codex-proxy 3.1.0__tar.gz → 5.0.0__tar.gz

codex_proxy-5.0.0/.dockerignore

@@ -0,0 +1,14 @@
+__pycache__
+*.pyc
+*.pyo
+.git
+.github
+.venv
+.env
+tests/
+dist/
+build/
+*.egg-info
+.ruff_cache
+.mypy_cache
+.pytest_cache

codex_proxy-5.0.0/CHANGELOG.md

@@ -0,0 +1,185 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [5.0.0] - 2026-06-07
+
+### Added — v5 Gateway Platform
+
+**Database Layer**
+- Async SQLAlchemy database with 13 tables (users, api_keys, providers, provider_keys, models, routing_rules, request_logs, budgets, cost_alerts, plugin_registry, plugin_instances, sessions, _schema_version)
+- SQLite (default) or PostgreSQL (via asyncpg) backend
+- Version-based migrations system
+- 7 CRUD modules for all entities
+
+**Multi-Provider Support**
+- `[[providers]]` TOML array — configure multiple providers in one instance
+- Per-provider clients, adapters, key rotators
+- Model-to-provider resolution across all providers
+- `/models` endpoint aggregates models from all providers
+
+**JWT Authentication**
+- POST `/auth/login`, `/auth/signup`, `/auth/refresh`, GET `/auth/me`
+- Password hashing with bcrypt (SHA-256 fallback)
+- JWT access + refresh tokens (PyJWT with stdlib HMAC-SHA256 fallback)
+- Admin user auto-seeded on first startup
+- Budget endpoints: GET/PUT `/auth/budget`
+
+**Smart Router**
+- 4 routing strategies: `fallback` (config order), `cost` (cheapest), `latency` (fastest), `weighted` (load balanced)
+- Rolling latency tracker with per-provider health detection
+- Unhealthy providers automatically skipped
+- GET `/api/router/status` for detailed metrics
+
+**Cost Tracking**
+- Per-model pricing: input_price_per_million, output_price_per_million
+- 25+ built-in model prices (GLM, GPT, Claude, Gemini, DeepSeek, Llama, etc.)
+- Automatic cost calculation on every request (DB lookup → KNOWN_PRICING fallback → $0)
+- Prices auto-seeded to DB on startup (idempotent)
+- Cost aggregation analytics: GET `/api/stats`, `/api/usage`
+
+**Budget Enforcement**
+- Daily and monthly spend limits per user
+- Requests blocked with 429 when budget exceeded
+- Alert threshold notifications
+- Budget status and alerts via API
+
+**Web Dashboard**
+- Embedded HTML+CSS+JS dashboard at GET `/dashboard`
+- Dark theme, auto-refresh every 10 seconds
+- Stats cards (requests, success rate, uptime, total cost)
+- Cost breakdown table with visual bars
+- Provider cards with model tags
+- Router status with latency and health info
+
+**v5 Config Sections** (all disabled by default for v4 compat)
+- `[database]` — persistent storage settings
+- `[auth]` — JWT authentication settings
+- `[router]` — smart routing settings
+- `[dashboard]` — web dashboard settings
+
+**New Dependencies**
+- `sqlalchemy>=2.0`, `aiosqlite>=0.20` (core)
+- Optional: `asyncpg>=0.29` (postgres), `bcrypt>=4.0`, `PyJWT>=2.8`, `cryptography>=42.0` (enterprise)
+
+**Testing**
+- 270+ tests (up from 217): 20 DB tests, 33 auth/cost/router tests
+- All existing v4 tests pass unchanged — zero regressions
+
+### Changed
+- Description updated: "LLM Gateway Platform — multi-provider proxy with smart routing, cost analytics, and web dashboard"
+- `_log_request()` now calculates real costs and resolves provider_id from DB
+- `/status` endpoint includes v5 features, auth, and router status
+- Startup output shows all providers and v5 mode
+
+### Backward Compatibility
+- All v5 features **disabled by default** — v4 config works as-is with zero changes
+- No breaking changes to existing API endpoints or behavior
+
+## [4.0.0] - 2026-06-07
+
+### Fixed
+
+- **`__main__.py`**: `--port 0` and `--host ""` no longer silently ignored (`is not None` check)
+- **`store.py`**: Removed production `assert` that could silently crash with `-O` flag
+- **`compaction.py`**: Compaction notice now says "dropped" instead of misleading "summarized"
+- **`tui.py`**: Success rate now correct after first request (off-by-one fix)
+- **`circuit_breaker.py`**: `get_status()` now includes `last_failure_time` field
+- **`Dockerfile`**: Binds to `0.0.0.0` for container networking
+
+### Changed
+
+- Removed duplicate `_mask_key()` from `server.py` — single source in `key_rotation.py`
+- `KeyRotator` now exposes `key_count` property (no more `_keys` access)
+- `ResponseStore` now exposes `clear()` method (no more `_store` access)
+- `server.py` no longer accesses private `_keys` on `KeyRotator`
+- `tui.py` no longer accesses private `_store` on `ResponseStore`
