Add OpenAI compatibility support and improve resource cleanup

- Introduced OpenAI compatibility configurations for external providers, enabling model alias routing via the OpenAI API format. - Enhanced provider logic in `GetProviderName` to handle OpenAI aliases and added new helper functions for compatibility checks. - Updated API handlers and client initialization to support OpenAI compatibility models. - Improved resource cleanup across clients by closing response bodies and streams using deferred functions.

Luis Pater committed 1 year agov4.1.0

Commit ed8873f1 parent 9102ff018 files · +671 −72Browse files

Unified · Split

18 files

README.md +80 −29
README_CN.md +70 −22
config.example.yaml +26 −9
internal/api/handlers/claude/code_handlers.go +1 −1
internal/api/handlers/gemini/gemini-cli_handlers.go +1 −1
internal/api/handlers/gemini/gemini_handlers.go +1 −1
internal/api/handlers/handlers.go +3 −1
internal/api/handlers/openai/openai_handlers.go +1 −1
internal/client/claude_client.go +4 −0
internal/client/codex_client.go +4 −0
internal/client/gemini-cli_client.go +6 −0
internal/client/gemini_client.go +4 −0
internal/client/openai-compatibility_client.go +360 −0
internal/client/qwen_client.go +4 −0
internal/cmd/run.go +12 −0
internal/config/config.go +29 −0
internal/constant/constant.go +6 −5
internal/util/provider.go +59 −2

