Compare commits
5 Commits
async-cont
...
batch-inst
| Author | SHA1 | Date | |
|---|---|---|---|
|
cea31fed38 | ||
|
|
8f8147828b | ||
|
158792d82f | ||
|
|
858d048d81 | ||
|
|
2f518d4c7a |
@@ -23,11 +23,11 @@ A collection of enhancements, plugins, and prompts for [open-webui](https://gith
|
||||
### 🔥 Top 6 Popular Plugins
|
||||
| Rank | Plugin | Version | Downloads | Views | 📅 Updated |
|
||||
| :---: | :--- | :---: | :---: | :---: | :---: |
|
||||
| 🥇 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) |  |  |  |  |
|
||||
| 🥇 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) |  |  |  |  |
|
||||
| 🥈 | [Smart Infographic](https://openwebui.com/posts/smart_infographic_ad6f0c7f) |  |  |  |  |
|
||||
| 🥉 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) |  |  |  |  |
|
||||
| 4️⃣ | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) |  |  |  |  |
|
||||
| 5️⃣ | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) |  |  |  |  |
|
||||
| 🥉 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) |  |  |  |  |
|
||||
| 4️⃣ | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) |  |  |  |  |
|
||||
| 5️⃣ | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) |  |  |  |  |
|
||||
| 6️⃣ | [AI Task Instruction Generator](https://openwebui.com/posts/ai_task_instruction_generator_9bab8b37) |  |  |  |  |
|
||||
|
||||
### 📈 Total Downloads Trend
|
||||
|
||||
@@ -20,11 +20,11 @@ OpenWebUI 增强功能集合。包含个人开发与收集的插件、提示词
|
||||
### 🔥 热门插件 Top 6
|
||||
| 排名 | 插件 | 版本 | 下载 | 浏览 | 📅 更新 |
|
||||
| :---: | :--- | :---: | :---: | :---: | :---: |
|
||||
| 🥇 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) |  |  |  |  |
|
||||
| 🥇 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) |  |  |  |  |
|
||||
| 🥈 | [Smart Infographic](https://openwebui.com/posts/smart_infographic_ad6f0c7f) |  |  |  |  |
|
||||
| 🥉 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) |  |  |  |  |
|
||||
| 4️⃣ | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) |  |  |  |  |
|
||||
| 5️⃣ | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) |  |  |  |  |
|
||||
| 🥉 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) |  |  |  |  |
|
||||
| 4️⃣ | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) |  |  |  |  |
|
||||
| 5️⃣ | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) |  |  |  |  |
|
||||
| 6️⃣ | [AI Task Instruction Generator](https://openwebui.com/posts/ai_task_instruction_generator_9bab8b37) |  |  |  |  |
|
||||
|
||||
### 📈 总下载量累计趋势
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
{
|
||||
"schemaVersion": 1,
|
||||
"label": "downloads",
|
||||
"message": "8.8k",
|
||||
"message": "9.1k",
|
||||
"color": "blue",
|
||||
"namedLogo": "openwebui"
|
||||
}
|
||||
@@ -1,6 +1,6 @@
|
||||
{
|
||||
"schemaVersion": 1,
|
||||
"label": "followers",
|
||||
"message": "344",
|
||||
"message": "353",
|
||||
"color": "blue"
|
||||
}
|
||||
@@ -1,6 +1,6 @@
|
||||
{
|
||||
"schemaVersion": 1,
|
||||
"label": "points",
|
||||
"message": "351",
|
||||
"message": "359",
|
||||
"color": "orange"
|
||||
}
|
||||
@@ -1,6 +1,6 @@
|
||||
{
|
||||
"schemaVersion": 1,
|
||||
"label": "upvotes",
|
||||
"message": "300",
|
||||
"message": "305",
|
||||
"color": "brightgreen"
|
||||
}
|
||||
@@ -1,14 +1,14 @@
|
||||
{
|
||||
"total_posts": 27,
|
||||
"total_downloads": 8765,
|
||||
"total_views": 92460,
|
||||
"total_upvotes": 300,
|
||||
"total_downloads": 9120,
|
||||
"total_views": 95785,
|
||||
"total_upvotes": 305,
|
||||
"total_downvotes": 4,
|
||||
"total_saves": 431,
|
||||
"total_comments": 73,
|
||||
"total_saves": 452,
|
||||
"total_comments": 77,
|
||||
"by_type": {
|
||||
"tool": 2,
|
||||
"filter": 4,
|
||||
"tool": 2,
|
||||
"pipe": 1,
|
||||
"action": 12,
|
||||
"prompt": 1
|
||||
@@ -21,13 +21,13 @@
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Intelligently analyzes text content and generates interactive mind maps to help users structure and visualize knowledge.",
|
||||
"downloads": 1730,
|
||||
"views": 14700,
|
||||
"upvotes": 30,
|
||||
"saves": 67,
|
||||
"comments": 21,
|
||||
"created_at": "2025-12-31",
|
||||
"updated_at": "2026-02-28",
|
||||
"downloads": 1797,
|
||||
"views": 15350,
|
||||
"upvotes": 31,
|
||||
"saves": 72,
|
||||
"comments": 23,
|
||||
"created_at": "2025-12-30",
|
||||
"updated_at": "2026-02-27",
|
||||
"url": "https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a"
|
||||
},
|
||||
{
|
||||
@@ -37,11 +37,11 @@
|
||||
"version": "1.5.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "AI-powered infographic generator based on AntV Infographic. Supports professional templates, auto-icon matching, and SVG/PNG downloads.",
|
||||
"downloads": 1330,
|
||||
"views": 13250,
|
||||
"upvotes": 27,
|
||||
"saves": 50,
|
||||
"comments": 10,
|
||||
"downloads": 1362,
|
||||
"views": 13589,
|
||||
"upvotes": 28,
|
||||
"saves": 53,
|
||||
"comments": 12,
|
||||
"created_at": "2025-12-28",
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/smart_infographic_ad6f0c7f"
|
||||
@@ -53,15 +53,31 @@
|
||||
"version": "1.2.8",
|
||||
"author": "Fu-Jie",
|
||||
