Bifrost Integration Guide¶

Learn how to integrate AccuKnox Prompt Firewall with Bifrost using a custom Go plugin to monitor and filter prompts and responses via the AccuKnox API.

alt text

Step 1: Prerequisites¶

We expect you to have the bifrost-http binary in-place. You can download it from the Bifrost Releases.

This integration guide uses the bifrost-http binary to test the configuration. You can use this config with any Bifrost deployment method.

Step 2: Download AccuKnox Plugin¶

Download the latest AccuKnox plugin for Bifrost:

Using wget¶

wget https://github.com/accuknox/bifrost-accuknox-integration/releases/latest/download/accuknox-plugin.so

Using curl¶

curl -LO https://github.com/accuknox/bifrost-accuknox-integration/releases/latest/download/accuknox-plugin.so

Source Code

The plugin source code is available at github.com/accuknox/bifrost-accuknox-integration for reference.

Step 3: Configure Bifrost¶

Create or update your Bifrost configuration file with the following settings:

{
    "log_level": "debug",
    "server": {
      "port": 8080,
      "host": "0.0.0.0"
    },
    "plugins": [
      {
        "enabled": true,
        "name": "accuknox-logger",
        "path": "./accuknox-plugin.so",
        "config": {
          "enabled": true,
          "api_key": "<YOUR_ACCUKNOX_JWT_TOKEN>",
          "user_info": "<username@accuknox.com>"
        }
      }
    ],
    "providers": {
      "openai": {
        "keys": [
          {
            "value": "<YOUR_OPENAI_API_KEY>",
            "models": ["gpt-3.5-turbo"],
            "weight": 1.0
          }
        ]
      }
    }
}

Step 4: Configuration Details¶

Plugin Configuration¶

enabled: Set to true to activate the plugin
name: Plugin identifier (must match the name returned by GetName())
path: Path to the compiled .so file
api_key: Your AccuKnox Prompt Firewall JWT token (obtained from AccuKnox dashboard)
user_info: Your email or username for tracking

Provider Configuration¶

Add your LLM provider credentials (OpenAI, Anthropic, etc.)
Each key can be restricted to specific models
Use weight for load balancing across multiple keys

Step 5: Run Bifrost Server¶

Copy the Plugin¶

cp accuknox-plugin.so /path/to/bifrost/
cd /path/to/bifrost/

Start the Server¶

./bifrost-http -app-dir . -log-level debug -log-style pretty

Step 6: Test the Integration¶

Send a test request using curl:

curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-3.5-turbo",
    "messages": [{"role": "user", "content": "Say hello"}],
    "max_tokens": 50
  }'