214 Commits

Author	SHA1	Message	Date
Deadwalk	bfadc4e2e0	fix(anthropic): improve MCP tool message handling and error resilience ... 修复Anthropic协议中MCP工具消息处理问题，增强错误恢复能力： - 改进MessageContent的JSON解析，支持原始JSON数据保留 - 移除对空内容的过滤，避免丢失MCP工具消息 - 增加详细调试日志，便于问题排查 - 优化错误响应处理，保留响应体信息 - 修复路由转发时可能的数据丢失问题 fix(anthropic): improve MCP tool message handling and error resilience - Enhance MessageContent JSON parsing to preserve raw JSON data - Remove empty content filtering to prevent MCP tool message loss - Add detailed debug logging for troubleshooting - Optimize error response handling with response body preservation - Fix potential data loss during routing 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>	2025-09-30 08:29:38 +08:00
Deadwalk	a587ac145a	feat(anthropic): support both string and array content formats ... - Add MessageContent type to handle both string and array content formats - Implement UnmarshalJSON to automatically detect content format - Add ToContentArray() method for unified content processing - Update Message and Response structs to use MessageContent - Fix all content processing logic in main.go and controller - Resolve JSON parsing errors for different content formats 支持Anthropic协议字符串和数组内容格式： - 添加MessageContent类型处理字符串和数组两种内容格式 - 实现UnmarshalJSON自动检测内容格式 - 添加ToContentArray()方法统一内容处理 - 更新Message和Response结构体使用MessageContent - 修复main.go和controller中的所有内容处理逻辑 - 解决不同内容格式的JSON解析错误	2025-09-29 16:52:19 +08:00
Deadwalk	7b9631408f	fix(anthropic): filter empty content messages to prevent API errors ... - Add message content validation in getAndValidateAnthropicRequest - Filter out messages with empty content to prevent 'all messages must have non-empty content' errors - Support both text and non-text content validation (images, tool_use, tool_result) - Add detailed logging for filtered messages - Ensure at least one valid message remains after filtering 修复Anthropic协议空消息内容问题： - 在getAndValidateAnthropicRequest中添加消息内容验证 - 过滤空内容消息以防止'all messages must have non-empty content'错误 - 支持文本和非文本内容验证（图片、工具调用、工具结果） - 为被过滤的消息添加详细日志记录 - 确保过滤后至少保留一条有效消息	2025-09-29 16:51:55 +08:00
Deadwalk	48396d3f33	fix: 修复CodeReview发现的安全问题和代码质量问题 | fix security and code quality issues identified by CodeReview ... - 修复JSON注入漏洞：使用json.Marshal()安全转义字符串内容 - 定义常量CHARS_PER_TOKEN替换硬编码的token估算数字4 - 处理UnmarshalJSON错误，避免静默失败并记录错误日志 - 定义常量替换硬编码的API端点路径，提高可维护性 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>	2025-09-28 16:41:48 +08:00
Deadwalk	e27612a620	refactor: 优化Anthropic协议日志输出 | optimize Anthropic protocol log output ... - 清理不必要的调试日志，减少生产环境噪音 - 将详细日志从Infof降级为Debugf级别 - 保留关键错误日志和重要流程信息 - 优化日志结构，提高可读性和维护性 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>	2025-09-28 16:22:18 +08:00
Deadwalk	008ffe4662	feat: 支持Anthropic API协议 | support Anthropic API protocol ... - 添加Anthropic适配器实现 | Add Anthropic adaptor implementation - 支持Anthropic消息格式转换 | Support Anthropic message format conversion - 添加Vertex AI Claude适配器支持 | Add Vertex AI Claude adapter support - 更新Anthropic的中继模式定义 | Update relay mode definitions for Anthropic - 添加Anthropic控制器和路由 | Add Anthropic controller and routing 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: zgdmemail@gmail.com	2025-09-28 11:13:55 +08:00
longkeyy	7ac553541b	feat: update openrouter models and price 20250213 (#2084)	2025-02-16 18:01:59 +08:00
longkeyy	a5c517c27a	feat: update ali models and price 20250213 (#2086)	2025-02-16 18:01:24 +08:00
JustSong	3f421c4f04	feat: support Gemini openai compatible api	2025-02-16 17:59:39 +08:00
JustSong	cafd0a0327	feat: add OpenAI compatible channel (close #2091)	2025-02-16 17:38:06 +08:00
Laisky.Cai	fef7ae048b	feat: support gemini-2.0-flash (#2055) ... * feat: support gemini-2.0-flash - Enhance model support by adding new entries and refining checks for system instruction compatibility. - Update logging display behavior and adjust default quotas for better user experience. - Revamp pricing structures in the billing system to reflect current model values and deprecate outdated entries. - Streamline code by replacing hardcoded values with configurations for maintainability. * feat: add new Gemini 2.0 flash models to adapter and billing ratio * fix: update GetRequestURL to support gemini-1.5 model in versioning	2025-02-12 00:34:25 +08:00
JustSong	53da209134	feat: add AliBailian adaptor and update channel options	2025-02-12 00:15:43 +08:00
JustSong	eb96aa635e	feat: update OpenRouter channel name and add model list for OpenRouter adaptor	2025-02-11 00:20:55 +08:00
JustSong	c715f2bc1d	feat: add new models for xai	2025-02-09 21:21:28 +08:00
JustSong	696265774e	feat: add MiniMax model constants to the adaptor	2025-02-09 18:55:32 +08:00
JustSong	974729426d	feat: refactor Xunfei API version handling and update model list	2025-02-09 18:50:51 +08:00