"description": "A content normalizer filter that fixes common Markdown formatting issues in LLM outputs, such as broken code blocks, LaTeX formulas, and list formatting. Including LaTeX command protection.",
|
||||
"downloads": 807,
|
||||
"views": 8499,
|
||||
"downloads": 838,
|
||||
"views": 8739,
|
||||
"upvotes": 21,
|
||||
"saves": 44,
|
||||
"saves": 46,
|
||||
"comments": 5,
|
||||
"created_at": "2026-01-12",
|
||||
"updated_at": "2026-03-09",
|
||||
"updated_at": "2026-03-08",
|
||||
"url": "https://openwebui.com/posts/markdown_normalizer_baaa8732"
|
||||
},
|
||||
{
|
||||
"title": "Async Context Compression",
|
||||
"slug": "async_context_compression_b1655bc8",
|
||||
"type": "filter",
|
||||
"version": "1.5.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Reduces token consumption in long conversations while maintaining coherence through intelligent summarization and message compression.",
|
||||
"downloads": 801,
|
||||
"views": 7258,
|
||||
"upvotes": 18,
|
||||
"saves": 54,
|
||||
"comments": 0,
|
||||
"created_at": "2025-11-08",
|
||||
"updated_at": "2026-03-14",
|
||||
"url": "https://openwebui.com/posts/async_context_compression_b1655bc8"
|
||||
},
|
||||
{
|
||||
"title": "Export to Word Enhanced",
|
||||
"slug": "export_to_word_enhanced_formatting_fca6a315",
|
||||
@@ -69,31 +85,15 @@
|
||||
"version": "0.4.4",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Export current conversation from Markdown to Word (.docx) with Mermaid diagrams rendered client-side (Mermaid.js, SVG+PNG), LaTeX math, real hyperlinks, improved tables, syntax highlighting, and blockquote support.",
|
||||
"downloads": 767,
|
||||
"views": 5898,
|
||||
"upvotes": 18,
|
||||
"saves": 38,
|
||||
"downloads": 799,
|
||||
"views": 6146,
|
||||
"upvotes": 20,
|
||||
"saves": 41,
|
||||
"comments": 5,
|
||||
"created_at": "2026-01-03",
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315"
|
||||
},
|
||||
{
|
||||
"title": "Async Context Compression",
|
||||
"slug": "async_context_compression_b1655bc8",
|
||||
"type": "filter",
|
||||
"version": "1.4.1",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Reduces token consumption in long conversations while maintaining coherence through intelligent summarization and message compression.",
|
||||
"downloads": 760,
|
||||
"views": 6985,
|
||||
"upvotes": 17,
|
||||
"saves": 50,
|
||||
"comments": 0,
|
||||
"created_at": "2025-11-08",
|
||||
"updated_at": "2026-03-11",
|
||||
"url": "https://openwebui.com/posts/async_context_compression_b1655bc8"
|
||||
},
|
||||
{
|
||||
"title": "AI Task Instruction Generator",
|
||||
"slug": "ai_task_instruction_generator_9bab8b37",
|
||||
@@ -101,10 +101,10 @@
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 666,
|
||||
"views": 7490,
|
||||
"downloads": 692,
|
||||
"views": 7783,
|
||||
"upvotes": 10,
|
||||
"saves": 19,
|
||||
"saves": 20,
|
||||
"comments": 0,
|
||||
"created_at": "2026-01-28",
|
||||
"updated_at": "2026-01-28",
|
||||
@@ -117,10 +117,10 @@
|
||||
"version": "0.3.7",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Extracts tables from chat messages and exports them to Excel (.xlsx) files with smart formatting.",
|
||||
"downloads": 604,
|
||||
"views": 3426,
|
||||
"downloads": 616,
|
||||
"views": 3508,
|
||||
"upvotes": 11,
|
||||
"saves": 11,
|
||||
"saves": 12,
|
||||
"comments": 0,
|
||||
"created_at": "2025-05-30",
|
||||
"updated_at": "2026-02-13",
|
||||
@@ -133,13 +133,13 @@
|
||||
"version": "0.3.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Standalone OpenWebUI tool for managing native Workspace Skills (list/show/install/create/update/delete) for any model.",
|
||||
"downloads": 434,
|
||||
"views": 5597,
|
||||
"downloads": 500,
|
||||
"views": 6112,
|
||||
"upvotes": 8,
|
||||
"saves": 22,
|
||||
"saves": 23,
|
||||
"comments": 4,
|
||||
"created_at": "2026-02-28",
|
||||
"updated_at": "2026-03-11",
|
||||
"updated_at": "2026-03-14",
|
||||
"url": "https://openwebui.com/posts/openwebui_skills_manager_tool_b4bce8e4"
|
||||
},
|
||||
{
|
||||
@@ -149,10 +149,10 @@
|
||||
"version": "0.10.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "A powerful Agent SDK integration for OpenWebUI. It deeply bridges GitHub Copilot SDK with OpenWebUI's ecosystem, enabling the Agent to autonomously perform intent recognition, web search, and context compaction. It seamlessly reuses your existing Tools, MCP servers, OpenAPI servers, and Skills for a professional, full-featured experience.",
|
||||
"downloads": 399,
|
||||
"views": 5542,
|
||||
"downloads": 403,
|
||||
"views": 5699,
|
||||
"upvotes": 16,
|
||||
"saves": 11,
|
||||
"saves": 12,
|
||||
"comments": 8,
|
||||
"created_at": "2026-01-26",
|
||||
"updated_at": "2026-03-07",
|
||||
@@ -165,8 +165,8 @@
|
||||
"version": "0.2.4",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Quickly generates beautiful flashcards from text, extracting key points and categories.",
|
||||
"downloads": 325,
|
||||
"views": 4650,
|
||||
"downloads": 331,
|
||||
"views": 4722,
|
||||
"upvotes": 13,
|
||||
"saves": 22,
|
||||
"comments": 2,
|
||||
@@ -181,8 +181,8 @@