README.md+80−29

@@ -6,9 +6,9 @@ A proxy server that provides OpenAI/Gemini/Claude compatible API interfaces for · 5 unmodified lines
6	6
7	7	It now also supports OpenAI Codex (GPT models) and Claude Code via OAuth.
8	8
9		−so you can use local or multi‑account CLI access with OpenAI‑compatible clients and SDKs.
	9	+So you can use local or multi-account CLI access with OpenAI-compatible clients and SDKs.
10	10
11		−Now, We added the first Chinese provider: [Qwen Code](https://github.com/QwenLM/qwen-code).
	11	+The first Chinese provider has now been added: [Qwen Code](https://github.com/QwenLM/qwen-code).
12	12
13	13	## Features
14	14
@@ -19,12 +19,13 @@ Now, We added the first Chinese provider: [Qwen Code](https://github.com/QwenLM/ · 4 unmodified lines
19	19	- Streaming and non-streaming responses
20	20	- Function calling/tools support
21	21	- Multimodal input support (text and images)
22		−- Multiple accounts with round‑robin load balancing (Gemini, OpenAI, Claude and Qwen)
	22	+- Multiple accounts with round-robin load balancing (Gemini, OpenAI, Claude and Qwen)
23	23	- Simple CLI authentication flows (Gemini, OpenAI, Claude and Qwen)
24	24	- Generative Language API Key support
25		−- Gemini CLI multi‑account load balancing
26		−- Claude Code multi‑account load balancing
27		−- Qwen Code multi‑account load balancing
	25	+- Gemini CLI multi-account load balancing
	26	+- Claude Code multi-account load balancing
	27	+- Qwen Code multi-account load balancing
	28	+- OpenAI-compatible upstream providers via config (e.g., OpenRouter)
28	29
29	30	## Installation
30	31
@@ -59,13 +60,13 @@ You can authenticate for Gemini, OpenAI, and/or Claude. All can coexist in the s · 28 unmodified lines
59	60	```bash
60	61	./cli-proxy-api --login
61	62	```
62		− If you are an old gemini code user, you may need to specify a project ID:
	63	+ If you are an existing Gemini Code user, you may need to specify a project ID:
63	64	```bash
64	65	./cli-proxy-api --login --project_id <your_project_id>
65	66	```
66	67	The local OAuth callback uses port `8085`.
67	68
68		− Options: add `--no-browser` to print the login URL instead of opening a browser. The local OAuth callback uses port `1455`.
	69	+ Options: add `--no-browser` to print the login URL instead of opening a browser. The local OAuth callback uses port `8085`.
69	70
70	71	- OpenAI (Codex/GPT via OAuth):
71	72	```bash
@@ -126,7 +127,7 @@ Request body example: · 54 unmodified lines
126	127	```
127	128
128	129	Notes:
129		−- Use a `gemini-` model for Gemini (e.g., `gemini-2.5-pro`), a `gpt-` model for OpenAI (e.g., `gpt-5`), a `claude-` model for Claude (e.g., `claude-3-5-sonnet-20241022`), or a `qwen-` model for Qwen (e.g., `qwen3-coder-plus`). The proxy will route to the correct provider automatically.
	130	+- Use a `gemini-` model for Gemini (e.g., "gemini-2.5-pro"), a `gpt-` model for OpenAI (e.g., "gpt-5"), a `claude-` model for Claude (e.g., "claude-3-5-sonnet-20241022"), or a `qwen-` model for Qwen (e.g., "qwen3-coder-plus"). The proxy will route to the correct provider automatically.
130	131
131	132	#### Claude Messages (SSE-compatible)
132	133
@@ -226,7 +227,7 @@ console.log(await claudeResponse.json()); · 93 unmodified lines
226	227	- claude-3-5-haiku-20241022
227	228	- qwen3-coder-plus
228	229	- qwen3-coder-flash
229		−- Gemini models auto‑switch to preview variants when needed
	230	+- Gemini models auto-switch to preview variants when needed
230	231
231	232	## Configuration
232	233
@@ -238,21 +239,28 @@ The server uses a YAML configuration file (`config.yaml`) located in the project · 5 unmodified lines
238	239
239	240	### Configuration Options
240	241
241		−\| Parameter \| Type \| Default \| Description \|
242		−\|---------------------------------------\|----------\|--------------------\|-----------------------------------------------------------------------------------------------\|
243		−\| `port` \| integer \| 8317 \| The port number on which the server will listen \|
244		−\| `auth-dir` \| string \| "~/.cli-proxy-api" \| Directory where authentication tokens are stored. Supports using `~` for home directory \|
245		−\| `proxy-url` \| string \| "" \| Proxy url, support socks5/http/https protocol, example: socks5://user:pass@192.168.1.1:1080/ \|
246		−\| `request-retry` \| integer \| 0 \| Request retry times, if the response http code is 403, 408, 500, 502, 503, 504, it will retry \|
247		−\| `quota-exceeded` \| object \| {} \| Configuration for handling quota exceeded \|
248		−\| `quota-exceeded.switch-project` \| boolean \| true \| Whether to automatically switch to another project when a quota is exceeded \|
249		−\| `quota-exceeded.switch-preview-model` \| boolean \| true \| Whether to automatically switch to a preview model when a quota is exceeded \|
250		−\| `debug` \| boolean \| false \| Enable debug mode for verbose logging \|
251		−\| `api-keys` \| string[] \| [] \| List of API keys that can be used to authenticate requests \|
252		−\| `generative-language-api-key` \| string[] \| [] \| List of Generative Language API keys \|
253		−\| `claude-api-key` \| object \| {} \| List of Claude API keys \|
254		−\| `claude-api-key.api-key` \| string \| "" \| Claude API key \|
255		−\| `claude-api-key.base-url` \| string \| "" \| Custom Claude API endpoint, if you use the third party API endpoint \|
	242	+\| Parameter \| Type \| Default \| Description \|
	243	+\|-----------------------------------------\|----------\|--------------------\|---------------------------------------------------------------------------------------------------------\|
	244	+\| `port` \| integer \| 8317 \| The port number on which the server will listen. \|
	245	+\| `auth-dir` \| string \| "~/.cli-proxy-api" \| Directory where authentication tokens are stored. Supports using `~` for the home directory. \|
	246	+\| `proxy-url` \| string \| "" \| Proxy URL. Supports socks5/http/https protocols. Example: socks5://user:pass@192.168.1.1:1080/ \|
	247	+\| `request-retry` \| integer \| 0 \| Number of times to retry a request. Retries will occur if the HTTP response code is 403, 408, 500, 502, 503, or 504. \|
	248	+\| `quota-exceeded` \| object \| {} \| Configuration for handling quota exceeded. \|
	249	+\| `quota-exceeded.switch-project` \| boolean \| true \| Whether to automatically switch to another project when a quota is exceeded. \|
	250	+\| `quota-exceeded.switch-preview-model` \| boolean \| true \| Whether to automatically switch to a preview model when a quota is exceeded. \|
	251	+\| `debug` \| boolean \| false \| Enable debug mode for verbose logging. \|
	252	+\| `api-keys` \| string[] \| [] \| List of API keys that can be used to authenticate requests. \|
	253	+\| `generative-language-api-key` \| string[] \| [] \| List of Generative Language API keys. \|
	254	+\| `claude-api-key` \| object \| {} \| List of Claude API keys. \|
	255	+\| `claude-api-key.api-key` \| string \| "" \| Claude API key. \|
	256	+\| `claude-api-key.base-url` \| string \| "" \| Custom Claude API endpoint, if you use a third-party API endpoint. \|
	257	+\| `openai-compatibility` \| object[] \| [] \| Upstream OpenAI-compatible providers configuration (name, base-url, api-keys, models). \|
	258	+\| `openai-compatibility.*.name` \| string \| "" \| The name of the provider. It will be used in the user agent and other places. \|
	259	+\| `openai-compatibility.*.base-url` \| string \| "" \| The base URL of the provider. \|
	260	+\| `openai-compatibility.*.api-keys` \| string[] \| [] \| The API keys for the provider. Add multiple keys if needed. Omit if unauthenticated access is allowed. \|
	261	+\| `openai-compatibility.*.models` \| object[] \| [] \| The actual model name. \|
	262	+\| `openai-compatibility..models..name` \| string \| "" \| The models supported by the provider. \|
	263	+\| `openai-compatibility..models..alias` \| string \| "" \| The alias used in the API. \|
256	264
257	265	### Example Configuration File
258	266
@@ -266,10 +274,10 @@ auth-dir: "~/.cli-proxy-api" · 7 unmodified lines
266	274	# Enable debug logging
267	275	debug: false
268	276
269		−# Proxy url, support socks5/http/https protocol, example: socks5://user:pass@192.168.1.1:1080/
	277	+# Proxy URL. Supports socks5/http/https protocols. Example: socks5://user:pass@192.168.1.1:1080/
270	278	proxy-url: ""
271	279
272		−# Request retry times, if the response http code is 403, 408, 500, 502, 503, 504, it will retry
	280	+# Number of times to retry a request. Retries will occur if the HTTP response code is 403, 408, 500, 502, 503, or 504.
273	281	request-retry: 3
274	282
275	283	# Quota exceeded behavior
@@ -294,8 +302,51 @@ claude-api-key: · 18 unmodified lines
294	302	- api-key: "sk-atSM..." # use the official claude API key, no need to set the base url
295	303	- api-key: "sk-atSM..."
296	304	base-url: "https://www.example.com" # use the custom claude API endpoint
	305	+
	306	+# OpenAI compatibility providers
	307	+openai-compatibility:
	308	+ - name: "openrouter" # The name of the provider; it will be used in the user agent and other places.
	309	+ base-url: "https://openrouter.ai/api/v1" # The base URL of the provider.
	310	+ api-keys: # The API keys for the provider. Add multiple keys if needed. Omit if unauthenticated access is allowed.
	311	+ - "sk-or-v1-...b780"
	312	+ - "sk-or-v1-...b781"
	313	+ models: # The models supported by the provider.
	314	+ - name: "moonshotai/kimi-k2:free" # The actual model name.
	315	+ alias: "kimi-k2" # The alias used in the API.
	316	+```
	317	+
	318	+### OpenAI Compatibility Providers
	319	+
	320	+Configure upstream OpenAI-compatible providers (e.g., OpenRouter) via `openai-compatibility`.
	321	+
	322	+- name: provider identifier used internally
	323	+- base-url: provider base URL
	324	+- api-keys: optional list of API keys (omit if provider allows unauthenticated requests)
	325	+- models: list of mappings from upstream model `name` to local `alias`
	326	+
	327	+Example:
	328	+
	329	+```yaml
	330	+openai-compatibility:
	331	+ - name: "openrouter"
	332	+ base-url: "https://openrouter.ai/api/v1"
	333	+ api-keys:
	334	+ - "sk-or-v1-...b780"
	335	+ - "sk-or-v1-...b781"
	336	+ models:
	337	+ - name: "moonshotai/kimi-k2:free"
	338	+ alias: "kimi-k2"
297	339	```
298	340
	341	+Usage:
	342	+
	343	+Call OpenAI's endpoint `/v1/chat/completions` with `model` set to the alias (e.g., `kimi-k2`). The proxy routes to the configured provider/model automatically.
	344	+
	345	+Also, you may call Claude's endpoint `/v1/messages`, Gemini's `/v1beta/models/model-name:streamGenerateContent` or `/v1beta/models/model-name:generateContent`.
	346	+
	347	+And you can always use Gemini CLI with `CODE_ASSIST_ENDPOINT` set to `http://127.0.0.1:8317` for these OpenAI-compatible provider's models.
	348	+
	349	+
299	350	### Authentication Directory
300	351
301	352	The `auth-dir` parameter specifies where authentication tokens are stored. When you run the login command, the application will create JSON files in this directory containing the authentication tokens for your Google accounts. Multiple accounts can be used for load balancing.
@@ -327,8 +378,8 @@ export CODE_ASSIST_ENDPOINT="http://127.0.0.1:8317" · 25 unmodified lines
327	378	The server will relay the `loadCodeAssist`, `onboardUser`, and `countTokens` requests. And automatically load balance the text generation requests between the multiple accounts.
328	379
329	380	> [!NOTE]
330		−> This feature only allows local access because I couldn't find a way to authenticate the requests.
331		−> I hardcoded `127.0.0.1` into the load balancing.
	381	+> This feature only allows local access because there is currently no way to authenticate the requests.
	382	+> 127.0.0.1 is hardcoded for load balancing.
332	383
333	384	## Claude Code with multiple account load balancing
334	385

