Module provider 

AI Provider abstraction layer.

Unified interface for multiple AI providers (OpenAI, Anthropic, Google, StepFun, Bedrock, etc.).

Architecture

• [types] — shared data types (Message, StreamChunk, etc.)
• [traits] — the Provider trait and ModelInfo
• [registry] — ProviderRegistry (name → provider map)
• [parse] — model-string parser ("openai/gpt-4o" → (provider, model))
• [init_vault] — Vault-based provider initialization
• [init_config] — TOML-config-based initialization
• [init_env] — environment-variable fallback
• [init_dispatch] / [init_dispatch_impl] — per-provider constructors

Provider implementations live in their own modules (openai, anthropic, bedrock/, etc.).

Modules

anthropic

Anthropic provider implementation using the Messages API

bedrock

Amazon Bedrock provider for the Converse API.

copilot

GitHub Copilot provider implementation using raw HTTP.

gemini_web

Gemini Web provider drives the Gemini chat UI’s undocumented BardChatUi endpoint using browser cookies stored in HashiCorp Vault.

glm5

GLM-5 FP8 provider for Vast.ai serverless deployments

google

Google Gemini provider implementation

limits

Canonical context-window limits for known LLM models.

local_cuda

Stub when CUDA is not compiled in.

metrics

Provider metrics wrapper

models

Model catalog from CodeTether API

moonshot

Moonshot AI provider implementation (direct API)

openai

OpenAI provider implementation

openai_codex

OpenAI Codex provider using ChatGPT Plus/Pro subscription via OAuth

openrouter

OpenRouter provider implementation using raw HTTP

retry

Provider HTTP retry logic.

stepfun

StepFun provider implementation (direct API, not via OpenRouter)

util

Provider-level re-exports of shared crate utilities.

vertex_anthropic

Vertex AI Anthropic provider implementation

vertex_glm

Vertex AI GLM provider implementation (MaaS endpoint)

zai

Z.AI provider implementation (direct API)

Structs

CompletionRequest

Request to generate a completion.

CompletionResponse

Response from a completion request.

EmbeddingRequest

Request to generate embeddings.

EmbeddingResponse

Response from an embedding request.

Message

A message in a conversation.

ModelInfo

Metadata about a model offered by a provider.

ProviderRegistry

Registry of available providers.

ToolDefinition

Schema-driven tool definition passed to the model.

Usage

Token usage statistics.

Enums

ContentPart

One content block within a Message.

FinishReason

Reason the model stopped generating.

Role

Participant role in a conversation.

StreamChunk

A streaming chunk produced by Provider::complete_stream.

Traits

Provider

Trait that all AI providers must implement.

Functions

parse_model_string

Parse a model string into (provider, model).