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "A comprehensive thinking lens that dives deep into any content - from context to logic, insights, and action paths.",
|
||||
"downloads": 224,
|
||||
"views": 1852,
|
||||
"downloads": 229,
|
||||
"views": 1887,
|
||||
"upvotes": 6,
|
||||
"saves": 15,
|
||||
"comments": 0,
|
||||
@@ -197,8 +197,8 @@
|
||||
"version": "0.4.4",
|
||||
"author": "Fu-Jie",
|
||||
"description": "将对话导出为 Word (.docx)，支持 Mermaid 图表 (客户端渲染 SVG+PNG)、LaTeX 数学公式、真实超链接、增强表格格式、代码高亮和引用块。",
|
||||
"downloads": 171,
|
||||
"views": 2974,
|
||||
"downloads": 172,
|
||||
"views": 3038,
|
||||
"upvotes": 14,
|
||||
"saves": 7,
|
||||
"comments": 4,
|
||||
@@ -213,8 +213,8 @@
|
||||
"version": "0.1.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Automatically extracts project rules from conversations and injects them into the folder's system prompt.",
|
||||
"downloads": 125,
|
||||
"views": 2137,
|
||||
"downloads": 130,
|
||||
"views": 2181,
|
||||
"upvotes": 7,
|
||||
"saves": 13,
|
||||
"comments": 0,
|
||||
@@ -229,8 +229,8 @@
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Intelligently analyzes text content and generates interactive mind maps to help users structure and visualize knowledge.",
|
||||
"downloads": 100,
|
||||
"views": 2203,
|
||||
"downloads": 116,
|
||||
"views": 2375,
|
||||
"upvotes": 5,
|
||||
"saves": 4,
|
||||
"comments": 0,
|
||||
@@ -246,12 +246,12 @@
|
||||
"author": "Fu-Jie",
|
||||
"description": "A specialized filter to bypass OpenWebUI's default RAG for GitHub Copilot SDK models. It moves uploaded files to a safe location ('copilot_files') so the Copilot Pipe can process them natively without interference.",
|
||||
"downloads": 93,
|
||||
"views": 2452,
|
||||
"views": 2474,
|
||||
"upvotes": 4,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
"created_at": "2026-02-09",
|
||||
"updated_at": "2026-03-04",
|
||||
"updated_at": "2026-03-03",
|
||||
"url": "https://openwebui.com/posts/github_copilot_sdk_files_filter_403a62ee"
|
||||
},
|
||||
{
|
||||
@@ -261,8 +261,8 @@
|
||||
"version": "1.5.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "基于 AntV Infographic 的智能信息图生成插件。支持多种专业模板，自动图标匹配，并提供 SVG/PNG 下载功能。",
|
||||
"downloads": 71,
|
||||
"views": 1545,
|
||||
"downloads": 72,
|
||||
"views": 1572,
|
||||
"upvotes": 10,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
@@ -277,8 +277,8 @@
|
||||
"version": "0.9.2",
|
||||
"author": "Fu-Jie",
|
||||
"description": "智能分析文本内容,生成交互式思维导图,帮助用户结构化和可视化知识。",
|
||||
"downloads": 53,
|
||||
"views": 789,
|
||||
"downloads": 56,
|
||||
"views": 814,
|
||||
"upvotes": 6,
|
||||
"saves": 2,
|
||||
"comments": 0,
|
||||
@@ -293,8 +293,8 @@
|
||||
"version": "1.2.2",
|
||||
"author": "Fu-Jie",
|
||||
"description": "通过智能摘要和消息压缩，降低长对话的 token 消耗，同时保持对话连贯性。",
|
||||
"downloads": 40,
|
||||
"views": 876,
|
||||
"downloads": 42,
|
||||
"views": 904,
|
||||
"upvotes": 7,
|
||||
"saves": 5,
|
||||
"comments": 0,
|
||||
@@ -302,6 +302,22 @@
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/异步上下文压缩_5c0617cb"
|
||||
},
|
||||
{
|
||||
"title": "精读",
|
||||
"slug": "精读_99830b0f",
|
||||
"type": "action",
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "全方位的思维透镜 —— 从背景全景到逻辑脉络，从深度洞察到行动路径。",
|
||||
"downloads": 37,
|
||||
"views": 708,
|
||||
"upvotes": 5,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
"created_at": "2026-01-08",
|
||||
"updated_at": "2026-01-08",
|
||||
"url": "https://openwebui.com/posts/精读_99830b0f"
|
||||
},
|
||||
{
|
||||
"title": "闪记卡 (Flash Card)",
|
||||
"slug": "闪记卡生成插件_4a31eac3",
|
||||
@@ -310,7 +326,7 @@
|
||||
"author": "Fu-Jie",
|
||||
"description": "快速将文本提炼为精美的学习记忆卡片，支持核心要点提取与分类。",
|
||||
"downloads": 34,
|
||||
"views": 917,
|
||||
"views": 926,
|
||||
"upvotes": 7,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
@@ -318,22 +334,6 @@
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/闪记卡生成插件_4a31eac3"
|
||||
},
|
||||
{
|
||||
"title": "精读",
|
||||
"slug": "精读_99830b0f",
|
||||
"type": "action",
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "全方位的思维透镜 —— 从背景全景到逻辑脉络，从深度洞察到行动路径。",
|
||||
"downloads": 32,
|
||||
"views": 678,
|
||||
"upvotes": 5,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
"created_at": "2026-01-08",
|
||||
"updated_at": "2026-01-08",
|
||||
"url": "https://openwebui.com/posts/精读_99830b0f"
|
||||
},
|
||||
{
|
||||
"title": "An Unconventional Use of Open Terminal ⚡",
|
||||
"slug": "an_unconventional_use_of_open_terminal_35498f8f",
|
||||
@@ -342,11 +342,11 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 3009,
|
||||
"views": 3335,
|
||||
"upvotes": 7,
|
||||
"saves": 1,
|
||||
"comments": 2,
|
||||
"created_at": "2026-03-07",
|
||||
"created_at": "2026-03-06",
|
||||
"updated_at": "2026-03-07",
|
||||
"url": "https://openwebui.com/posts/an_unconventional_use_of_open_terminal_35498f8f"
|
||||
},
|
||||
@@ -358,11 +358,11 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 1762,
|
||||
"views": 1803,
|
||||
"upvotes": 5,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
"created_at": "2026-02-28",
|
||||
"created_at": "2026-02-27",
|
||||
"updated_at": "2026-02-28",
|
||||