README_CN.md+70−22

@@ -6,9 +6,9 @@ · 5 unmodified lines
6	6
7	7	现已支持通过 OAuth 登录接入 OpenAI Codex（GPT 系列）和 Claude Code。
8	8
9		−可与本地或多账户方式配合，使用任何 OpenAI 兼容的客户端与 SDK。
	9	+您可以使用本地或多账户的CLI方式，通过任何与OpenAI兼容的客户端和SDK进行访问。
10	10
11		−现在，我们添加了第一个中国提供商：[Qwen Code](https://github.com/QwenLM/qwen-code)。
	11	+现已新增首个中国提供商：[Qwen Code](https://github.com/QwenLM/qwen-code)。
12	12
13	13	## 功能特性
14	14
@@ -25,6 +25,7 @@ · 10 unmodified lines
25	25	- 支持 Gemini CLI 多账户轮询
26	26	- 支持 Claude Code 多账户轮询
27	27	- 支持 Qwen Code 多账户轮询
	28	+- 通过配置接入上游 OpenAI 兼容提供商（例如 OpenRouter）
28	29
29	30	## 安装
30	31
@@ -59,12 +60,14 @@ · 28 unmodified lines
59	60	```bash
60	61	./cli-proxy-api --login
61	62	```
62		− 如果您是旧版 gemini code 用户，可能需要指定项目 ID：
	63	+ 如果您是现有的 Gemini Code 用户，可能需要指定一个项目ID：
63	64	```bash
64	65	./cli-proxy-api --login --project_id <your_project_id>
65	66	```
66	67	本地 OAuth 回调端口为 `8085`。
67	68
	69	+ 选项：加上 `--no-browser` 可打印登录地址而不自动打开浏览器。本地 OAuth 回调端口为 `8085`。
	70	+
68	71	- OpenAI（Codex/GPT，OAuth）：
69	72	```bash
70	73	./cli-proxy-api --codex-login
@@ -123,7 +126,7 @@ POST http://localhost:8317/v1/chat/completions · 52 unmodified lines
123	126	```
124	127
125	128	说明：
126		−- 使用 `gemini-` 模型（如 `gemini-2.5-pro`）走 Gemini，使用 `gpt-` 模型（如 `gpt-5`）走 OpenAI，使用 `claude-` 模型（如 `claude-3-5-sonnet-20241022`）走 Claude，使用 `qwen-` 模型（如 `qwen3-coder-plus`）走 Qwen，服务会自动路由到对应提供商。
	129	+- 使用 "gemini-" 模型（例如 "gemini-2.5-pro"）来调用 Gemini，使用 "gpt-" 模型（例如 "gpt-5"）来调用 OpenAI，使用 "claude-" 模型（例如 "claude-3-5-sonnet-20241022"）来调用 Claude，或者使用 "qwen-" 模型（例如 "qwen3-coder-plus"）来调用 Qwen。代理服务会自动将请求路由到相应的提供商。
127	130
128	131	#### Claude 消息（SSE 兼容）
129	132
@@ -235,21 +238,28 @@ console.log(await claudeResponse.json()); · 105 unmodified lines
235	238
236	239	### 配置选项
237	240
238		−\| 参数 \| 类型 \| 默认值 \| 描述 \|
239		−\|---------------------------------------\|----------\|--------------------\|------------------------------------------------------------------------\|
240		−\| `port` \| integer \| 8317 \| 服务器监听的端口号 \|
241		−\| `auth-dir` \| string \| "~/.cli-proxy-api" \| 存储身份验证令牌的目录。支持使用 `~` 表示主目录 \|
242		−\| `proxy-url` \| string \| "" \| 代理 URL，支持 socks5/http/https 协议，示例：socks5://user:pass@192.168.1.1:1080/ \|
243		−\| `request-retry` \| integer \| 0 \| 请求失败重试次数，如果响应的http代码为403、408、500、502、503、504则会自动重试 \|
244		−\| `quota-exceeded` \| object \| {} \| 处理配额超限的配置 \|
245		−\| `quota-exceeded.switch-project` \| boolean \| true \| 当配额超限时是否自动切换到另一个项目 \|
246		−\| `quota-exceeded.switch-preview-model` \| boolean \| true \| 当配额超限时是否自动切换到预览模型 \|
247		−\| `debug` \| boolean \| false \| 启用调试模式以进行详细日志记录 \|
248		−\| `api-keys` \| string[] \| [] \| 可用于验证请求的 API 密钥列表 \|
249		−\| `generative-language-api-key` \| string[] \| [] \| 生成式语言 API 密钥列表 \|
250		−\| `claude-api-key` \| object \| {} \| Claude API 密钥列表 \|
251		−\| `claude-api-key.api-key` \| string \| "" \| Claude API 密钥 \|
252		−\| `claude-api-key.base-url` \| string \| "" \| 自定义 Claude API 端点（如果你使用的是第三方 Claude API 端点） \|
	241	+\| 参数 \| 类型 \| 默认值 \| 描述 \|
	242	+\|---------------------------------------\|----------\|--------------------\|---------------------------------------------------------------------------------------------\|
	243	+\| `port` \| integer \| 8317 \| 服务器将监听的端口号。 \|
	244	+\| `auth-dir` \| string \| "~/.cli-proxy-api" \| 存储身份验证令牌的目录。支持使用 `~` 来表示主目录。 \|
	245	+\| `proxy-url` \| string \| "" \| 代理URL。支持socks5/http/https协议。例如：socks5://user:pass@192.168.1.1:1080/ \|
	246	+\| `request-retry` \| integer \| 0 \| 请求重试次数。如果HTTP响应码为403、408、500、502、503或504，将会触发重试。 \|
	247	+\| `quota-exceeded` \| object \| {} \| 用于处理配额超限的配置。 \|
	248	+\| `quota-exceeded.switch-project` \| boolean \| true \| 当配额超限时，是否自动切换到另一个项目。 \|
	249	+\| `quota-exceeded.switch-preview-model` \| boolean \| true \| 当配额超限时，是否自动切换到预览模型。 \|
	250	+\| `debug` \| boolean \| false \| 启用调试模式以获取详细日志。 \|
	251	+\| `api-keys` \| string[] \| [] \| 可用于验证请求的API密钥列表。 \|
	252	+\| `generative-language-api-key` \| string[] \| [] \| 生成式语言API密钥列表。 \|
	253	+\| `claude-api-key` \| object \| {} \| Claude API密钥列表。 \|
	254	+\| `claude-api-key.api-key` \| string \| "" \| Claude API密钥。 \|
	255	+\| `claude-api-key.base-url` \| string \| "" \| 自定义的Claude API端点，如果您使用第三方的API端点。 \|
	256	+\| `openai-compatibility` \| object[] \| [] \| 上游OpenAI兼容提供商的配置（名称、基础URL、API密钥、模型）。 \|
	257	+\| `openai-compatibility.*.name` \| string \| "" \| 提供商的名称。它将被用于用户代理（User Agent）和其他地方。 \|
	258	+\| `openai-compatibility.*.base-url` \| string \| "" \| 提供商的基础URL。 \|
	259	+\| `openai-compatibility.*.api-keys` \| string[] \| [] \| 提供商的API密钥。如果需要，可以添加多个密钥。如果允许未经身份验证的访问，则可以省略。 \|
	260	+\| `openai-compatibility.*.models` \| object[] \| [] \| 实际的模型名称。 \|
	261	+\| `openai-compatibility..models..name` \| string \| "" \| 提供商支持的模型。 \|
	262	+\| `openai-compatibility..models..alias` \| string \| "" \| 在API中使用的别名。 \|
253	263
254	264	### 配置文件示例
255	265
@@ -263,10 +273,10 @@ auth-dir: "~/.cli-proxy-api" · 7 unmodified lines
263	273	# 启用调试日志
264	274	debug: false
265	275
266		−# 代理 URL，支持 socks5/http/https 协议，示例：socks5://user:pass@192.168.1.1:1080/
	276	+# 代理URL。支持socks5/http/https协议。例如：socks5://user:pass@192.168.1.1:1080/
267	277	proxy-url: ""
268	278
269		−# 请求失败重试次数，如果响应的http代码为403、408、500、502、503、504则会自动重试
	279	+# 请求重试次数。如果HTTP响应码为403、408、500、502、503或504，将会触发重试。
270	280	request-retry: 3
271	281
272	282
@@ -292,8 +302,46 @@ claude-api-key: · 19 unmodified lines
292	302	- api-key: "sk-atSM..." # use the official claude API key, no need to set the base url
293	303	- api-key: "sk-atSM..."
294	304	base-url: "https://www.example.com" # use the custom claude API endpoint
	305	+
	306	+# OpenAI 兼容提供商
	307	+openai-compatibility:
	308	+ - name: "openrouter" # 提供商的名称；它将被用于用户代理和其它地方。
	309	+ base-url: "https://openrouter.ai/api/v1" # 提供商的基础URL。
	310	+ api-keys: # 提供商的API密钥。如果需要，可以添加多个密钥。如果允许未经身份验证的访问，则可以省略。
	311	+ - "sk-or-v1-...b780"
	312	+ - "sk-or-v1-...b781"
	313	+ models: # 提供商支持的模型。
	314	+ - name: "moonshotai/kimi-k2:free" # 实际的模型名称。
	315	+ alias: "kimi-k2" # 在API中使用的别名。
295	316	```
296	317
	318	+### OpenAI 兼容上游提供商
	319	+
	320	+通过 `openai-compatibility` 配置上游 OpenAI 兼容提供商（例如 OpenRouter）。
	321	+
	322	+- name：内部识别名
	323	+- base-url：提供商基础地址
	324	+- api-keys：可选，多密钥轮询（若提供商支持无鉴权可省略）
	325	+- models：将上游模型 `name` 映射为本地可用 `alias`
	326	+
	327	+示例：
	328	+
	329	+```yaml
	330	+openai-compatibility:
	331	+ - name: "openrouter"
	332	+ base-url: "https://openrouter.ai/api/v1"
	333	+ api-keys:
	334	+ - "sk-or-v1-...b780"
	335	+ - "sk-or-v1-...b781"
	336	+ models:
	337	+ - name: "moonshotai/kimi-k2:free"
	338	+ alias: "kimi-k2"
	339	+```
	340	+
	341	+使用方式：在 `/v1/chat/completions` 中将 `model` 设为别名（如 `kimi-k2`），代理将自动路由到对应提供商与模型。
	342	+
	343	+并且，对于这些与OpenAI兼容的提供商模型，您始终可以通过将CODE_ASSIST_ENDPOINT设置为 http://127.0.0.1:8317 来使用Gemini CLI。
	344	+
297	345	### 身份验证目录
298	346
299	347	`auth-dir` 参数指定身份验证令牌的存储位置。当您运行登录命令时，应用程序将在此目录中创建包含 Google 账户身份验证令牌的 JSON 文件。多个账户可用于轮询。
@@ -325,7 +373,7 @@ export CODE_ASSIST_ENDPOINT="http://127.0.0.1:8317" · 25 unmodified lines
325	373	服务器将中继 `loadCodeAssist`、`onboardUser` 和 `countTokens` 请求。并自动在多个账户之间轮询文本生成请求。
326	374
327	375	> [!NOTE]
328		−> 此功能仅允许本地访问，因为找不到一个可以验证请求的方法。
	376	+> 此功能仅允许本地访问，因为找不到一个可以验证请求的方法。
329	377	> 所以只能强制只有 `127.0.0.1` 可以访问。
330	378
331	379	## Claude Code 的使用方法

