Complete guide for integrating Kimi 2.5 AI model with OpenClaw running on Deep Layer cloud services

Using Kimi 2.5 with OpenClaw on Deep Layer

This guide provides detailed instructions for integrating Kimi 2.5 (hosted on Groq) with OpenClaw running on Deep Layer cloud services.

Overview

Kimi 2.5 is available through Groq's API as moonshotai/kimi-k2-instruct-0905 with the following specifications:
- Speed: 200 tokens/second
- Pricing: $1.00 input / $3.00 output per 1M tokens
- Context Window: 262,144 tokens
- Max Completion: 16,384 tokens
- Rate Limits: 250K TPM, 1K RPM (Developer plan)

Prerequisites

Groq API Key: Sign up at console.groq.com and generate an API key
OpenClaw Instance: Running on Deep Layer cloud services
Access to OpenClaw Configuration: Ability to modify the gateway configuration

Configuration Steps

Step 1: Add Kimi 2.5 to OpenClaw Model Catalog

Add the following configuration to your OpenClaw gateway config under agents.defaults.models:

agents:
  defaults:
    models:
      custom-provider/moonshotai/kimi-k2-instruct-0905:
        alias: "kimi-2.5"
        params:
          baseUrl: "https://api.groq.com/openai/v1"
          apiKey: "${GROQ_API_KEY}"  # Use environment variable
          model: "moonshotai/kimi-k2-instruct-0905"
        streaming: true

Step 2: Configure Environment Variables

Set the Groq API key as an environment variable on your Deep Layer instance:

# Add to your environment (via Deep Layer dashboard or SSH)
export GROQ_API_KEY="gsk_your_actual_api_key_here"

# Make it persistent
 echo "export GROQ_API_KEY=gsk_your_actual_api_key_here" >> ~/.bashrc

Step 3: Set Kimi 2.5 as Default Model (Optional)

To use Kimi 2.5 as your default model, update the agent defaults:

agents:
  defaults:
    model: "custom-provider/moonshotai/kimi-k2-instruct-0905"

Step 4: Apply Configuration

Apply the configuration changes to your OpenClaw gateway:

# Apply the new configuration
openclaw gateway config.apply

# Or restart the gateway
openclaw gateway restart

Usage Examples

Basic Usage

Once configured, you can use Kimi 2.5 in your OpenClaw sessions:

# Use with specific model flag
openclaw --model custom-provider/moonshotai/kimi-k2-instruct-0905

# Or if you've set it as default, just use normally
openclaw

In Configuration Files

Reference Kimi 2.5 in your agent configurations:

agents:
  list:
    my-agent:
      model: "custom-provider/moonshotai/kimi-k2-instruct-0905"
      # ... other agent settings

Via API

When using OpenClaw's API, specify the model:

{
  "model": "custom-provider/moonshotai/kimi-k2-instruct-0905",
  "messages": [
    {"role": "user", "content": "Explain quantum computing"}
  ]
}

Deep Layer Specific Considerations

Environment Variable Management

On Deep Layer, you can set environment variables through:

Dashboard: Environment Variables section in your instance settings
SSH: Direct access to configure persistent environment variables
Startup Scripts: Add to your instance's startup configuration

Network Configuration

Deep Layer instances have outbound internet access by default, so connecting to Groq's API (api.groq.com) should work without additional firewall rules.

Monitoring and Logging

Monitor your Kimi 2.5 usage through:

Groq Console: Track API usage and costs
OpenClaw Logs: Check gateway logs for model interactions
Deep Layer Monitoring: Monitor instance performance

Troubleshooting

Common Issues

1. Authentication Errors
Error: 401 Unauthorized
Solution: Verify your GROQ_API_KEY is correctly set and active in the Groq console.

2. Model Not Found
Error: Model not found: moonshotai/kimi-k2-instruct-0905
Solution: Check that the exact model ID is used: moonshotai/kimi-k2-instruct-0905

3. Rate Limiting
Error: Rate limit exceeded
Solution: You're hitting Groq's rate limits. The Developer plan allows 250K tokens per minute and 1K requests per minute.

Verification Steps

Test API Connectivity:
bash curl https://api.groq.com/openai/v1/models \n -H "Authorization: Bearer $GROQ_API_KEY"
Check OpenClaw Model List:
bash openclaw models list
Test Model Directly:
bash echo "Hello, how are you?" | openclaw --model custom-provider/moonshotai/kimi-k2-instruct-0905

Cost Optimization

Token Usage Monitoring

Monitor your token consumption:
- Input tokens: $1.00 per 1M tokens
- Output tokens: $3.00 per 1M tokens
- Typical conversation: ~1-5K tokens per exchange

Usage Patterns

For cost-effective usage:
1. Batch requests when possible
2. Use appropriate context length - Kimi 2.5 supports up to 262K context
3. Monitor usage in Groq console
4. Set up alerts for unusual spending patterns

Advanced Configuration

Custom Parameters

You can add additional OpenAI-compatible parameters:

agents:
  defaults:
    models:
      custom-provider/moonshotai/kimi-k2-instruct-0905:
        alias: "kimi-2.5"
        params:
          baseUrl: "https://api.groq.com/openai/v1"
          apiKey: "${GROQ_API_KEY}"
          model: "moonshotai/kimi-k2-instruct-0905"
          temperature: 0.7
          max_tokens: 4000
          top_p: 0.9
        streaming: true

Multiple Model Setup

Configure multiple Kimi variants or other models:

agents:
  defaults:
    models:
      custom-provider/moonshotai/kimi-k2-instruct-0905:
        alias: "kimi-2.5"
        params:
          baseUrl: "https://api.groq.com/openai/v1"
          apiKey: "${GROQ_API_KEY}"
          model: "moonshotai/kimi-k2-instruct-0905"
        streaming: true

      custom-provider/llama-3.3-70b:
        alias: "llama-3.3"
        params:
          baseUrl: "https://api.groq.com/openai/v1"
          apiKey: "${GROQ_API_KEY}"
          model: "llama-3.3-70b-versatile"
        streaming: true

Support and Resources

Groq Documentation: console.groq.com/docs
OpenClaw Documentation: docs.openclaw.ai
Deep Layer Support: Through your Deep Layer dashboard
Model Status: Check status.groq.com

Security Best Practices

API Key Storage: Use environment variables, not hardcoded values
Key Rotation: Regularly rotate your Groq API keys
Access Control: Limit who can access your OpenClaw configuration
Monitoring: Set up alerts for unusual usage patterns
Backup: Keep backups of your configuration files

This guide is specific to Deep Layer's OpenClaw deployment. For other hosting providers, adjust the network and environment configuration accordingly.