"url": "https://openwebui.com/posts/github_copilot_sdk_pipe_v090_copilot_sdk_skills_co_99a42452"
|
||||
},
|
||||
@@ -374,11 +374,11 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 2758,
|
||||
"views": 2797,
|
||||
"upvotes": 8,
|
||||
"saves": 4,
|
||||
"comments": 1,
|
||||
"created_at": "2026-02-23",
|
||||
"created_at": "2026-02-22",
|
||||
"updated_at": "2026-02-28",
|
||||
"url": "https://openwebui.com/posts/github_copilot_sdk_pipe_v070_native_tool_ui_zero_c_4af38131"
|
||||
},
|
||||
@@ -390,7 +390,7 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 2430,
|
||||
"views": 2442,
|
||||
"upvotes": 7,
|
||||
"saves": 5,
|
||||
"comments": 0,
|
||||
@@ -406,12 +406,12 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 1989,
|
||||
"views": 2014,
|
||||
"upvotes": 13,
|
||||
"saves": 23,
|
||||
"comments": 9,
|
||||
"created_at": "2026-01-25",
|
||||
"updated_at": "2026-01-29",
|
||||
"updated_at": "2026-01-28",
|
||||
"url": "https://openwebui.com/posts/open_webui_prompt_plus_ai_powered_prompt_manager_s_15fa060e"
|
||||
},
|
||||
{
|
||||
@@ -422,7 +422,7 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 263,
|
||||
"views": 271,
|
||||
"upvotes": 2,
|
||||
"saves": 0,
|
||||
"comments": 0,
|
||||
@@ -438,7 +438,7 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 1579,
|
||||
"views": 1588,
|
||||
"upvotes": 16,
|
||||
"saves": 13,
|
||||
"comments": 2,
|
||||
@@ -452,11 +452,11 @@
|
||||
"name": "Fu-Jie",
|
||||
"profile_url": "https://openwebui.com/u/Fu-Jie",
|
||||
"profile_image": "https://community.s3.openwebui.com/uploads/users/b15d1348-4347-42b4-b815-e053342d6cb0/profile_d9510745-4bd4-4f8f-a997-4a21847d9300.webp",
|
||||
"followers": 344,
|
||||
"followers": 353,
|
||||
"following": 6,
|
||||
"total_points": 351,
|
||||
"post_points": 298,
|
||||
"comment_points": 53,
|
||||
"contributions": 66
|
||||
"total_points": 359,
|
||||
"post_points": 303,
|
||||
"comment_points": 56,
|
||||
"contributions": 68
|
||||
}
|
||||
}
|
||||
@@ -1,19 +1,17 @@
|
||||
{
|
||||
"total_posts": 27,
|
||||
"total_downloads": 7786,
|
||||
"total_views": 82342,
|
||||
"total_upvotes": 281,
|
||||
"total_downloads": 8947,
|
||||
"total_views": 94188,
|
||||
"total_upvotes": 301,
|
||||
"total_downvotes": 4,
|
||||
"total_saves": 398,
|
||||
"total_comments": 63,
|
||||
"total_saves": 444,
|
||||
"total_comments": 75,
|
||||
"by_type": {
|
||||
"post": 6,
|
||||
"tool": 2,
|
||||
"pipe": 1,
|
||||
"filter": 4,
|
||||
"pipe": 1,
|
||||
"action": 12,
|
||||
"prompt": 1,
|
||||
"review": 1
|
||||
"prompt": 1
|
||||
},
|
||||
"posts": [
|
||||
{
|
||||
@@ -23,11 +21,11 @@
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Intelligently analyzes text content and generates interactive mind maps to help users structure and visualize knowledge.",
|
||||
"downloads": 1542,
|
||||
"views": 12996,
|
||||
"upvotes": 28,
|
||||
"saves": 66,
|
||||
"comments": 18,
|
||||
"downloads": 1772,
|
||||
"views": 15047,
|
||||
"upvotes": 30,
|
||||
"saves": 70,
|
||||
"comments": 21,
|
||||
"created_at": "2025-12-30",
|
||||
"updated_at": "2026-02-27",
|
||||
"url": "https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a"
|
||||
@@ -39,11 +37,11 @@
|
||||
"version": "1.5.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "AI-powered infographic generator based on AntV Infographic. Supports professional templates, auto-icon matching, and SVG/PNG downloads.",
|
||||
"downloads": 1230,
|
||||
"views": 12309,
|
||||
"upvotes": 25,
|
||||
"saves": 46,
|
||||
"comments": 10,
|
||||
"downloads": 1350,
|
||||
"views": 13453,
|
||||
"upvotes": 27,
|
||||
"saves": 52,
|
||||
"comments": 12,
|
||||
"created_at": "2025-12-28",
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/smart_infographic_ad6f0c7f"
|
||||
@@ -52,16 +50,16 @@
|
||||
"title": "Markdown Normalizer",
|
||||
"slug": "markdown_normalizer_baaa8732",
|
||||
"type": "filter",
|
||||
"version": "1.2.7",
|
||||
"version": "1.2.8",
|
||||
"author": "Fu-Jie",
|
||||
"description": "A content normalizer filter that fixes common Markdown formatting issues in LLM outputs, such as broken code blocks, LaTeX formulas, and list formatting. Including LaTeX command protection.",
|
||||
"downloads": 719,
|
||||
"views": 7704,
|
||||
"upvotes": 20,
|
||||
"saves": 42,
|
||||
"downloads": 824,
|
||||
"views": 8622,
|
||||
"upvotes": 21,
|
||||
"saves": 45,
|
||||
"comments": 5,
|
||||
"created_at": "2026-01-12",
|
||||
"updated_at": "2026-03-03",
|
||||
"updated_at": "2026-03-08",
|
||||
"url": "https://openwebui.com/posts/markdown_normalizer_baaa8732"
|
||||
},
|
||||
{
|
||||
@@ -71,10 +69,10 @@
|
||||
"version": "0.4.4",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Export current conversation from Markdown to Word (.docx) with Mermaid diagrams rendered client-side (Mermaid.js, SVG+PNG), LaTeX math, real hyperlinks, improved tables, syntax highlighting, and blockquote support.",
|
||||
"downloads": 700,
|
||||
"views": 5399,
|
||||
"upvotes": 17,
|
||||
"saves": 37,
|
||||
"downloads": 780,
|
||||
"views": 6015,
|
||||
"upvotes": 19,
|
||||
"saves": 39,
|
||||
"comments": 5,
|
||||
