Support additional API providers (OpenRouter, Gemini, etc.)

[somdipto]

What Needs Doing:

• Define a base provider interface (common methods: send, stream, healthCheck)

• Build adapters for OpenRouter and Gemini

• Make config / UI switch to select provider

• Add tests (unit & integration)

• Update documentation and usage examples

Why This Matters:

• Enables switching providers easily

• Prepares system for future integrations

• Reduces dependency risk

Assign to me or anyone interested.

[Dylan-86]

Ok, I will add here the guide to configure Bytebot to support more AI models, including OpenRouter. It's not as straight forward as it seemed when I read the documentation.

Guide: Integrating Bytebot with OpenRouter via LiteLLM

1. Prerequisites

• Working Docker + Docker Compose setup

• Bytebot cloned locally (packages/ available)

• An OpenRouter API key (from https://openrouter.ai/keys)

2. Configure docker-compose.proxy.yml

Add your API key to the bytebot-llm-proxy service. Example:

  bytebot-llm-proxy:
    build:
      context: ../packages/
      dockerfile: bytebot-llm-proxy/Dockerfile
    ports:
      - "4000:4000"
    environment:
      - GEMINI_API_KEY=sk-or-v1-xxxxxxxxxxxxxxxxxxxxxxxxxxxx
    networks:
      - bytebot-network

Note: The variable name can stay GEMINI_API_KEY — it’s just used as a placeholder for the OpenRouter key.

3. Update LiteLLM configuration

Edit packages/bytebot-llm-proxy/litellm-config.yaml.Use OpenRouter model IDs (the ones listed on https://openrouter.ai/docs).Here’s a working example:

model_list:

  - model_name: grok
    litellm_params:
      model: openrouter/x-ai/grok-4-fast:free
      api_base: https://openrouter.ai/api/v1
      api_key: os.environ/GEMINI_API_KEY
      drop_params: true

  - model_name: ds3
    litellm_params:
      model: openrouter/deepseek/deepseek-chat-v3.1:free
      api_base: https://openrouter.ai/api/v1
      api_key: os.environ/GEMINI_API_KEY
      drop_params: true

  - model_name: gpt-oss
    litellm_params:
      model: openrouter/openai/gpt-oss-120b:free
      api_base: https://openrouter.ai/api/v1
      api_key: os.environ/GEMINI_API_KEY
      drop_params: true

Key points:

• model must exactly match the OpenRouter model ID (e.g. openrouter/x-ai/grok-4-fast:free).

• api_key: os.environ/GEMINI_API_KEY tells LiteLLM to pull the API key from the container environment.

• drop_params: true avoids passing extra params OpenRouter doesn’t support.

4. Rebuild and restart containers

docker compose -f docker/docker-compose.proxy.yml down docker compose -f docker/docker-compose.proxy.yml up -d --build

5. Verify models are registered

Check if LiteLLM proxy exposes the models:

curl http://localhost:4000/v1/models

Expected output includes your configured models:

{
  "data": [
    {"id":"grok","object":"model"},
    {"id":"ds3","object":"model"},
    {"id":"gpt-oss","object":"model"}
  ]
}

Also verify Bytebot agent sees them:

curl http://localhost:9991/tasks/models

6. Troubleshooting

• 401 Unauthorized → API key missing. Ensure GEMINI_API_KEY is set in docker-compose.proxy.yml.

• Invalid model ID → Wrong OpenRouter model name. Check OpenRouter model catalog.

• YAML parse errors → Misaligned indentation. Each - model_name must align under model_list:.

7. Final Notes

• The GEMINI_API_KEY name is arbitrary; it doesn’t have to match the model provider.

• You can add as many models as you like from OpenRouter by repeating the block in litellm-config.yaml.

• Always restart the proxy after modifying the config.

[somdipto]

hey dylan , have already done the changes , if you could review the pr , please go ahead

[Dylan-86]

hey dylan , have already done the changes , if you could review the pr , please go ahead

Hey! I just tested locally.
I could load and run it, I set my .env like this:

# Database
DATABASE_URL=postgresql://postgres:postgres@postgres:5432/bytebotdb

# Proxy URL for the LLM proxy service
BYTEBOT_LLM_PROXY_URL=http://bytebot-llm-proxy:4000


OPENROUTER_API_KEY=MY-API-KEY
ANTHROPIC_API_KEY=
OPENAI_API_KEY=
GEMINI_API_KEY=

# Bytebot configs
BYTEBOT_DESKTOP_BASE_URL=http://localhost:9990
BYTEBOT_ANALYTICS_ENDPOINT=

Notes!!!

1. I had to manually add the models inside litellm-config.yaml like this

model_list:
  # Anthropic Models
  - model_name: claude-opus-4
    litellm_params:
      model: anthropic/claude-opus-4-20250514
      api_key: os.environ/ANTHROPIC_API_KEY
  - model_name: claude-sonnet-4
    litellm_params:
      model: anthropic/claude-sonnet-4-20250514
      api_key: os.environ/ANTHROPIC_API_KEY

  # OpenAI Models
  - model_name: gpt-4.1
    litellm_params:
      model: openai/gpt-4.1
      api_key: os.environ/OPENAI_API_KEY
  - model_name: gpt-4o
    litellm_params:
      model: openai/gpt-4o
      api_key: os.environ/OPENAI_API_KEY

  # Gemini Models
  - model_name: gemini-2.5-pro
    litellm_params:
      model: gemini/gemini-2.5-pro
      api_key: os.environ/GEMINI_API_KEY
  - model_name: gemini-2.5-flash
    litellm_params:
      model: gemini/gemini-2.5-flash
      api_key: os.environ/GEMINI_API_KEY

  - model_name: grok
    litellm_params:
      model: openrouter/x-ai/grok-4-fast:free
      api_base: https://openrouter.ai/api/v1
      api_key: os.environ/OPENROUTER_API_KEY
      drop_params: true

  - model_name: ds3
    litellm_params:
      model: openrouter/deepseek/deepseek-chat-v3.1:free
      api_base: https://openrouter.ai/api/v1
      api_key: os.environ/OPENROUTER_API_KEY
      drop_params: true
  - model_name: gpt-oss
    litellm_params:
      model: openrouter/openai/gpt-oss-120b:free
      api_base: https://openrouter.ai/api/v1
      api_key: os.environ/OPENROUTER_API_KEY
      drop_params: true

2. I had to add this line inside docker-compose.proxy.yml

  bytebot-llm-proxy:
    build:
      context: ../packages/
      dockerfile: bytebot-llm-proxy/Dockerfile
    ports:
      - "4000:4000"
    environment:
      - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY}
      - OPENAI_API_KEY=${OPENAI_API_KEY}
      - GEMINI_API_KEY=${GEMINI_API_KEY}
      - OPENROUTER_API_KEY=${OPENROUTER_API_KEY} # <<---- ADD THIS LINE!!
    networks:
      - bytebot-network

And I'm running the code with

docker compose -f docker/docker-compose.proxy.yml --env-file .env up -d --build

For the rest, it is working well.

I would strongly suggest adding some instructions about the above notes inside the README, honestly :)