config.example.yaml+26−9

@@ -1,23 +1,29 @@
1		−# Server configuration
	1	+# Server port
2	2	port: 8317
	3	+
	4	+# Authentication directory (supports ~ for home directory)
3	5	auth-dir: "~/.cli-proxy-api"
4		−debug: true
	6	+
	7	+# Enable debug logging
	8	+debug: false
	9	+
	10	+# Proxy URL. Supports socks5/http/https protocols. Example: socks5://user:pass@192.168.1.1:1080/
5	11	proxy-url: ""
6	12
7		−# Request retry times, if the response http code is 403, 408, 500, 502, 503, 504, it will retry
	13	+# Number of times to retry a request. Retries will occur if the HTTP response code is 403, 408, 500, 502, 503, or 504.
8	14	request-retry: 3
9	15
10	16	# Quota exceeded behavior
11	17	quota-exceeded:
12		− switch-project: true
13		− switch-preview-model: true
	18	+ switch-project: true # Whether to automatically switch to another project when a quota is exceeded
	19	+ switch-preview-model: true # Whether to automatically switch to a preview model when a quota is exceeded
14	20
15		−# API keys for client authentication
	21	+# API keys for authentication
16	22	api-keys:
17		− - "12345"
18		− - "23456"
	23	+ - "your-api-key-1"
	24	+ - "your-api-key-2"
19	25
20		−# Generative language API keys
	26	+# API keys for official Generative Language API
21	27	generative-language-api-key:
22	28	- "AIzaSy...01"
23	29	- "AIzaSy...02"
@@ -29,3 +35,14 @@ claude-api-key: · 5 unmodified lines
29	35	- api-key: "sk-atSM..." # use the official claude API key, no need to set the base url
30	36	- api-key: "sk-atSM..."
31	37	base-url: "https://www.example.com" # use the custom claude API endpoint
	38	+
	39	+# OpenAI compatibility providers
	40	+openai-compatibility:
	41	+ - name: "openrouter" # The name of the provider; it will be used in the user agent and other places.
	42	+ base-url: "https://openrouter.ai/api/v1" # The base URL of the provider.
	43	+ api-keys: # The API keys for the provider. Add multiple keys if needed. Omit if unauthenticated access is allowed.
	44	+ - "sk-or-v1-...b780"
	45	+ - "sk-or-v1-...b781"
	46	+ models: # The models supported by the provider.
	47	+ - name: "moonshotai/kimi-k2:free" # The actual model name.
	48	+ alias: "kimi-k2" # The alias used in the API.

internal/api/handlers/claude/code_handlers.go+1−1

@@ -148,7 +148,7 @@ outLoop: · 147 unmodified lines
148	148	// Detects when the HTTP client has disconnected and cleans up resources
149	149	case <-c.Request.Context().Done():
150	150	if c.Request.Context().Err().Error() == "context canceled" {
151		− log.Debugf("GeminiClient disconnected: %v", c.Request.Context().Err())
	151	+ log.Debugf("claude client disconnected: %v", c.Request.Context().Err())
152	152	cliCancel() // Cancel the backend request to prevent resource leaks
153	153	return
154	154	}

internal/api/handlers/gemini/gemini-cli_handlers.go+1−1

@@ -188,7 +188,7 @@ outLoop: · 187 unmodified lines
188	188	// Handle client disconnection.
189	189	case <-c.Request.Context().Done():
190	190	if c.Request.Context().Err().Error() == "context canceled" {
191		− log.Debugf("Client disconnected: %v", c.Request.Context().Err())
	191	+ log.Debugf("gemini cli client disconnected: %v", c.Request.Context().Err())
192	192	cliCancel() // Cancel the backend request.
193	193	return
194	194	}

internal/api/handlers/gemini/gemini_handlers.go+1−1

@@ -290,7 +290,7 @@ outLoop: · 289 unmodified lines
290	290	// Handle client disconnection.
291	291	case <-c.Request.Context().Done():
292	292	if c.Request.Context().Err().Error() == "context canceled" {
293		− log.Debugf("GeminiClient disconnected: %v", c.Request.Context().Err())
	293	+ log.Debugf("gemini client disconnected: %v", c.Request.Context().Err())
294	294	cliCancel() // Cancel the backend request.
295	295	return
296	296	}