"created_at": "2026-01-03",
|
||||
"updated_at": "2026-02-13",
|
||||
@@ -84,16 +82,16 @@
|
||||
"title": "Async Context Compression",
|
||||
"slug": "async_context_compression_b1655bc8",
|
||||
"type": "filter",
|
||||
"version": "1.3.0",
|
||||
"version": "1.4.2",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Reduces token consumption in long conversations while maintaining coherence through intelligent summarization and message compression.",
|
||||
"downloads": 669,
|
||||
"views": 6274,
|
||||
"upvotes": 16,
|
||||
"saves": 47,
|
||||
"downloads": 776,
|
||||
"views": 7102,
|
||||
"upvotes": 17,
|
||||
"saves": 53,
|
||||
"comments": 0,
|
||||
"created_at": "2025-11-08",
|
||||
"updated_at": "2026-03-03",
|
||||
"updated_at": "2026-03-13",
|
||||
"url": "https://openwebui.com/posts/async_context_compression_b1655bc8"
|
||||
},
|
||||
{
|
||||
@@ -103,10 +101,10 @@
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 583,
|
||||
"views": 6659,
|
||||
"upvotes": 9,
|
||||
"saves": 17,
|
||||
"downloads": 676,
|
||||
"views": 7619,
|
||||
"upvotes": 10,
|
||||
"saves": 19,
|
||||
"comments": 0,
|
||||
"created_at": "2026-01-28",
|
||||
"updated_at": "2026-01-28",
|
||||
@@ -119,29 +117,45 @@
|
||||
"version": "0.3.7",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Extracts tables from chat messages and exports them to Excel (.xlsx) files with smart formatting.",
|
||||
"downloads": 563,
|
||||
"views": 3153,
|
||||
"downloads": 612,
|
||||
"views": 3475,
|
||||
"upvotes": 11,
|
||||
"saves": 11,
|
||||
"saves": 12,
|
||||
"comments": 0,
|
||||
"created_at": "2025-05-30",
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/export_mulit_table_to_excel_244b8f9d"
|
||||
},
|
||||
{
|
||||
"title": "OpenWebUI Skills Manager Tool",
|
||||
"slug": "openwebui_skills_manager_tool_b4bce8e4",
|
||||
"type": "tool",
|
||||
"version": "0.3.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Standalone OpenWebUI tool for managing native Workspace Skills (list/show/install/create/update/delete) for any model.",
|
||||
"downloads": 463,
|
||||
"views": 5862,
|
||||
"upvotes": 8,
|
||||
"saves": 23,
|
||||
"comments": 4,
|
||||
"created_at": "2026-02-28",
|
||||
"updated_at": "2026-03-13",
|
||||
"url": "https://openwebui.com/posts/openwebui_skills_manager_tool_b4bce8e4"
|
||||
},
|
||||
{
|
||||
"title": "GitHub Copilot Official SDK Pipe",
|
||||
"slug": "github_copilot_official_sdk_pipe_ce96f7b4",
|
||||
"type": "pipe",
|
||||
"version": "0.9.1",
|
||||
"version": "0.10.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "A powerful Agent SDK integration for OpenWebUI. It deeply bridges GitHub Copilot SDK with OpenWebUI's ecosystem, enabling the Agent to autonomously perform intent recognition, web search, and context compaction. It seamlessly reuses your existing Tools, MCP servers, OpenAPI servers, and Skills for a professional, full-featured experience.",
|
||||
"downloads": 335,
|
||||
"views": 4905,
|
||||
"downloads": 402,
|
||||
"views": 5629,
|
||||
"upvotes": 16,
|
||||
"saves": 10,
|
||||
"comments": 6,
|
||||
"saves": 12,
|
||||
"comments": 8,
|
||||
"created_at": "2026-01-26",
|
||||
"updated_at": "2026-03-03",
|
||||
"updated_at": "2026-03-07",
|
||||
"url": "https://openwebui.com/posts/github_copilot_official_sdk_pipe_ce96f7b4"
|
||||
},
|
||||
{
|
||||
@@ -151,31 +165,15 @@
|
||||
"version": "0.2.4",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Quickly generates beautiful flashcards from text, extracting key points and categories.",
|
||||
"downloads": 312,
|
||||
"views": 4448,
|
||||
"downloads": 327,
|
||||
"views": 4685,
|
||||
"upvotes": 13,
|
||||
"saves": 20,
|
||||
"saves": 22,
|
||||
"comments": 2,
|
||||
"created_at": "2025-12-30",
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/flash_card_65a2ea8f"
|
||||
},
|
||||
{
|
||||
"title": "OpenWebUI Skills Manager Tool",
|
||||
"slug": "openwebui_skills_manager_tool_b4bce8e4",
|
||||
"type": "tool",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 303,
|
||||
"views": 4265,
|
||||
"upvotes": 7,
|
||||
"saves": 13,
|
||||
"comments": 2,
|
||||
"created_at": "2026-02-28",
|
||||
"updated_at": "2026-03-05",
|
||||
"url": "https://openwebui.com/posts/openwebui_skills_manager_tool_b4bce8e4"
|
||||
},
|
||||
{
|
||||
"title": "Deep Dive",
|
||||
"slug": "deep_dive_c0b846e4",
|
||||
@@ -183,8 +181,8 @@
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "A comprehensive thinking lens that dives deep into any content - from context to logic, insights, and action paths.",
|
||||
"downloads": 219,
|
||||
"views": 1764,
|
||||
"downloads": 228,
|
||||
"views": 1874,
|
||||
"upvotes": 6,
|
||||
"saves": 15,
|
||||
"comments": 0,
|
||||
@@ -199,8 +197,8 @@
|
||||
"version": "0.4.4",
|
||||
"author": "Fu-Jie",
|
||||
"description": "将对话导出为 Word (.docx)，支持 Mermaid 图表 (客户端渲染 SVG+PNG)、LaTeX 数学公式、真实超链接、增强表格格式、代码高亮和引用块。",
|
||||
"downloads": 165,
|
||||
"views": 2831,
|
||||
"downloads": 172,
|
||||
"views": 3019,
|
||||
"upvotes": 14,
|
||||
"saves": 7,
|
||||
"comments": 4,
|
||||
@@ -215,15 +213,31 @@
|
||||
