What LLM Providers Are Supported by ChocolateLMLite? Complete List & Configuration Guide

ChocolateLMLite supports any OpenAI-compatible API endpoint, including pre-configured integrations for OpenRouter, Ollama, NanoGPT, LM Studio, llama-server, OpenAI, Google Gemini, and Anthropic Claude.

ChocolateLMLite (gpsnmeajp/chocolatelmlite) is built around the OpenAI API specification, allowing it to connect to virtually any large language model service that exposes a standard /v1/chat/completions endpoint. The repository ships with hardcoded base URLs and default model configurations in src/FileManager.cs and README.md that cover both cloud-hosted and local inference options.

Cloud-Based LLM Providers Supported by ChocolateLMLite

The application includes built-in support for several popular cloud providers through configurable base URLs documented in the source code.

OpenRouter (Recommended)

According to README.md (lines 91-94), OpenRouter is the default and most recommended provider. The pre-configured base URL is https://openrouter.ai/api/v1, which supports chat completion, image generation, and tool use. The opencode.json configuration file demonstrates OpenRouter routing with specific models like openai/gpt-oss-120b and qwen/qwen3-32b.

Ollama Cloud and NanoGPT

For alternative cloud options, ChocolateLMLite includes:

• Ollama Cloud: Base URL https://ollama.com/v1/ with a free tier available (README.md lines 92-94)
• NanoGPT: Base URL https://nano-gpt.com/api/v1 operating on a paid tier (README.md lines 93-94)

Major API Providers (OpenAI, Gemini, Claude)

The codebase explicitly documents support for industry-standard APIs:

• OpenAI: https://api.openai.com/v1/ — works through the generic OpenAI-compatible wrapper (README.md lines 106-107)
• Google Gemini: https://generativelanguage.googleapis.com/v1beta/openai/ — verified with free tier support (README.md lines 107-108)
• Anthropic Claude: https://api.anthropic.com/v1/ — supported via the OpenAI-compatible endpoint wrapper (README.md lines 108-109)

Additional providers like Deepseek, Phi, and Qwen are mentioned in the documentation (README.md line 20) and configured via OpenRouter routing.

Local and Self-Hosted LLM Providers

ChocolateLMLite equally supports local inference servers running on the host machine or local network.

LM Studio

For users running LM Studio, the default local endpoint is http://127.0.0.1:1234/v1 as documented in README.md (lines 96-99). This configuration works with any model exposed through the LM Studio server interface without cloud dependencies.

Self-Hosted Ollama and llama-server

The local server options extend to:

• Ollama (self-hosted): http://127.0.0.1:11434/v1 — OpenAI-compatible local server (README.md lines 96-99)
• llama-server: http://127.0.0.1:8080/v1 — Supported for chat completions, though image generation is not yet implemented for this backend (README.md lines 96-99)

Default Model Configuration in ChocolateLMLite

The source code hardcodes specific default models in src/FileManager.cs. At lines 90-91, the DefaultModel constant is set to google/gemini-2.5-flash. For image generation capabilities, lines 124-125 define ImageGenerationModel as google/gemini-2.5-flash-image.

You can override these defaults through configuration keys documented in API.md, which exposes settings like DefaultModel and ImageGenerationModel for runtime customization. For example, to configure a local Ollama endpoint:

{
  "BaseUrl": "http://127.0.0.1:11434/v1",
  "DefaultModel": "llama3.1",
  "ImageGenerationModel": ""
}

Summary

• ChocolateLMLite uses an OpenAI-compatible API architecture that supports virtually any LLM provider with a standard /v1/chat/completions endpoint.
• Cloud providers include OpenRouter (default), Ollama Cloud, NanoGPT, OpenAI, Google Gemini, and Anthropic Claude.
• Local providers include LM Studio (127.0.0.1:1234), self-hosted Ollama (127.0.0.1:11434), and llama-server (127.0.0.1:8080).
• Default models are configured in src/FileManager.cs with google/gemini-2.5-flash for chat and google/gemini-2.5-flash-image for image generation.
• Provider URLs and model routing are documented in README.md and opencode.json.

Frequently Asked Questions

Does ChocolateLMLite support the official OpenAI API?

Yes. While the repository notes this as unverified in testing, the codebase is explicitly architected to accept the OpenAI endpoint at https://api.openai.com/v1/. The OpenAI-compatible request wrapper handles authentication and payload formatting for this and any similar provider.

Can I run ChocolateLMLite with completely local models without internet access?

Yes. The application supports fully offline inference through LM Studio (port 1234), self-hosted Ollama (port 11434), and llama-server (port 8080). These configurations allow you to run models entirely on your local machine as documented in README.md lines 96-99.

What is the default LLM provider when I first install ChocolateLMLite?

The default configuration uses OpenRouter as the primary provider with google/gemini-2.5-flash set as the default chat model, as defined in src/FileManager.cs at lines 90-91. For image generation, it defaults to google/gemini-2.5-flash-image at lines 124-125.

Is image generation supported across all LLM providers in ChocolateLMLite?

No. According to the source code in README.md (lines 96-99), image generation is fully supported with OpenRouter and compatible cloud providers, but is not yet implemented for the llama-server local backend. Verify specific provider capabilities in the ImageGenerationModel configuration before attempting multimodal requests.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}