[Enhancement]: Please add OpenRouter connection

[iain0901]

Target Component

Core Services (Frontend UI/Backend API)

Enhancement Description

There are many models on OpenRouter and can accept many context lengths

Technical Details

No response

Designs and Mockups

No response

Alternative Solutions

No response

Verification

• I have checked that this enhancement hasn't been already proposed
• This enhancement aligns with PentAGI's goal of autonomous penetration testing
• I have considered the security implications of this enhancement
• I have provided clear use cases and benefits

[asdek]

Hi @iain0901,

thank you for your question, PentAGI has already supported OpenRouter and DeepInfra at least.

Using OpenRouter

1. Configure your .env file with the following settings:

   LLM_SERVER_URL=https://openrouter.ai/api/v1
   LLM_SERVER_KEY=your_api_key
   LLM_SERVER_MODEL=
   LLM_SERVER_CONFIG_PATH=/opt/pentagi/conf/openrouter.provider.yml
   

2. Run the ctester utility with the pre-configured OpenRouter provider file:

   docker run --rm \
     -v $(pwd)/.env:/opt/pentagi/.env \
     vxcontrol/pentagi /opt/pentagi/bin/ctester -config /opt/pentagi/conf/openrouter.provider.yml -verbose

Using Deep Infra

1. Configure your .env file with the following settings:

   LLM_SERVER_URL=https://api.deepinfra.com/v1/openai
   LLM_SERVER_KEY=your_api_key
   LLM_SERVER_MODEL=
   LLM_SERVER_CONFIG_PATH=/opt/pentagi/conf/deepinfra.provider.yml
   

2. Run the ctester utility with the pre-configured Deep Infra provider file:

   docker run --rm \
     -v $(pwd)/.env:/opt/pentagi/.env \
     vxcontrol/pentagi /opt/pentagi/bin/ctester -config /opt/pentagi/conf/deepinfra.provider.yml -verbose

Customizing Models and Context Lengths

You can customize provider configurations to test specific models with various context lengths. For example:

simple:
  model: "openai/gpt-4o"
  temperature: 0.7
  max_tokens: 4000

agent:
  model: "anthropic/claude-3.7-sonnet"
  temperature: 0.1
  max_tokens: 8000

# other agents

Testing Workflow

1. Run baseline tests with the default configuration.
2. Experiment with different models for each agent type.
3. Compare results by examining success rates and latencies.
4. Deploy the optimal configuration in your production environment.

For more detailed instructions, please refer to the Testing LLM Agents section in our README or you can see #17

Notes: both files /opt/pentagi/conf/deepinfra.provider.yml and /opt/pentagi/conf/openrouter.provider.yml are available in original the latest docker image vxcontrol/pentagi

In additional you could check reports for these providers in the repo:

• Deep Infra
• OpenRouter