Description

Running OpenClaw with custom model. We use K2Think as base model.

This setup uses the IFM Model API with:

• Model: MBZUAI-IFM/K2-Think-v2
• Base URL: https://api.k2think.ai
• Chat endpoint: /v1/chat/completions
• Authentication: Authorization: Bearer <API_KEY> (mapped from OPENCLAW_OPENAI_API_KEY)

K2 Think v2 is positioned for advanced reasoning tasks such as coding, math, and complex problem solving.

Based on IFM docs defaults, keep these in mind for planning usage:

• Rate limit: 30 requests/minute (unless approved otherwise)
• Concurrency: 10 concurrent requests (unless approved otherwise)
• Max request tokens: 65536 (unless approved otherwise)

How To Use

1. Run:

./scripts/build.sh

2. Set environment variables:

export PORT=3000
export OPENCLAW_OPENAI_API_KEY="your-k2think-token" # get token from https://platform.ifm.ai/
export OPENCLAW_GATEWAY_TOKEN="any-random-string"
export OPENCLAW_TELEGRAM_BOT_TOKEN="telegram-token"

3. Run:

./scripts/start.sh

4. Start chat with your bot from Telegram.