Groq

Configure evalflow to run fast, low-cost evals using Groq’s inference API.

Use Groq when you want fast CI checks without incurring high API costs. Groq offers a free tier with no credit card required, making it a strong default for teams that want routine regression checks on every pull request without slowing builds down.

Groq is also the default judge provider that evalflow uses to score outputs — so even if you test against a different provider, Groq is already part of your eval pipeline unless you override the judge block.

Configure evalflow.yaml

Add a groq block under providers and set default_provider to groq:

providers:
  groq:
    api_key_env: "GROQ_API_KEY"
    default_model: "llama-3.1-8b-instant"

eval:
  default_provider: "groq"

api_key_env is the name of the environment variable that holds your key — evalflow reads the variable at runtime and never stores the key itself.

Set your API key

Get a free key at console.groq.com, then export it:

export GROQ_API_KEY="your-key-here"

Add this line to your shell profile (~/.bashrc, ~/.zshrc, etc.) or a .env file so you do not have to re-export it each session. Never commit your .env file to version control.

Verify the connection

Run evalflow doctor to confirm evalflow can see the key before running any evals:

evalflow doctor

✓ GROQ_API_KEY set

Run evals

evalflow eval --provider groq

Running test cases against llama-3.1-8b-instant...
Quality Gate: PASS

If eval.default_provider is already set to groq in your evalflow.yaml, you can omit the --provider flag:

evalflow eval

Provider notes

• Default model: llama-3.1-8b-instant. This model is fast and well-suited for CI use cases where latency matters.
• Free tier: Groq’s free tier covers a generous number of tokens per day, enough for routine eval runs on most projects.
• Judge model: evalflow’s default judge is llama-3.1-8b-instant on Groq. If you want a different judge, update the judge block in evalflow.yaml:

judge:
  provider: "groq"
  model: "llama-3.1-8b-instant"

Best for

• CI/CD pipelines where fast feedback loops matter
• Teams on a budget or without a billing account on other providers
• Default judge-model usage without additional configuration