LLM Providers

Configure Strix to work with different LLM providers including OpenAI, Anthropic, Google, and local models

Strix uses LiteLLM to support a wide range of LLM providers. You can use cloud-based models, local models, or the Strix Router for unified access.

Recommended Models

For best results, we recommend using one of these frontier models:

• OpenAI GPT-5 - openai/gpt-5
• Anthropic Claude Sonnet 4.6 - anthropic/claude-sonnet-4-6
• Google Gemini 3 Pro Preview - vertex_ai/gemini-3-pro-preview

Strix Router

Strix Router provides a single API key for accessing multiple LLM providers with intelligent routing and $10 free credit on signup.

export STRIX_LLM="strix/gpt-5"
export LLM_API_KEY="your-strix-router-key"

Available models via Strix Router:

• strix/gpt-5
• strix/claude-sonnet-4-6
• strix/gemini-3-pro-preview

When using the strix/ prefix, the API base URL is automatically set to https://models.strix.ai/api/v1. You don’t need to configure LLM_API_BASE.

Cloud Providers

OpenAI

export STRIX_LLM="openai/gpt-5"
export LLM_API_KEY="sk-..."

Supported models:

• openai/gpt-5
• openai/gpt-4o
• openai/o1
• openai/o3-mini

Get your API key at platform.openai.com.

Anthropic

export STRIX_LLM="anthropic/claude-sonnet-4-6"
export LLM_API_KEY="sk-ant-..."

Supported models:

• anthropic/claude-sonnet-4-6
• anthropic/claude-opus-4
• anthropic/claude-3.5-sonnet

Get your API key at console.anthropic.com.

Google Cloud (Vertex AI)

export STRIX_LLM="vertex_ai/gemini-3-pro-preview"
# Authentication via gcloud CLI or service account

Supported models:

• vertex_ai/gemini-3-pro-preview
• vertex_ai/gemini-2.0-flash-exp
• vertex_ai/gemini-1.5-pro

Vertex AI uses Google Cloud authentication. You need to authenticate via gcloud auth application-default login or set GOOGLE_APPLICATION_CREDENTIALS to your service account JSON file.

AWS Bedrock

export STRIX_LLM="bedrock/anthropic.claude-sonnet-4-6-v1:0"
# Authentication via AWS credentials

Supported models:

• bedrock/anthropic.claude-sonnet-4-6-v1:0
• bedrock/anthropic.claude-opus-4-v1:0
• bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0

AWS Bedrock uses AWS credentials. Configure your credentials via AWS CLI (aws configure) or environment variables (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION).

Azure OpenAI

export STRIX_LLM="azure/your-deployment-name"
export LLM_API_KEY="your-azure-api-key"
export LLM_API_BASE="https://your-resource.openai.azure.com"

You can also use Azure-specific environment variables:

export AZURE_API_KEY="your-azure-api-key"
export AZURE_API_BASE="https://your-resource.openai.azure.com"
export AZURE_API_VERSION="2024-02-15-preview"

Local Models

Ollama

Ollama lets you run LLMs locally on your machine.

export STRIX_LLM="ollama/llama3.1:70b"
export LLM_API_BASE="http://localhost:11434"

You don’t need to set LLM_API_KEY when using Ollama. Make sure Ollama is running (ollama serve) before starting Strix.

Recommended models for security testing:

• ollama/llama3.1:70b
• ollama/qwen2.5:72b
• ollama/deepseek-v3

LM Studio

LM Studio provides a local server for running LLMs with an OpenAI-compatible API.

export STRIX_LLM="openai/model-name"
export LLM_API_BASE="http://localhost:1234/v1"

When using LM Studio, the model name should match the model loaded in LM Studio, or use a generic name like openai/local-model.

Other Local Providers

Strix works with any OpenAI-compatible API endpoint:

export STRIX_LLM="openai/your-model"
export LLM_API_BASE="http://your-server:port/v1"

Advanced Configuration

Custom Timeouts

Adjust LLM request timeouts for slower models or connections:

export LLM_TIMEOUT="600"  # 10 minutes

Retry Configuration

Control how many times Strix retries failed LLM requests:

export STRIX_LLM_MAX_RETRIES="10"

Reasoning Effort

Control the reasoning effort level for better or faster responses:

# For thorough analysis (slower)
export STRIX_REASONING_EFFORT="xhigh"

# For quick scans (faster)
export STRIX_REASONING_EFFORT="medium"

# For minimal overhead
export STRIX_REASONING_EFFORT="low"

Provider-Specific Notes

Using Multiple Providers

You can use different models for different purposes by switching STRIX_LLM:

# Quick scan with smaller model
export STRIX_LLM="openai/gpt-4o"
strix --target ./app --scan-mode quick

# Deep analysis with larger model
export STRIX_LLM="openai/gpt-5"
strix --target ./app --scan-mode deep

Authentication Priority

Strix checks for credentials in this order:

1. LLM_API_KEY environment variable
2. Provider-specific environment variables (e.g., OPENAI_API_KEY, ANTHROPIC_API_KEY)
3. Provider-specific authentication mechanisms (e.g., gcloud, AWS credentials)

Base URL Priority

Strix checks for base URLs in this order:

1. Automatic detection for strix/ models
2. LLM_API_BASE environment variable
3. OPENAI_API_BASE environment variable
4. LITELLM_BASE_URL environment variable
5. OLLAMA_API_BASE environment variable
6. Provider default URLs

Troubleshooting

Connection Failed

If you see “LLM CONNECTION FAILED”, verify:

1. Your API key is correct and has the necessary permissions
2. Your model name is correct (e.g., openai/gpt-5, not gpt-5)
3. Your API base URL is correct (for local models)
4. Your network can reach the API endpoint
5. Your model/deployment exists and is accessible

Model Not Found

Ensure you’re using the correct provider prefix:

• ✅ openai/gpt-5
• ❌ gpt-5

Rate Limiting

If you hit rate limits, you can:

1. Reduce STRIX_REASONING_EFFORT to make fewer requests
2. Use a different model with higher rate limits
3. Increase your API plan limits

Local Model Performance

For best results with local models:

1. Use models with at least 70B parameters
2. Ensure you have sufficient RAM/VRAM
3. Use GPU acceleration when possible
4. Consider using --scan-mode quick for faster scans

See Also

• Environment Variables
• Config File
• LiteLLM Providers Documentation