internal/api/handlers/handlers.go+3−1

@@ -136,6 +136,8 @@ func (h *BaseAPIHandler) GetClient(modelName string, isGenerateContent ...bool) · 135 unmodified lines
136	136	log.Debugf("Claude Model %s is quota exceeded for account %s", modelName, cliClient.GetEmail())
137	137	} else if cliClient.Provider() == "qwen" {
138	138	log.Debugf("Qwen Model %s is quota exceeded for account %s", modelName, cliClient.GetEmail())
	139	+ } else if cliClient.Type() == "openai-compatibility" {
	140	+ log.Debugf("OpenAI Compatibility Model %s is quota exceeded for provider %s", modelName, cliClient.Provider())
139	141	}
140	142	cliClient = nil
141	143	continue
@@ -145,7 +147,7 @@ func (h *BaseAPIHandler) GetClient(modelName string, isGenerateContent ...bool) · 3 unmodified lines
145	147	}
146	148
147	149	if len(reorderedClients) == 0 {
148		− if util.GetProviderName(modelName) == "claude" {
	150	+ if util.GetProviderName(modelName, h.Cfg) == "claude" {
149	151	// log.Debugf("Claude Model %s is quota exceeded for all accounts", modelName)
150	152	return nil, &interfaces.ErrorMessage{StatusCode: 429, Error: fmt.Errorf(`{"type":"error","error":{"type":"rate_limit_error","message":"This request would exceed your account's rate limit. Please try again later."}}`)}
151	153	}

internal/api/handlers/openai/openai_handlers.go+1−1

@@ -278,7 +278,7 @@ outLoop: · 277 unmodified lines
278	278	// Handle client disconnection.
279	279	case <-c.Request.Context().Done():
280	280	if c.Request.Context().Err().Error() == "context canceled" {
281		− log.Debugf("Client disconnected: %v", c.Request.Context().Err())
	281	+ log.Debugf("qwen client disconnected: %v", c.Request.Context().Err())
282	282	cliCancel() // Cancel the backend request.
283	283	return
284	284	}

internal/client/claude_client.go+4−0

@@ -183,6 +183,7 @@ func (c *ClaudeClient) SendRawMessage(ctx context.Context, modelName string, raw · 182 unmodified lines
183	183	return nil, &interfaces.ErrorMessage{StatusCode: 500, Error: errReadAll}
184	184	}
185	185
	186	+ _ = respBody.Close()
186	187	c.AddAPIResponseData(ctx, bodyBytes)
187	188
188	189	var param any
@@ -238,6 +239,9 @@ func (c *ClaudeClient) SendRawMessageStream(ctx context.Context, modelName strin · 49 unmodified lines
238	239	return
239	240	}
240	241	delete(c.modelQuotaExceeded, modelName)
	242	+ defer func() {
	243	+ _ = stream.Close()
	244	+ }()
241	245
242	246	scanner := bufio.NewScanner(stream)
243	247	buffer := make([]byte, 10240*1024)

internal/client/codex_client.go+4−0

@@ -132,6 +132,7 @@ func (c *CodexClient) SendRawMessage(ctx context.Context, modelName string, rawJ · 131 unmodified lines
132	132	return nil, &interfaces.ErrorMessage{StatusCode: 500, Error: errReadAll}
133	133	}
134	134
	135	+ _ = respBody.Close()
135	136	c.AddAPIResponseData(ctx, bodyBytes)
136	137
137	138	var param any
@@ -188,6 +189,9 @@ func (c *CodexClient) SendRawMessageStream(ctx context.Context, modelName string · 50 unmodified lines
188	189	return
189	190	}
190	191	delete(c.modelQuotaExceeded, modelName)
	192	+ defer func() {
	193	+ _ = stream.Close()
	194	+ }()
191	195
192	196	scanner := bufio.NewScanner(stream)
193	197	buffer := make([]byte, 10240*1024)

internal/client/gemini-cli_client.go+6−0

@@ -497,6 +497,7 @@ func (c *GeminiCLIClient) SendRawMessage(ctx context.Context, modelName string, · 496 unmodified lines
497	497	return nil, &interfaces.ErrorMessage{StatusCode: 500, Error: errReadAll}
498	498	}
499	499
	500	+ _ = respBody.Close()
500	501	c.AddAPIResponseData(ctx, bodyBytes)
501	502
502	503	newCtx := context.WithValue(ctx, "alt", alt)
@@ -571,6 +572,11 @@ func (c *GeminiCLIClient) SendRawMessageStream(ctx context.Context, modelName st · 68 unmodified lines
571	572	delete(c.modelQuotaExceeded, modelName)
572	573	break
573	574	}
	575	+ defer func() {
	576	+ if stream != nil {
	577	+ _ = stream.Close()
	578	+ }
	579	+ }()
574	580
575	581	newCtx := context.WithValue(ctx, "alt", alt)
576	582	var param any

internal/client/gemini_client.go+4−0

@@ -249,6 +249,7 @@ func (c *GeminiClient) SendRawMessage(ctx context.Context, modelName string, raw · 248 unmodified lines
249	249	return nil, &interfaces.ErrorMessage{StatusCode: 500, Error: errReadAll}
250	250	}
251	251
	252	+ _ = respBody.Close()
252	253	c.AddAPIResponseData(ctx, bodyBytes)
253	254
254	255	var param any
@@ -301,6 +302,9 @@ func (c *GeminiClient) SendRawMessageStream(ctx context.Context, modelName strin · 46 unmodified lines
301	302	return
302	303	}
303	304	delete(c.modelQuotaExceeded, modelName)
	305	+ defer func() {
	306	+ _ = stream.Close()
	307	+ }()
304	308
305	309	newCtx := context.WithValue(ctx, "alt", alt)
306	310	var param any

