Nodes/ComfyUI Griptape Nodes/Griptape Agent Config: Groq Drivers
ComfyUI Node

Griptape Agent Config: Groq Drivers

Chat plus Whisper transcription

By griptape-ai·Created 2 years ago·Updated about a year ago· 238
Griptape Agent Config: Groq Drivers
    • CONFIG
    prompt_model_commentPrompt Driver
    modelgemma2-9b-it
    max_attempts_on_fail2
    temperature0.10
    seed10342349342
    use_native_toolstrue
    max_tokens-1
    min_p0.10
    response_formatdefault
    api_key_env_varGROQ_API_KEY
    audio_transcription_model_commentAudio Transcription Driver
    audio_transcription_modelwhisper-large-v3-turbo

    Groq is the "tokens are basically free and absurdly fast" provider, and this config node wires Griptape to it - a prompt driver for chat and an audio transcription driver for Whisper, all on one GROQ_API_KEY. The free tier is famously generous, which makes this the config people reach for when they want to test a Griptape agent without spending real money. It's not the most feature-complete config in the pack, but for cheap chat it's hard to beat.

    How it works

    The node builds a DriversConfig with two drivers bundled in:

    • A prompt driver using Groq's chat models - the dropdown lists a dozen, from gemma2-9b-it (default) through Llama variants and Whisper's neighbors.
    • An audio transcription driver built on Whisper (whisper-large-v3-turbo default), which the pack implements by pointing the OpenAI transcription driver at Groq's OpenAI-compatible endpoint, https://api.groq.com/openai/v1/.

    Both come out bundled in one CONFIG output. The notable gap is deliberate: Groq doesn't offer an embedding driver, so the changelog notes that to use off-prompt tools (which need embeddings), you have to bring your own embedding driver separately.

    The inputs that matter

    • model (default gemma2-9b-it) - the chat model from Groq's lineup.
    • api_key_env_var (default GROQ_API_KEY) - key name; grab a free one at https://console.groq.com/keys.
    • audio_transcription_model (default whisper-large-v3-turbo, with whisper-large-v3 and distil-whisper-large-v3-en options) - for the transcription driver.
    • Shared prompt knobs: temperature, seed, min_p, max_tokens, response_format, use_native_tools, max_attempts_on_fail.

    Output is a single CONFIG.

    Installing it

    Standard one-pack install:

    cd ComfyUI/custom_nodes
    git clone https://github.com/griptape-ai/ComfyUI-Griptape
    

    or ComfyUI Manager → search "Griptape" → install. Restart, set GROQ_API_KEY in Settings → Griptape (or as an env var), done. Same shared caveat as always: the pack drags in a torch build via griptape[all] that can fight your ComfyUI torch - README troubleshooting has the recovery steps.

    Common issues

    • [WARNING]: GROQ_API_KEY is not set - the usual key-missing console warning; verify the settings entry or env var name.
    • Agent errors on tool/memory use - that's the missing embedding driver biting you. Groq has no embedding endpoint, so attach a separate embedding driver (e.g. OpenAI's or a local one) before expecting off-prompt tools or retrieval to work.
    • Whisper transcriptions failing - check audio_transcription_model is one of the three listed; Groq only serves those Whisper flavors through its API.

    If you just want a fast, nearly-free chat agent to experiment with, this config plus the free Groq key is genuinely the cheapest way into the Griptape graph. Just remember you're giving up embeddings and image generation compared to the OpenAI config.

    CategoryGriptape/Agent Configs

    Inputs (12)

    NameTypeDefaultDescription
    prompt_model_commentoptSTRINGPrompt Driver
    modeloptCOMBOgemma2-9b-itThe model to use for the chat prompt.
    max_attempts_on_failoptINT21–100Maximum attempts on failure
    temperatureoptFLOAT0.100–1Temperature for sampling
    seedoptINT10342349342Seed for random number generation
    use_native_toolsoptBOOLEANtrueWhether to use native tools.
    max_tokensoptINT-1Maximum tokens to generate. If <=0, it will use the default based on the tokenizer.
    min_poptFLOAT0.100–1Minimum probability for sampling. Lower values will be more random.
    response_formatoptCOMBOdefaultFormat of the response
    api_key_env_varoptSTRINGGROQ_API_KEYEnter the name of the environment variable that contains the API key, not the API key itself.
    audio_transcription_model_commentoptSTRINGAudio Transcription Driver
    audio_transcription_modeloptCOMBOwhisper-large-v3-turboSelect the audio transcription model to use.

    Outputs (1)

    NameTypeDescription
    CONFIGCONFIG