🥇 Groq · Llama 3.3 70B (free)

Recommended #1. Groq serves Llama 3.3 70B fast enough to fit comfortably inside Bouncer's 1.5-second budget.

1. Get a free key

1. Open console.groq.com/keys and sign in.
2. Click Create API Key, name it bouncer, and copy it (it starts with gsk_).

2. Connect it (one line)

bouncer config --preset groq-gold --key gsk_your_key_here

To keep the key out of your shell history, set it as an environment variable and leave out --key:

export GROQ_API_KEY=gsk_your_key_here
bouncer config --preset groq-gold

Bouncer saves the config, then sends one test request:

  Intelligence  🥇 LEVEL 3 (Gold 70B/Flash) · Groq Cloud · llama-3.3-70b-versatile
  🥇 LEVEL 3 unlocked! Semantic file resolution + AI rescue summaries are on.
  Testing Groq Cloud (llama-3.3-70b-versatile)… ok (180ms → TRIVIAL, confidence 0.9)

Preset details

Endpoint	https://api.groq.com/openai/v1/chat/completions
Model	llama-3.3-70b-versatile
Level	🥇 3, confidence gate 0.80
Sent per prompt	prompt text + up to 40 file paths (never file contents)

Use a different Groq model with --model, for example bouncer config --preset groq-gold --model llama-3.1-8b-instant (grades as 🥈 Silver).

Troubleshooting

• failed (HTTP 401): the key is wrong or was revoked. Create a new one.
• failed (HTTP 429): you hit Groq's free-tier rate limit. Bouncer falls back to Level 1 until it resets, and your prompts are never held up.
• Remove it: bouncer config --clear.