rig-provider

To add a new LLM provider to the Rig framework, follow the OpenAI reference implementation pattern to build provider structs, declare capabilities, implement trait bounds, and map API responses. This ensures your integration aligns with upstream APIs and framework standards.

WebAssembly compatibility in Rig providers requires using specific patterns to ensure the LLM API integration operates seamlessly within WASM environments. The implementation must pass quality gates that verify trait bounds, error types, and avoid non-compatible dependencies.

Error handling and telemetry for a custom LLM provider require robust error mapping alongside the addition of telemetry spans. These practices ensure that API failures are caught correctly and execution traces are observable during the agent's lifecycle.

Yes, Rig supports observing and controlling the agent execution lifecycle through prompt hooks. This mechanism allows developers to intercept and manage the flow of interactions with the large language model provider during runtime.

Before submitting a new Rig model provider, your code must pass quality gates checking trait bounds, error types, comments, documentation, and outstanding TODO items. Pre-submission checks also verify architectural alignment and correct message type conversions.

To ensure API alignment when integrating an LLM API into Rig, keep request and response fields strictly consistent with the upstream provider APIs. The implementation must accurately map message types and adhere to the framework's client type aliases.