+- `tui.py` no longer shadows `state_colors` variable name
+- Hardcoded `MAX_RETRIES`, `RETRY_DELAY`, httpx timeouts are now configurable via config
+- Example config URL corrected to `ZiryaNoov/codex-proxy`
+
+### Added
+
+- **CORS middleware** — configurable via `cors_origins` in `[server]` config
+- **Admin authentication** — optional Bearer token on `/reload` and `/status` endpoints
+- **Rate limiting** — per-client sliding window rate limiter (`[rate_limit]` config section)
+- **Request size limits** — configurable `max_request_body_bytes` (default 10MB)
+- **Configurable timeouts** — `connect_timeout`, `read_timeout`, `max_retries`, `retry_delay`
+- **Provider adapters** for Together AI and Fireworks AI (11 adapters total)
+- **`rate_limiter.py`** — new module with `RateLimiter` class
+- **217 tests** (up from 184) with new test files:
+  - `test_rate_limiter.py` — rate limiter unit tests
+  - `test_server_features.py` — rate limiting, admin auth, request size integration tests
+  - `test_edge_cases.py` — compaction edge cases, config TOML, provider adapter registry
+- `pyproject.toml` now includes `readme = "README.md"` for PyPI
+
+## [3.1.0] - 2026-06-01
+
+### Added
+- Dynamic, unified configuration and service reloading across HTTP and TUI (`reload_config_internal`).
+- Instant, non-blocking Unix/macOS keyboard input using `termios`/`tty` cbreak mode.
+
+## [2.0.1] - 2026-05-31
+
+### Fixed
+
+- **Compaction config threading**: `max_messages` and `keep_last` from config.toml
+  are now correctly passed to the compaction engine (were previously hardcoded)
+- **Circuit breaker type annotation**: `AppState.circuit_breaker` now correctly
+  typed as `CircuitBreaker | None`
+- **Circuit breaker coverage**: Now protects streaming HTTP and WebSocket paths
+  (was only protecting non-streaming HTTP)
+- **Config reload**: `/reload` now recreates circuit breaker with new config values
+- **Example config**: `codex-proxy --init` now generates `[circuit_breaker]` and
+  `[compaction]` sections
+
+### Added
+
+- `--print-config` now shows circuit_breaker and compaction settings
+- MIT `LICENSE` file
+
+## [2.0.0] - 2026-05-31
+
+### Added
+
+- **6 new provider adapters** for broader LLM ecosystem support:
+  - Anthropic — `x-api-key` header and `anthropic-version` handling
+  - Gemini — Google's OpenAI-compatible endpoint
+  - DeepSeek — stream_options stripping
+  - Mistral — OpenAI-compatible with stream_options handling
+  - Cohere — OpenAI-compatible endpoint
+  - NVIDIA — NIM OpenAI-compatible endpoint
+- **Circuit breaker** for upstream resilience — protects against failing providers
+  with configurable failure threshold, recovery timeout, and half-open state
+- **Context compaction** for long conversations — automatically trims message
+  history to fit within limits while preserving system prompts and recent context
+- **CI/CD pipeline** (GitHub Actions) — automated lint, test, and release workflows
+- **Docker Compose** support — one-command containerized deployment with health checks
+- **Pre-commit hooks** — Ruff linting and formatting on every commit
+- **112+ tests** covering translator, config, store, server, providers, circuit
+  breaker, and compaction modules
+- **CONTRIBUTING.md** — development setup, code style, PR process, and commit conventions
+- **CODE_OF_CONDUCT.md** — community guidelines
+- **Issue templates** and **PR template** for structured contributions
+
+## [1.0.0] - 2026-05-19
+
+### Added
+
+- Core **Responses API to Chat Completions** protocol translation
+- **Streaming SSE** support with real-time token delivery
+- **WebSocket** support with full Realtime API envelope handling
+- **Reasoning/thinking token** passthrough from supported models
+- **Tool calls** — full function calling support (definitions + results)
+- **Multi-turn conversations** via `previous_response_id` with in-memory response store
+- **Multi-provider** backend — switch between Z.AI, Groq, Together AI, OpenRouter,
+  Ollama, Fireworks, and more with a single config change
+- **Usage tracking** — captures token counts from streaming responses
+- **Auto-retry** — one retry on 5xx/transport errors for non-streaming requests
+- **TOML configuration** file with env var and CLI flag overrides
+- **FastAPI** server with `/responses`, `/models`, `/health`, `/status`, and `/reload` endpoints
+- **Dockerfile** for containerized deployment
+- **CLI** with `--init`, `--print-config`, `--port`, `--host`, and `--config` options