"version": "0.1.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Automatically extracts project rules from conversations and injects them into the folder's system prompt.",
|
||||
"downloads": 112,
|
||||
"views": 1992,
|
||||
"downloads": 128,
|
||||
"views": 2154,
|
||||
"upvotes": 7,
|
||||
"saves": 11,
|
||||
"saves": 13,
|
||||
"comments": 0,
|
||||
"created_at": "2026-01-20",
|
||||
"updated_at": "2026-01-20",
|
||||
"url": "https://openwebui.com/posts/folder_memory_auto_evolving_project_context_4a9875b2"
|
||||
},
|
||||
{
|
||||
"title": "🧠 Smart Mind Map Tool: Auto-Generate Interactive Knowledge Graphs",
|
||||
"slug": "smart_mind_map_tool_auto_generate_interactive_know_d25f4e3d",
|
||||
"type": "tool",
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "Intelligently analyzes text content and generates interactive mind maps to help users structure and visualize knowledge.",
|
||||
"downloads": 106,
|
||||
"views": 2284,
|
||||
"upvotes": 5,
|
||||
"saves": 4,
|
||||
"comments": 0,
|
||||
"created_at": "2026-03-04",
|
||||
"updated_at": "2026-03-05",
|
||||
"url": "https://openwebui.com/posts/smart_mind_map_tool_auto_generate_interactive_know_d25f4e3d"
|
||||
},
|
||||
{
|
||||
"title": "GitHub Copilot SDK Files Filter",
|
||||
"slug": "github_copilot_sdk_files_filter_403a62ee",
|
||||
@@ -231,8 +245,8 @@
|
||||
"version": "0.1.3",
|
||||
"author": "Fu-Jie",
|
||||
"description": "A specialized filter to bypass OpenWebUI's default RAG for GitHub Copilot SDK models. It moves uploaded files to a safe location ('copilot_files') so the Copilot Pipe can process them natively without interference.",
|
||||
"downloads": 76,
|
||||
"views": 2311,
|
||||
"downloads": 93,
|
||||
"views": 2462,
|
||||
"upvotes": 4,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
@@ -247,8 +261,8 @@
|
||||
"version": "1.5.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "基于 AntV Infographic 的智能信息图生成插件。支持多种专业模板，自动图标匹配，并提供 SVG/PNG 下载功能。",
|
||||
"downloads": 68,
|
||||
"views": 1431,
|
||||
"downloads": 72,
|
||||
"views": 1566,
|
||||
"upvotes": 10,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
@@ -263,8 +277,8 @@
|
||||
"version": "0.9.2",
|
||||
"author": "Fu-Jie",
|
||||
"description": "智能分析文本内容,生成交互式思维导图,帮助用户结构化和可视化知识。",
|
||||
"downloads": 52,
|
||||
"views": 761,
|
||||
"downloads": 56,
|
||||
"views": 807,
|
||||
"upvotes": 6,
|
||||
"saves": 2,
|
||||
"comments": 0,
|
||||
@@ -279,8 +293,8 @@
|
||||
"version": "1.2.2",
|
||||
"author": "Fu-Jie",
|
||||
"description": "通过智能摘要和消息压缩，降低长对话的 token 消耗，同时保持对话连贯性。",
|
||||
"downloads": 39,
|
||||
"views": 838,
|
||||
"downloads": 42,
|
||||
"views": 892,
|
||||
"upvotes": 7,
|
||||
"saves": 5,
|
||||
"comments": 0,
|
||||
@@ -288,22 +302,6 @@
|
||||
"updated_at": "2026-02-13",
|
||||
"url": "https://openwebui.com/posts/异步上下文压缩_5c0617cb"
|
||||
},
|
||||
{
|
||||
"title": "🧠 Smart Mind Map Tool: Auto-Generate Interactive Knowledge Graphs",
|
||||
"slug": "smart_mind_map_tool_auto_generate_interactive_know_d25f4e3d",
|
||||
"type": "tool",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 34,
|
||||
"views": 767,
|
||||
"upvotes": 2,
|
||||
"saves": 3,
|
||||
"comments": 0,
|
||||
"created_at": "2026-03-04",
|
||||
"updated_at": "2026-03-05",
|
||||
"url": "https://openwebui.com/posts/smart_mind_map_tool_auto_generate_interactive_know_d25f4e3d"
|
||||
},
|
||||
{
|
||||
"title": "闪记卡 (Flash Card)",
|
||||
"slug": "闪记卡生成插件_4a31eac3",
|
||||
@@ -312,7 +310,7 @@
|
||||
"author": "Fu-Jie",
|
||||
"description": "快速将文本提炼为精美的学习记忆卡片，支持核心要点提取与分类。",
|
||||
"downloads": 34,
|
||||
"views": 888,
|
||||
"views": 922,
|
||||
"upvotes": 7,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
@@ -327,8 +325,8 @@
|
||||
"version": "1.0.0",
|
||||
"author": "Fu-Jie",
|
||||
"description": "全方位的思维透镜 —— 从背景全景到逻辑脉络，从深度洞察到行动路径。",
|
||||
"downloads": 31,
|
||||
"views": 647,
|
||||
"downloads": 34,
|
||||
"views": 699,
|
||||
"upvotes": 5,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
@@ -339,28 +337,28 @@
|
||||
{
|
||||
"title": "An Unconventional Use of Open Terminal ⚡",
|
||||
"slug": "an_unconventional_use_of_open_terminal_35498f8f",
|
||||
"type": "post",
|
||||
"type": "action",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 14,
|
||||
"upvotes": 1,
|
||||
"saves": 0,
|
||||
"comments": 0,
|
||||
"views": 3205,
|
||||
"upvotes": 7,
|
||||
"saves": 1,
|
||||
"comments": 2,
|
||||
"created_at": "2026-03-06",
|
||||
"updated_at": "2026-03-06",
|
||||
"updated_at": "2026-03-07",
|
||||
"url": "https://openwebui.com/posts/an_unconventional_use_of_open_terminal_35498f8f"
|
||||
},
|
||||
{
|
||||
"title": "🚀 GitHub Copilot SDK Pipe v0.9.0: Skills & RichUI",
|
||||
"slug": "github_copilot_sdk_pipe_v090_copilot_sdk_skills_co_99a42452",
|
||||
"type": "post",
|
||||
"type": "pipe",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 1585,
|
||||
"views": 1781,
|
||||
"upvotes": 5,
|
||||
"saves": 1,
|
||||
"comments": 0,
|
||||
@@ -371,12 +369,12 @@
|
||||
{
|
||||
"title": "🚀 GitHub Copilot SDK Pipe v0.7.0: Skills & Rich UI 🛠️",
|
||||
"slug": "github_copilot_sdk_pipe_v070_native_tool_ui_zero_c_4af38131",
|
||||
"type": "post",
|
||||