JustSong	57c1367ec8	feat: add Xunfei V2 channel support and update related configurations	2025-02-09 18:31:54 +08:00
JustSong	44233d5c04	feat: add completion tokens details and reasoning effort fields to model (close #2050)	2025-02-09 18:14:01 +08:00
JustSong	bf45a955c3	fix: update system prompt handling by renaming field and ensuring proper usage in request processing (close #2069)	2025-02-09 14:41:42 +08:00
JustSong	dd65b997dd	feat: add Baidu V2 channel support and improve model handling	2025-02-09 12:37:26 +08:00
longkeyy	3e3b8230ac	fix: add read/write locks for ModelRatio and GroupRatio to prevent concurrent map read/write issues (#2067)	2025-02-09 11:02:45 +08:00
JustSong	07808122a6	fix: fix Debugf not using DebugEnabled (close #2068)	2025-02-09 10:57:22 +08:00
JustSong	293ba93ad6	fix: remove outdated model from ModelList and add new deepseek models	2025-02-07 00:13:57 +08:00
JustSong	e146b14d46	fix: add default API version handling and enhance error message checks for Gemini	2025-02-07 00:01:38 +08:00
JustSong	e19045f925	chore: add deepseek-reasoner	2025-02-06 23:38:29 +08:00
JustSong	2250f311e1	feat: add latest version of Claude model to constants	2025-02-02 22:25:52 +08:00
lyj	e43f758623	Update model.go (#1963)	2025-02-02 22:24:31 +08:00
jiz4oh	ea3331b79a	fix: remove duplicated model (#2012) ... the model `llama-3.2-11b-vision-preview` was declared at 3915ce9814/relay/adaptor/groq/constants.go (L11)	2025-02-02 22:21:30 +08:00
Jeremy JIANG	7a97ddc03c	fix: fix hunyuan paramete & update price (#2019) ... * fix: omit null TopP and Temperature fields in request body * feat: update Hunyuan model ratios	2025-02-02 22:20:38 +08:00
JustSong	d7028b55fd	feat: update model list in Zhipu constants for expanded options	2025-02-02 22:15:42 +08:00
Jeremy JIANG	36a03f465b	feat: update Zhipu model prices (#2020)	2025-02-02 22:06:07 +08:00
JustSong	c81ee68dd0	feat: improve error logging for token encoder initialization with offline usage guidance	2025-02-02 17:24:23 +08:00
JustSong	4f68f3e1b3	chore: update log content	2025-01-31 20:16:56 +08:00
JustSong	0bab887b2d	chore: update log content	2025-01-31 20:15:04 +08:00
JustSong	dc470ce82e	feat: show stream & elapsed time in log detail	2025-01-31 19:34:22 +08:00
JustSong	ea0721d525	feat: update log content format	2025-01-31 18:15:43 +08:00
Fennng	09911a301d	feat: support hunyuan-embedding (#2035) ... * feat: support hunyuan-embedding * chore: improve implementation --------- Co-authored-by: LUO Feng <luofeng@flowpp.com> Co-authored-by: JustSong <quanpengsong@gmail.com>	2025-01-31 16:48:02 +08:00
Laisky.Cai	d88e07fd9a	feat: add deepseek-reasoner & gemini-2.0-flash-thinking-exp-01-21 (#2045) ... * feat: add MILLI_USD constant and update pricing for deepseek services * feat: add support for new Gemini model version 'gemini-2.0-flash-thinking-exp-01-21'	2025-01-31 15:15:59 +08:00
Qiying Wang	a858292b54	feat: support gpt-4o-2024-11-20 (#1941)	2024-12-22 19:49:50 +08:00
Yuwei Ba	ff589b5e4a	chore: update model mapping implementation for audio (#1932) ... * fixed model mapping * chore: update implementation --------- Co-authored-by: JustSong <quanpengsong@gmail.com>	2024-12-22 19:33:11 +08:00
ZhangTianrong	59eae186a3	fix: remove the duplicate claude-3-5-haiku-20241022 in Anthropic's base model list (#1957) ... * Update constants.go Remove the duplicate `claude-3-5-haiku-20241022` causing issue 1928 * fix: fix syntax error --------- Co-authored-by: JustSong <quanpengsong@gmail.com>	2024-12-22 18:58:29 +08:00
Wei Tingjiang	49ffb1c60d	feat: enhance response handling to support gemini-2.0-thinking (#1995)	2024-12-22 18:25:44 +08:00
longkeyy	2f16649896	feat: update qwen model and price (#1966)	2024-12-22 18:22:57 +08:00
Laisky.Cai	e9f117ff72	feat: add gemini-2.0-flash-exp and fix race condition in processChannelRelayError (#1983) ... Co-authored-by: JustSong <39998050+songquanpeng@users.noreply.github.com>	2024-12-21 20:32:30 +08:00
Laisky.Cai	6bb5247bd6	feat: add support for new OpenAI models and update billing ratios (#1990)	2024-12-21 20:28:51 +08:00
Laisky.Cai	305ce14fe3	feat: support replicate chat models (#1989) ... * feat: add Replicate adaptor and integrate into channel and API types * feat: support llm chat on replicate	2024-12-21 14:41:19 +08:00
JustSong	6ab87f8a08	feat: add warning in log when system prompt is reset	2024-11-10 17:18:46 +08:00
JustSong	6eb0770a89	feat: support set system prompt for channel (close #1920)	2024-11-10 14:53:34 +08:00
JustSong	92cd46d64f	feat: able to use ENFORCE_INCLUDE_USAGE to enforce include usage in response	2024-11-10 00:36:08 +08:00
lihangfu	2b2dc2c733	fix: update Spark Lite's domain to lite (#1896)	2024-11-09 23:55:55 +08:00