Model Providers

Light-Fabric provides a unified, high-performance interface for interacting with diverse Large Language Model (LLM) providers. This abstraction is centered around the Provider trait, allowing applications to remain model-agnostic while leveraging advanced capabilities like native tool calling and prompt caching.

The Provider Trait

All model integrations implement the Provider trait, which supports:

  • One-shot and Multi-turn Chat: Simplified APIs for simple prompts and full conversation histories.
  • Structured Tool Calling: Native integration for function calling (OpenAI-style).
  • Capabilities Detection: Programmatic checks for vision, native tool support, and prompt caching.

Supported Cloud Providers

Light-Fabric supports all major LLM providers. Because the Provider trait is model-agnostic, the framework is compatible with the latest flagship releases as soon as they are available.

  • OpenAI: Native support for the GPT-5 series (5.4, mini, nano), the o4 reasoning models, and full legacy support for GPT-4o and GPT-4 Turbo.
  • Anthropic: Support for the Claude 4 generation, including Opus 4.7, Sonnet, and Haiku.
  • Google Gemini: Support for Gemini 3.1 Pro and Flash, leveraging Vertex AI or AI Studio for multi-modal and long-context tasks.
  • Azure OpenAI: Enterprise-grade OpenAI deployments with support for the latest model deployments.
  • AWS Bedrock: Access to the latest Claude and Titan models hosted on Amazon Web Services.
  • OpenRouter: Access to hundreds of open-source and proprietary models via a single unified API.
  • Telnyx: Support for models hosted on the Telnyx platform.
  • GLM (Zhipu AI): Support for the ChatGLM/GLM-5 series of models.

Local & Specialized Providers

  • Ollama: Seamless integration with local models running on your machine.
  • OpenAI-Compatible: A generic CompatibleProvider for any service implementing the OpenAI REST API.
  • GitHub Copilot: Integration with GitHub Copilot Chat for developer-centric workflows.

Meta-Providers (Orchestration)

These providers wrap other providers to add resilient or intelligent behavior:

  • ReliableProvider: Enhances any base provider with retries, exponential backoff, and automatic failover to fallback models.
  • RouterProvider: Dynamically routes requests to different models based on hints or input complexity.

CLI & Tooling Integrations

Light-Fabric includes specialized integrations for developer tools and terminal environments:

  • Claude Code CLI: Integration with Anthropic's Claude Code environment.
  • Gemini CLI: Terminal-based access to Google's Gemini models.
  • KiloCLI: Light-Fabric's native CLI integration for rapid testing and automation.

Key Capabilities

Providers can be queried for their support of advanced features:

  • Native Tool Calling: Efficiently generate structured function calls.
  • Vision: Process images alongside text prompts.
  • Prompt Caching: Leverage provider-side caching to reduce latency and costs for long contexts.