"type": "pipe",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 2608,
|
||||
"views": 2775,
|
||||
"upvotes": 8,
|
||||
"saves": 4,
|
||||
"comments": 1,
|
||||
@@ -387,14 +385,14 @@
|
||||
{
|
||||
"title": "🚀 GitHub Copilot SDK Pipe: AI That Executes, Not Just Talks",
|
||||
"slug": "github_copilot_sdk_for_openwebui_elevate_your_ai_t_a140f293",
|
||||
"type": "post",
|
||||
"type": "pipe",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 2390,
|
||||
"views": 2441,
|
||||
"upvotes": 7,
|
||||
"saves": 4,
|
||||
"saves": 5,
|
||||
"comments": 0,
|
||||
"created_at": "2026-02-10",
|
||||
"updated_at": "2026-02-10",
|
||||
@@ -403,15 +401,15 @@
|
||||
{
|
||||
"title": "🚀 Open WebUI Prompt Plus: AI-Powered Prompt Manager",
|
||||
"slug": "open_webui_prompt_plus_ai_powered_prompt_manager_s_15fa060e",
|
||||
"type": "post",
|
||||
"type": "action",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 1915,
|
||||
"upvotes": 12,
|
||||
"saves": 21,
|
||||
"comments": 8,
|
||||
"views": 1999,
|
||||
"upvotes": 13,
|
||||
"saves": 23,
|
||||
"comments": 9,
|
||||
"created_at": "2026-01-25",
|
||||
"updated_at": "2026-01-28",
|
||||
"url": "https://openwebui.com/posts/open_webui_prompt_plus_ai_powered_prompt_manager_s_15fa060e"
|
||||
@@ -424,7 +422,7 @@
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 251,
|
||||
"views": 267,
|
||||
"upvotes": 2,
|
||||
"saves": 0,
|
||||
"comments": 0,
|
||||
@@ -435,14 +433,14 @@
|
||||
{
|
||||
"title": " 🛠️ Debug Open WebUI Plugins in Your Browser",
|
||||
"slug": "debug_open_webui_plugins_in_your_browser_81bf7960",
|
||||
"type": "post",
|
||||
"type": "action",
|
||||
"version": "",
|
||||
"author": "",
|
||||
"description": "",
|
||||
"downloads": 0,
|
||||
"views": 1549,
|
||||
"views": 1583,
|
||||
"upvotes": 16,
|
||||
"saves": 12,
|
||||
"saves": 13,
|
||||
"comments": 2,
|
||||
"created_at": "2026-01-10",
|
||||
"updated_at": "2026-01-10",
|
||||
@@ -454,11 +452,11 @@
|
||||
"name": "Fu-Jie",
|
||||
"profile_url": "https://openwebui.com/u/Fu-Jie",
|
||||
"profile_image": "https://community.s3.openwebui.com/uploads/users/b15d1348-4347-42b4-b815-e053342d6cb0/profile_d9510745-4bd4-4f8f-a997-4a21847d9300.webp",
|
||||
"followers": 315,
|
||||
"followers": 348,
|
||||
"following": 6,
|
||||
"total_points": 329,
|
||||
"post_points": 279,
|
||||
"comment_points": 50,
|
||||
"contributions": 59
|
||||
"total_points": 352,
|
||||
"post_points": 299,
|
||||
"comment_points": 53,
|
||||
"contributions": 67
|
||||
}
|
||||
}
|
||||
@@ -8,7 +8,7 @@
|
||||
> *Blue: Downloads | Purple: Views (Real-time dynamic)*
|
||||
|
||||
### 📂 Content Distribution
|
||||
|
||||
|
||||
|
||||
|
||||
## 📈 Overview
|
||||
@@ -25,8 +25,8 @@
|
||||
|
||||
## 📂 By Type
|
||||
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
@@ -35,30 +35,30 @@
|
||||
|
||||
| Rank | Title | Type | Version | Downloads | Views | Upvotes | Saves | Updated |
|
||||
|:---:|------|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
|
||||
| 1 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) | action |  |  |  |  |  | 2026-02-28 |
|
||||
| 1 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) | action |  |  |  |  |  | 2026-02-27 |
|
||||
| 2 | [Smart Infographic](https://openwebui.com/posts/smart_infographic_ad6f0c7f) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 3 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) | filter |  |  |  |  |  | 2026-03-09 |
|
||||
| 4 | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 5 | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) | filter |  |  |  |  |  | 2026-03-11 |
|
||||
| 3 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) | filter |  |  |  |  |  | 2026-03-08 |
|
||||
| 4 | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) | filter |  |  |  |  |  | 2026-03-14 |
|
||||
| 5 | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 6 | [AI Task Instruction Generator](https://openwebui.com/posts/ai_task_instruction_generator_9bab8b37) | prompt |  |  |  |  |  | 2026-01-28 |
|
||||
| 7 | [Export to Excel](https://openwebui.com/posts/export_mulit_table_to_excel_244b8f9d) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 8 | [OpenWebUI Skills Manager Tool](https://openwebui.com/posts/openwebui_skills_manager_tool_b4bce8e4) | tool |  |  |  |  |  | 2026-03-11 |
|
||||
| 8 | [OpenWebUI Skills Manager Tool](https://openwebui.com/posts/openwebui_skills_manager_tool_b4bce8e4) | tool |  |  |  |  |  | 2026-03-14 |
|
||||
| 9 | [GitHub Copilot Official SDK Pipe](https://openwebui.com/posts/github_copilot_official_sdk_pipe_ce96f7b4) | pipe |  |  |  |  |  | 2026-03-07 |
|
||||
| 10 | [Flash Card](https://openwebui.com/posts/flash_card_65a2ea8f) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 11 | [Deep Dive](https://openwebui.com/posts/deep_dive_c0b846e4) | action |  |  |  |  |  | 2026-01-08 |
|
||||
| 12 | [导出为Word增强版](https://openwebui.com/posts/导出为_word_支持公式流程图表格和代码块_8a6306c0) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 13 | [📂 Folder Memory – Auto-Evolving Project Context](https://openwebui.com/posts/folder_memory_auto_evolving_project_context_4a9875b2) | filter |  |  |  |  |  | 2026-01-20 |