internal/client/openai-compatibility_client.go+360−0

@@ -0,0 +1,360 @@
	1	+// Package client defines the interface and base structure for AI API clients.
	2	+// It provides a common interface that all supported AI service clients must implement,
	3	+// including methods for sending messages, handling streams, and managing authentication.
	4	+package client
	5	+
	6	+import (
	7	+ "bufio"
	8	+ "bytes"
	9	+ "context"
	10	+ "fmt"
	11	+ "io"
	12	+ "net/http"
	13	+ "strings"
	14	+ "sync"
	15	+ "time"
	16	+
	17	+ "github.com/gin-gonic/gin"
	18	+ "github.com/luispater/CLIProxyAPI/internal/auth"
	19	+ "github.com/luispater/CLIProxyAPI/internal/config"
	20	+ . "github.com/luispater/CLIProxyAPI/internal/constant"
	21	+ "github.com/luispater/CLIProxyAPI/internal/interfaces"
	22	+ "github.com/luispater/CLIProxyAPI/internal/translator/translator"
	23	+ "github.com/luispater/CLIProxyAPI/internal/util"
	24	+ log "github.com/sirupsen/logrus"
	25	+ "github.com/tidwall/sjson"
	26	+)
	27	+
	28	+// OpenAICompatibilityClient implements the Client interface for external OpenAI-compatible API providers.
	29	+// This client handles requests to external services that support OpenAI-compatible APIs,
	30	+// such as OpenRouter, Together.ai, and other similar services.
	31	+type OpenAICompatibilityClient struct {
	32	+ ClientBase
	33	+ compatConfig *config.OpenAICompatibility
	34	+ currentAPIKeyIndex int
	35	+}
	36	+
	37	+// NewOpenAICompatibilityClient creates a new OpenAI compatibility client instance.
	38	+//
	39	+// Parameters:
	40	+// - cfg: The application configuration.
	41	+// - compatConfig: The OpenAI compatibility configuration for the specific provider.
	42	+//
	43	+// Returns:
	44	+// - *OpenAICompatibilityClient: A new OpenAI compatibility client instance.
	45	+// - error: An error if the client creation fails.
	46	+func NewOpenAICompatibilityClient(cfg config.Config, compatConfig config.OpenAICompatibility) (*OpenAICompatibilityClient, error) {
	47	+ if compatConfig == nil {
	48	+ return nil, fmt.Errorf("compatibility configuration is required")
	49	+ }
	50	+
	51	+ if len(compatConfig.APIKeys) == 0 {
	52	+ return nil, fmt.Errorf("at least one API key is required for OpenAI compatibility provider: %s", compatConfig.Name)
	53	+ }
	54	+
	55	+ httpClient := util.SetProxy(cfg, &http.Client{})
	56	+ client := &OpenAICompatibilityClient{
	57	+ ClientBase: ClientBase{
	58	+ RequestMutex: &sync.Mutex{},
	59	+ httpClient: httpClient,
	60	+ cfg: cfg,
	61	+ modelQuotaExceeded: make(map[string]*time.Time),
	62	+ },
	63	+ compatConfig: compatConfig,
	64	+ currentAPIKeyIndex: 0,
	65	+ }
	66	+
	67	+ return client, nil
	68	+}
	69	+
	70	+// Type returns the client type.
	71	+func (c *OpenAICompatibilityClient) Type() string {
	72	+ return OPENAI
	73	+}
	74	+
	75	+// Provider returns the provider name for this client.
	76	+func (c *OpenAICompatibilityClient) Provider() string {
	77	+ return c.compatConfig.Name
	78	+}
	79	+
	80	+// CanProvideModel checks if this client can provide the specified model alias.
	81	+//
	82	+// Parameters:
	83	+// - modelName: The name/alias of the model to check.
	84	+//
	85	+// Returns:
	86	+// - bool: True if the model alias is supported, false otherwise.
	87	+func (c *OpenAICompatibilityClient) CanProvideModel(modelName string) bool {
	88	+ for _, model := range c.compatConfig.Models {
	89	+ if model.Alias == modelName {
	90	+ return true
	91	+ }
	92	+ }
	93	+ return false
	94	+}
	95	+
	96	+// GetUserAgent returns the user agent string for OpenAI compatibility API requests.
	97	+func (c *OpenAICompatibilityClient) GetUserAgent() string {
	98	+ return fmt.Sprintf("cli-proxy-api-%s", c.compatConfig.Name)
	99	+}
	100	+
	101	+// TokenStorage returns nil as this client doesn't use traditional token storage.
	102	+func (c *OpenAICompatibilityClient) TokenStorage() auth.TokenStorage {
	103	+ return nil
	104	+}
	105	+
	106	+// GetCurrentAPIKey returns the current API key to use, with rotation support.
	107	+func (c *OpenAICompatibilityClient) GetCurrentAPIKey() string {
	108	+ if len(c.compatConfig.APIKeys) == 0 {
	109	+ return ""
	110	+ }
	111	+
	112	+ key := c.compatConfig.APIKeys[c.currentAPIKeyIndex]
	113	+ // Rotate to next key for load balancing
	114	+ c.currentAPIKeyIndex = (c.currentAPIKeyIndex + 1) % len(c.compatConfig.APIKeys)
	115	+ return key
	116	+}
	117	+
	118	+// GetActualModelName returns the actual model name to use with the external API
	119	+// based on the provided alias.
	120	+func (c *OpenAICompatibilityClient) GetActualModelName(alias string) string {
	121	+ for _, model := range c.compatConfig.Models {
	122	+ if model.Alias == alias {
	123	+ return model.Name
	124	+ }
	125	+ }
	126	+ return alias // fallback to alias if not found
	127	+}
	128	+
	129	+// APIRequest makes an HTTP request to the OpenAI-compatible API.
	130	+//
	131	+// Parameters:
	132	+// - ctx: The context for the request.
	133	+// - modelName: The model name to use.
	134	+// - endpoint: The API endpoint path.
	135	+// - rawJSON: The raw JSON request data.
	136	+// - alt: Alternative response format (not used for OpenAI compatibility).
	137	+// - stream: Whether this is a streaming request.
	138	+//
	139	+// Returns:
	140	+// - io.ReadCloser: The response body reader.
	141	+// - *interfaces.ErrorMessage: An error message if the request fails.
	142	+func (c OpenAICompatibilityClient) APIRequest(ctx context.Context, modelName string, endpoint string, rawJSON []byte, alt string, stream bool) (io.ReadCloser, interfaces.ErrorMessage) {
	143	+ // Replace the model alias with the actual model name in the request
	144	+ actualModelName := c.GetActualModelName(modelName)
	145	+ modifiedJSON, errReplace := sjson.SetBytes(rawJSON, "model", actualModelName)
	146	+ if errReplace != nil {
	147	+ return nil, &interfaces.ErrorMessage{
	148	+ StatusCode: http.StatusInternalServerError,
	149	+ Error: fmt.Errorf("failed to replace model name: %w", errReplace),
	150	+ }
	151	+ }
	152	+
	153	+ // Create the HTTP request
	154	+ url := strings.TrimSuffix(c.compatConfig.BaseURL, "/") + endpoint
	155	+ req, errReq := http.NewRequestWithContext(ctx, "POST", url, bytes.NewReader(modifiedJSON))
	156	+ if errReq != nil {
	157	+ return nil, &interfaces.ErrorMessage{
	158	+ StatusCode: http.StatusInternalServerError,
	159	+ Error: fmt.Errorf("failed to create request: %w", errReq),
	160	+ }
	161	+ }
	162	+
	163	+ // Set headers
	164	+ req.Header.Set("Content-Type", "application/json")
	165	+ apiKey := c.GetCurrentAPIKey()
	166	+ if apiKey != "" {
	167	+ req.Header.Set("Authorization", fmt.Sprintf("Bearer %s", apiKey))
	168	+ }
	169	+ req.Header.Set("User-Agent", c.GetUserAgent())
	170	+
	171	+ if stream {
	172	+ req.Header.Set("Accept", "text/event-stream")
	173	+ req.Header.Set("Cache-Control", "no-cache")
	174	+ }
	175	+
	176	+ log.Debugf("OpenAI Compatibility [%s] API request: %s", c.compatConfig.Name, util.HideAPIKey(apiKey))
	177	+
	178	+ // Send the request
	179	+ resp, err := c.httpClient.Do(req)
