: text_with_speaker_labels
)

puts response.response

Submit multiple transcripts

LeMUR can easily ingest multiple transcripts in a single API call. You can feed in up to a maximum of 100 files or 100 hours, whichever is lower.

transcript_group = transcriber.transcribe_group(
    [
        "https://example.org/customer1.mp3",
        "https://example.org/customer2.mp3",
        "https://example.org/customer3.mp3",
    ],
)

# Or use existing transcripts:
# transcript_group = aai.TranscriptGroup.get_by_ids([id1, id2, id3])

result = transcript_group.lemur.task(
  prompt="Provide a summary of these customer calls."
)

const { response } = await client.lemur.task({
  transcript_ids: [id1, id2, id3],
  prompt: 'Provide a summary of these customer calls.'
})

var params aai.LeMURTaskParams
params.TranscriptIDs = []string{id1, id2, id3}
params.Prompt = aai.String("Provide a summary of these customer calls.")

result, _ := client.LeMUR.Task(ctx, params)

var response = client.lemur().task(LemurTaskParams.builder()
        .prompt(prompt)
        .transcriptIds(List.of(id1, id2, id3))
        .build());

var lemurTaskParams = new LemurTaskParams
{
    Prompt = prompt,
    TranscriptIds = [id1, id2, id3]
};

response = client.lemur.task(
  prompt: prompt,
  transcript_ids: [id1, id2, id3]
)

Delete data

You can delete the data for a previously submitted LeMUR request. Response data from the LLM, as well as any context provided in the original request will be removed.

result = transcript.lemur.task(prompt)

deletion_response = aai.Lemur.purge_request_data(result.request_id)

const { response, request_id } = await client.lemur.task({
  transcript_ids: [transcript.id],
  prompt
})

const deletionResponse = await client.lemur.purgeRequestData(request_id)

var response = client.lemur().task(LemurTaskParams.builder()
        .prompt(prompt)
        .transcriptIds(List.of(transcript.getId()))
        .build());

var deletionResponse = client.lemur().purgeRequestData(response.getRequestId());

var response = await client.Lemur.TaskAsync(lemurTaskParams);

var deletionResponse = await client.Lemur.PurgeRequestDataAsync(response.RequestId);

response = client.lemur.task(
  prompt: prompt,
  transcript_ids: [transcript_id],
)

deletion_response = client.lemur.purge_request_data(request_id: response.request_id)

API reference

You can find detailed information about all LeMUR API endpoints and parameters in the LeMUR API reference.

​Change the model type

​Change the maximum output size

​Change the temperature

​Send customized input

​Submit multiple transcripts

​Delete data

​API reference