Crate codex_convert_proxy 

Codex Convert Proxy Library

A library for converting between OpenAI Responses API (used by Codex) and Chat Completions API (used by Chinese LLM providers).

Overview

Codex 0.118+ only supports the Responses API, but Chinese LLM providers (GLM, Kimi, DeepSeek, MiniMax) only support the Chat Completions API. This library provides bidirectional conversion between these formats.

Features

• lib: Provider trait + implementations + conversion + types (for embedding the conversion logic in other Rust applications)
• server (default): Full Pingora-based proxy binary (implies lib, adds config parsing, the HTTP server, CLI, and logging)
• telemetry: OpenTelemetry tracing/metrics exporters

Usage (as library)

use codex_convert_proxy::{
    response_to_chat, chat_to_response, chat_to_response_with_context,
    chat_chunk_to_response_events, event_to_sse,
    Provider, GLMProvider, create_provider,
    StreamState, ResponseRequestContext, ResponseStreamEvent,
    types::response_api::{ResponseRequest, InputItemOrString},
    util::parse_sse,
};

Providers

Each Chinese LLM provider may have slightly different API requirements. Use the appropriate provider when converting:

• GLMProvider - For Zhipu AI GLM models
• KimiProvider - For Moonshot AI Kimi models
• DeepSeekProvider - For DeepSeek models
• MiniMaxProvider - For MiniMax models
• DefaultProvider - Generic fallback for any OpenAI-compatible provider (used automatically by create_provider when the provider name is not in the built-in registry)

Re-exports

pub use error::ConversionError;

pub use error::ProxyError;

pub use convert::chat_chunk_to_response_events;

pub use convert::chat_to_response;

pub use convert::chat_to_response_with_context;

pub use convert::event_to_sse;

pub use convert::response_to_chat;

pub use convert::ResponseRequestContext;

pub use convert::streaming::ResponseStreamEvent;

pub use convert::streaming::StreamState;

pub use stats::RequestRecord;

pub use stats::RequestStats;

pub use stats::StatsSummary;

pub use stats::TokenUsage;

pub use providers::create_provider;

pub use providers::DefaultProvider;

pub use providers::DeepSeekProvider;

pub use providers::GLMProvider;

pub use providers::KimiProvider;

pub use providers::MiniMaxProvider;

pub use providers::Provider;

pub use config::BackendConfig;

pub use config::BackendInfo;

pub use config::BackendRouter;

pub use config::ProxyConfig;

pub use proxy::CodexProxy;

pub use cli::Cli;

pub use cli::Commands;

pub use cli::StartArgs;

pub use types::*;

Modules

cli

CLI argument parsing module.

config

Configuration module for proxy settings and backend routing.

constants

项目级配置常量

convert

Conversion between Responses API and Chat API.

error

Error types for the library.

logger

Logging module based on tracing ecosystem.

providers

Provider implementations for Chinese LLM services.

proxy

Proxy module for pingora integration.

server

Pingora server startup and configuration.

stats

Statistics and metrics module.

types

Type definitions for both API formats.

util

Utility modules.