	180	+ if err != nil {
	181	+ return nil, &interfaces.ErrorMessage{StatusCode: 500, Error: fmt.Errorf("failed to execute request: %v", err)}
	182	+ }
	183	+
	184	+ if resp.StatusCode < 200 \|\| resp.StatusCode >= 300 {
	185	+ defer func() {
	186	+ if err = resp.Body.Close(); err != nil {
	187	+ log.Printf("warn: failed to close response body: %v", err)
	188	+ }
	189	+ }()
	190	+ bodyBytes, _ := io.ReadAll(resp.Body)
	191	+ // log.Debug(string(jsonBody))
	192	+ return nil, &interfaces.ErrorMessage{StatusCode: resp.StatusCode, Error: fmt.Errorf("%s", string(bodyBytes))}
	193	+ }
	194	+
	195	+ return resp.Body, nil
	196	+}
	197	+
	198	+// SendRawMessage sends a raw message to the OpenAI-compatible API.
	199	+//
	200	+// Parameters:
	201	+// - ctx: The context for the request.
	202	+// - modelName: The model alias name to use.
	203	+// - rawJSON: The raw JSON request data.
	204	+// - alt: Alternative response format parameter.
	205	+//
	206	+// Returns:
	207	+// - []byte: The response data from the API.
	208	+// - *interfaces.ErrorMessage: An error message if the request fails.
	209	+func (c OpenAICompatibilityClient) SendRawMessage(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, interfaces.ErrorMessage) {
	210	+ handler := ctx.Value("handler").(interfaces.APIHandler)
	211	+ handlerType := handler.HandlerType()
	212	+ rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, false)
	213	+
	214	+ respBody, err := c.APIRequest(ctx, modelName, "/chat/completions", rawJSON, alt, false)
	215	+ if err != nil {
	216	+ if err.StatusCode == 429 {
	217	+ now := time.Now()
	218	+ c.modelQuotaExceeded[modelName] = &now
	219	+ }
	220	+ return nil, err
	221	+ }
	222	+ delete(c.modelQuotaExceeded, modelName)
	223	+ bodyBytes, errReadAll := io.ReadAll(respBody)
	224	+ if errReadAll != nil {
	225	+ return nil, &interfaces.ErrorMessage{StatusCode: 500, Error: errReadAll}
	226	+ }
	227	+
	228	+ _ = respBody.Close()
	229	+ c.AddAPIResponseData(ctx, bodyBytes)
	230	+
	231	+ var param any
	232	+ bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
	233	+
	234	+ return bodyBytes, nil
	235	+}
	236	+
	237	+// SendRawMessageStream sends a raw streaming message to the OpenAI-compatible API.
	238	+//
	239	+// Parameters:
	240	+// - ctx: The context for the request.
	241	+// - modelName: The model alias name to use.
	242	+// - rawJSON: The raw JSON request data.
	243	+// - alt: Alternative response format parameter.
	244	+//
	245	+// Returns:
	246	+// - <-chan []byte: A channel that will receive response chunks.
	247	+// - <-chan *interfaces.ErrorMessage: A channel that will receive error messages.
	248	+func (c OpenAICompatibilityClient) SendRawMessageStream(ctx context.Context, modelName string, rawJSON []byte, alt string) (<-chan []byte, <-chan interfaces.ErrorMessage) {
	249	+ handler := ctx.Value("handler").(interfaces.APIHandler)
	250	+ handlerType := handler.HandlerType()
	251	+ rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, true)
	252	+
	253	+ dataTag := []byte("data: ")
	254	+ doneTag := []byte("data: [DONE]")
	255	+ errChan := make(chan *interfaces.ErrorMessage)
	256	+ dataChan := make(chan []byte)
	257	+ // log.Debugf(string(rawJSON))
	258	+ // return dataChan, errChan
	259	+ go func() {
	260	+ defer close(errChan)
	261	+ defer close(dataChan)
	262	+
	263	+ // Set streaming flag in the request
	264	+ rawJSON, _ = sjson.SetBytes(rawJSON, "stream", true)
	265	+
	266	+ newCtx := context.WithValue(ctx, "gin", ctx.Value("gin").(*gin.Context))
	267	+
	268	+ stream, err := c.APIRequest(newCtx, modelName, "/chat/completions", rawJSON, alt, true)
	269	+ if err != nil {
	270	+ if err.StatusCode == 429 {
	271	+ now := time.Now()
	272	+ c.modelQuotaExceeded[modelName] = &now
	273	+ }
	274	+ errChan <- err
	275	+ return
	276	+ }
	277	+ delete(c.modelQuotaExceeded, modelName)
	278	+ defer func() {
	279	+ _ = stream.Close()
	280	+ }()
	281	+
	282	+ scanner := bufio.NewScanner(stream)
	283	+
	284	+ if translator.NeedConvert(handlerType, c.Type()) {
	285	+ var param any
	286	+ for scanner.Scan() {
	287	+ line := scanner.Bytes()
	288	+ if bytes.HasPrefix(line, dataTag) {
	289	+ if bytes.Equal(line, doneTag) {
	290	+ break
	291	+ }
	292	+ lines := translator.Response(handlerType, c.Type(), newCtx, modelName, line[6:], &param)
	293	+ for i := 0; i < len(lines); i++ {
	294	+ dataChan <- []byte(lines[i])
	295	+ }
	296	+ }
	297	+ }
	298	+ } else {
	299	+ // No translation needed, stream data directly
	300	+ for scanner.Scan() {
	301	+ line := scanner.Bytes()
	302	+ if bytes.HasPrefix(line, dataTag) {
	303	+ if bytes.Equal(line, doneTag) {
	304	+ break
	305	+ }
	306	+ c.AddAPIResponseData(newCtx, line[6:])
	307	+ dataChan <- line[6:]
	308	+ }
	309	+ }
	310	+ }
	311	+
	312	+ if scanner.Err() != nil {
	313	+ errChan <- &interfaces.ErrorMessage{StatusCode: 500, Error: scanner.Err()}
	314	+ }
	315	+ }()
	316	+
	317	+ return dataChan, errChan
	318	+}
	319	+
	320	+// SendRawTokenCount sends a token count request (not implemented for OpenAI compatibility).
	321	+// This method is required by the Client interface but not supported by OpenAI compatibility clients.
	322	+func (c OpenAICompatibilityClient) SendRawTokenCount(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, interfaces.ErrorMessage) {
	323	+ return nil, &interfaces.ErrorMessage{
	324	+ StatusCode: http.StatusNotImplemented,
	325	+ Error: fmt.Errorf("token counting not supported for OpenAI compatibility clients"),
	326	+ }
	327	+}
	328	+
	329	+// GetEmail returns a placeholder email for this OpenAI compatibility client.
	330	+// Since these clients don't use traditional email-based authentication,
	331	+// we return the provider name as an identifier.
	332	+func (c *OpenAICompatibilityClient) GetEmail() string {
	333	+ return fmt.Sprintf("openai-compatibility-%s", c.compatConfig.Name)
	334	+}
	335	+
	336	+// IsModelQuotaExceeded checks if the specified model has exceeded its quota.
	337	+// For OpenAI compatibility clients, this is based on tracked quota exceeded times.
	338	+func (c *OpenAICompatibilityClient) IsModelQuotaExceeded(model string) bool {
	339	+ if quota, exists := c.modelQuotaExceeded[model]; exists && quota != nil {
	340	+ // Check if quota exceeded time is less than 5 minutes ago
	341	+ if time.Since(quota) < 5time.Minute {
	342	+ return true
	343	+ }
	344	+ // Clear expired quota tracking
	345	+ delete(c.modelQuotaExceeded, model)
	346	+ }
	347	+ return false
	348	+}
	349	+
	350	+// SaveTokenToFile returns nil as this client type doesn't use traditional token storage.
	351	+func (c *OpenAICompatibilityClient) SaveTokenToFile() error {
	352	+ // No token file to save for OpenAI compatibility clients
	353	+ return nil
	354	+}
	355	+
	356	+// RefreshTokens is not applicable for OpenAI compatibility clients as they use API keys.
	357	+func (c *OpenAICompatibilityClient) RefreshTokens(ctx context.Context) error {
	358	+ // API keys don't need refreshing
	359	+ return nil
	360	+}