{codex_proxy-3.1.0 → codex_proxy-5.0.0}/CONTRIBUTING.md

@@ -54,7 +54,7 @@ pytest tests/test_translator.py -v
 pytest tests/ -v --tb=short
 ```
 
-The test suite uses **pytest** with 112+ tests covering the translator, config, store, server, providers, circuit breaker, and compaction modules.
+The test suite uses **pytest** with 217+ tests covering the translator, config, store, server, providers, circuit breaker, compaction, rate limiter, and admin auth modules.
 
 ## Project Structure
 
@@ -67,10 +67,15 @@ codex-proxy/
     translator.py        # Responses API <-> Chat Completions
     config.py            # TOML config loading
     store.py             # In-memory response store
-    providers.py         # Provider-specific adapters
+    providers.py         # Provider-specific adapters (11 providers)
     circuit_breaker.py   # Upstream resilience
     compaction.py        # Context compaction
-  tests/                 # Test suite
+    key_rotation.py      # Multi-key round-robin pool
+    plugins.py           # Hook-based plugin system
+    plugins_builtin.py   # Built-in plugins (LoggingPlugin)
+    rate_limiter.py      # Per-client sliding window rate limiter
+    tui.py               # Rich TUI dashboard
+  tests/                 # Test suite (217+ tests)
   .github/workflows/     # CI/CD pipelines
 ```
 
@@ -118,7 +123,7 @@ Common types: `feat`, `fix`, `refactor`, `test`, `docs`, `chore`, `ci`, `perf`
 
 ## Reporting Issues
 
-- Use [GitHub Issues](https://github.com/ZakPro/codex-proxy/issues) for bugs and feature requests
+- Use [GitHub Issues](https://github.com/ZiryaNoov/codex-proxy/issues) for bugs and feature requests
 - Include your Python version, OS, and relevant config (redact API keys)
 - For bugs, provide steps to reproduce and any error output
 

{codex_proxy-3.1.0 → codex_proxy-5.0.0}/Dockerfile

@@ -6,4 +6,5 @@ RUN pip install --no-cache-dir .
 EXPOSE 4242
 ENV CODEX_PROXY_API_KEY=""
 ENV CODEX_PROXY_BASE_URL=""
-CMD ["codex-proxy"]
+ENV CODEX_PROXY_HOST="0.0.0.0"
+CMD ["codex-proxy", "--host", "0.0.0.0"]