internal/client/qwen_client.go+4−0

@@ -128,6 +128,7 @@ func (c *QwenClient) SendRawMessage(ctx context.Context, modelName string, rawJS · 127 unmodified lines
128	128	return nil, &interfaces.ErrorMessage{StatusCode: 500, Error: errReadAll}
129	129	}
130	130
	131	+ _ = respBody.Close()
131	132	c.AddAPIResponseData(ctx, bodyBytes)
132	133
133	134	var param any
@@ -186,6 +187,9 @@ func (c *QwenClient) SendRawMessageStream(ctx context.Context, modelName string, · 52 unmodified lines
186	187	return
187	188	}
188	189	delete(c.modelQuotaExceeded, modelName)
	190	+ defer func() {
	191	+ _ = stream.Close()
	192	+ }()
189	193
190	194	scanner := bufio.NewScanner(stream)
191	195	buffer := make([]byte, 10240*1024)

internal/cmd/run.go+12−0

@@ -150,6 +150,18 @@ func StartService(cfg *config.Config, configPath string) { · 149 unmodified lines
150	150	}
151	151	}
152	152
	153	+ if len(cfg.OpenAICompatibility) > 0 {
	154	+ // Initialize clients for OpenAI compatibility configurations
	155	+ for _, compatConfig := range cfg.OpenAICompatibility {
	156	+ log.Debugf("Initializing OpenAI compatibility client for provider: %s", compatConfig.Name)
	157	+ compatClient, errClient := client.NewOpenAICompatibilityClient(cfg, &compatConfig)
	158	+ if errClient != nil {
	159	+ log.Fatalf("failed to create OpenAI compatibility client for %s: %v", compatConfig.Name, errClient)
	160	+ }
	161	+ cliClients = append(cliClients, compatClient)
	162	+ }
	163	+ }
	164	+
153	165	// Create and start the API server with the pool of clients in a separate goroutine.
154	166	apiServer := api.NewServer(cfg, cliClients)
155	167	log.Infof("Starting API server on port %d", cfg.Port)

internal/config/config.go+29−0

@@ -41,6 +41,9 @@ type Config struct { · 40 unmodified lines
41	41	RequestRetry int `yaml:"request-retry"`
42	42
43	43	ClaudeKey []ClaudeKey `yaml:"claude-api-key"`
	44	+
	45	+ // OpenAICompatibility defines OpenAI API compatibility configurations for external providers.
	46	+ OpenAICompatibility []OpenAICompatibility `yaml:"openai-compatibility"`
44	47	}
45	48
46	49	// QuotaExceeded defines the behavior when API quota limits are exceeded.
@@ -64,6 +67,32 @@ type ClaudeKey struct { · 17 unmodified lines
64	67	BaseURL string `yaml:"base-url"`
65	68	}
66	69
	70	+// OpenAICompatibility represents the configuration for OpenAI API compatibility
	71	+// with external providers, allowing model aliases to be routed through OpenAI API format.
	72	+type OpenAICompatibility struct {
	73	+ // Name is the identifier for this OpenAI compatibility configuration.
	74	+ Name string `yaml:"name"`
	75	+
	76	+ // BaseURL is the base URL for the external OpenAI-compatible API endpoint.
	77	+ BaseURL string `yaml:"base-url"`
	78	+
	79	+ // APIKeys are the authentication keys for accessing the external API services.
	80	+ APIKeys []string `yaml:"api-keys"`
	81	+
	82	+ // Models defines the model configurations including aliases for routing.
	83	+ Models []OpenAICompatibilityModel `yaml:"models"`
	84	+}
	85	+
	86	+// OpenAICompatibilityModel represents a model configuration for OpenAI compatibility,
	87	+// including the actual model name and its alias for API routing.
	88	+type OpenAICompatibilityModel struct {
	89	+ // Name is the actual model name used by the external provider.
	90	+ Name string `yaml:"name"`
	91	+
	92	+ // Alias is the model name alias that clients will use to reference this model.
	93	+ Alias string `yaml:"alias"`
	94	+}
	95	+
67	96	// LoadConfig reads a YAML configuration file from the given path,
68	97	// unmarshals it into a Config struct, applies environment variable overrides,
69	98	// and returns it.

internal/constant/constant.go+6−5

@@ -1,9 +1,10 @@
1	1	package constant
2	2
3	3	const (
4		− GEMINI = "gemini"
5		− GEMINICLI = "gemini-cli"
6		− CODEX = "codex"
7		− CLAUDE = "claude"
8		− OPENAI = "openai"
	4	+ GEMINI = "gemini"
	5	+ GEMINICLI = "gemini-cli"
	6	+ CODEX = "codex"
	7	+ CLAUDE = "claude"
	8	+ OPENAI = "openai"
	9	+ OPENAI_COMPATIBILITY = "openai-compatibility"
9	10	)

internal/util/provider.go+59−2

@@ -5,25 +5,33 @@ package util · 4 unmodified lines
5	5
6	6	import (
7	7	"strings"
	8	+
	9	+ "github.com/luispater/CLIProxyAPI/internal/config"
8	10	)
9	11
10	12	// GetProviderName determines the AI service provider based on the model name.
11	13	// It analyzes the model name string to identify which service provider it belongs to.
	14	+// First checks for OpenAI compatibility aliases, then falls back to standard provider detection.
12	15	//
13	16	// Supported providers:
14	17	// - "gemini" for Google's Gemini models
15	18	// - "gpt" for OpenAI's GPT models
16	19	// - "claude" for Anthropic's Claude models
17	20	// - "qwen" for Alibaba's Qwen models
	21	+// - "openai-compatibility" for external OpenAI-compatible providers
18	22	// - "unknow" for unrecognized model names
19	23	//
20	24	// Parameters:
21	25	// - modelName: The name of the model to identify the provider for.
	26	+// - cfg: The application configuration containing OpenAI compatibility settings.
22	27	//
23	28	// Returns:
24	29	// - string: The name of the provider.
25		−func GetProviderName(modelName string) string {
26		− if strings.Contains(modelName, "gemini") {
	30	+func GetProviderName(modelName string, cfg *config.Config) string {
	31	+ // First check if this model name is an OpenAI compatibility alias
	32	+ if IsOpenAICompatibilityAlias(modelName, cfg) {
	33	+ return "openai-compatibility"
	34	+ } else if strings.Contains(modelName, "gemini") { // Fall back to standard provider detection
27	35	return "gemini"
28	36	} else if strings.Contains(modelName, "gpt") {
29	37	return "gpt"
@@ -37,6 +45,55 @@ func GetProviderName(modelName string) string { · 7 unmodified lines
37	45	return "unknow"
38	46	}
39	47
	48	+// IsOpenAICompatibilityAlias checks if the given model name is an alias
	49	+// configured for OpenAI compatibility routing.
	50	+//
	51	+// Parameters:
	52	+// - modelName: The model name to check
	53	+// - cfg: The application configuration containing OpenAI compatibility settings
	54	+//
	55	+// Returns:
	56	+// - bool: True if the model name is an OpenAI compatibility alias, false otherwise
	57	+func IsOpenAICompatibilityAlias(modelName string, cfg *config.Config) bool {
	58	+ if cfg == nil {
	59	+ return false
	60	+ }
	61	+
	62	+ for _, compat := range cfg.OpenAICompatibility {
	63	+ for _, model := range compat.Models {
	64	+ if model.Alias == modelName {
	65	+ return true
	66	+ }
	67	+ }
	68	+ }
	69	+ return false
	70	+}
	71	+
	72	+// GetOpenAICompatibilityConfig returns the OpenAI compatibility configuration
	73	+// and model details for the given alias.
	74	+//
	75	+// Parameters:
	76	+// - alias: The model alias to find configuration for
	77	+// - cfg: The application configuration containing OpenAI compatibility settings
	78	+//
	79	+// Returns:
	80	+// - *config.OpenAICompatibility: The matching compatibility configuration, or nil if not found
	81	+// - *config.OpenAICompatibilityModel: The matching model configuration, or nil if not found
	82	+func GetOpenAICompatibilityConfig(alias string, cfg config.Config) (config.OpenAICompatibility, *config.OpenAICompatibilityModel) {
	83	+ if cfg == nil {
	84	+ return nil, nil
	85	+ }
	86	+
	87	+ for _, compat := range cfg.OpenAICompatibility {
	88	+ for _, model := range compat.Models {
	89	+ if model.Alias == alias {
	90	+ return &compat, &model
	91	+ }
	92	+ }
	93	+ }
	94	+ return nil, nil
	95	+}
	96	+
40	97	// InArray checks if a string exists in a slice of strings.
41	98	// It iterates through the slice and returns true if the target string is found,
42	99	// otherwise it returns false.