|
||||
| 14 | [🧠 Smart Mind Map Tool: Auto-Generate Interactive Knowledge Graphs](https://openwebui.com/posts/smart_mind_map_tool_auto_generate_interactive_know_d25f4e3d) | tool |  |  |  |  |  | 2026-03-05 |
|
||||
| 15 | [GitHub Copilot SDK Files Filter](https://openwebui.com/posts/github_copilot_sdk_files_filter_403a62ee) | filter |  |  |  |  |  | 2026-03-04 |
|
||||
| 15 | [GitHub Copilot SDK Files Filter](https://openwebui.com/posts/github_copilot_sdk_files_filter_403a62ee) | filter |  |  |  |  |  | 2026-03-03 |
|
||||
| 16 | [智能信息图](https://openwebui.com/posts/智能信息图_e04a48ff) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 17 | [思维导图](https://openwebui.com/posts/智能生成交互式思维导图帮助用户可视化知识_8d4b097b) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 18 | [异步上下文压缩](https://openwebui.com/posts/异步上下文压缩_5c0617cb) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 19 | [闪记卡 (Flash Card)](https://openwebui.com/posts/闪记卡生成插件_4a31eac3) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 20 | [精读](https://openwebui.com/posts/精读_99830b0f) | action |  |  |  |  |  | 2026-01-08 |
|
||||
| 19 | [精读](https://openwebui.com/posts/精读_99830b0f) | action |  |  |  |  |  | 2026-01-08 |
|
||||
| 20 | [闪记卡 (Flash Card)](https://openwebui.com/posts/闪记卡生成插件_4a31eac3) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 21 | [An Unconventional Use of Open Terminal ⚡](https://openwebui.com/posts/an_unconventional_use_of_open_terminal_35498f8f) | action |  |  |  |  |  | 2026-03-07 |
|
||||
| 22 | [🚀 GitHub Copilot SDK Pipe v0.9.0: Skills & RichUI](https://openwebui.com/posts/github_copilot_sdk_pipe_v090_copilot_sdk_skills_co_99a42452) | pipe |  |  |  |  |  | 2026-02-28 |
|
||||
| 23 | [🚀 GitHub Copilot SDK Pipe v0.7.0: Skills & Rich UI 🛠️](https://openwebui.com/posts/github_copilot_sdk_pipe_v070_native_tool_ui_zero_c_4af38131) | pipe |  |  |  |  |  | 2026-02-28 |
|
||||
| 24 | [🚀 GitHub Copilot SDK Pipe: AI That Executes, Not Just Talks](https://openwebui.com/posts/github_copilot_sdk_for_openwebui_elevate_your_ai_t_a140f293) | pipe |  |  |  |  |  | 2026-02-10 |
|
||||
| 25 | [🚀 Open WebUI Prompt Plus: AI-Powered Prompt Manager](https://openwebui.com/posts/open_webui_prompt_plus_ai_powered_prompt_manager_s_15fa060e) | action |  |  |  |  |  | 2026-01-29 |
|
||||
| 25 | [🚀 Open WebUI Prompt Plus: AI-Powered Prompt Manager](https://openwebui.com/posts/open_webui_prompt_plus_ai_powered_prompt_manager_s_15fa060e) | action |  |  |  |  |  | 2026-01-28 |
|
||||
| 26 | [Review of Claude Haiku 4.5](https://openwebui.com/posts/review_of_claude_haiku_45_41b0db39) | review |  |  |  |  |  | 2026-01-14 |
|
||||
| 27 | [ 🛠️ Debug Open WebUI Plugins in Your Browser](https://openwebui.com/posts/debug_open_webui_plugins_in_your_browser_81bf7960) | action |  |  |  |  |  | 2026-01-10 |
|
||||
|
||||
@@ -8,7 +8,7 @@
|
||||
> *蓝色: 总下载量 | 紫色: 总浏览量 (实时动态生成)*
|
||||
|
||||
### 📂 内容分类占比 (Distribution)
|
||||
|
||||
|
||||
|
||||
|
||||
## 📈 总览
|
||||
@@ -25,8 +25,8 @@
|
||||
|
||||
## 📂 按类型分类
|
||||
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
- 
|
||||
@@ -35,30 +35,30 @@
|
||||
|
||||
| 排名 | 标题 | 类型 | 版本 | 下载 | 浏览 | 点赞 | 收藏 | 更新日期 |
|
||||
|:---:|------|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
|
||||
| 1 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) | action |  |  |  |  |  | 2026-02-28 |
|
||||
| 1 | [Smart Mind Map](https://openwebui.com/posts/turn_any_text_into_beautiful_mind_maps_3094c59a) | action |  |  |  |  |  | 2026-02-27 |
|
||||
| 2 | [Smart Infographic](https://openwebui.com/posts/smart_infographic_ad6f0c7f) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 3 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) | filter |  |  |  |  |  | 2026-03-09 |
|
||||
| 4 | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 5 | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) | filter |  |  |  |  |  | 2026-03-11 |
|
||||
| 3 | [Markdown Normalizer](https://openwebui.com/posts/markdown_normalizer_baaa8732) | filter |  |  |  |  |  | 2026-03-08 |
|
||||
| 4 | [Async Context Compression](https://openwebui.com/posts/async_context_compression_b1655bc8) | filter |  |  |  |  |  | 2026-03-14 |
|
||||
| 5 | [Export to Word Enhanced](https://openwebui.com/posts/export_to_word_enhanced_formatting_fca6a315) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 6 | [AI Task Instruction Generator](https://openwebui.com/posts/ai_task_instruction_generator_9bab8b37) | prompt |  |  |  |  |  | 2026-01-28 |
|
||||
| 7 | [Export to Excel](https://openwebui.com/posts/export_mulit_table_to_excel_244b8f9d) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 8 | [OpenWebUI Skills Manager Tool](https://openwebui.com/posts/openwebui_skills_manager_tool_b4bce8e4) | tool |  |  |  |  |  | 2026-03-11 |
|
||||
| 8 | [OpenWebUI Skills Manager Tool](https://openwebui.com/posts/openwebui_skills_manager_tool_b4bce8e4) | tool |  |  |  |  |  | 2026-03-14 |
|
||||
| 9 | [GitHub Copilot Official SDK Pipe](https://openwebui.com/posts/github_copilot_official_sdk_pipe_ce96f7b4) | pipe |  |  |  |  |  | 2026-03-07 |
|
||||
| 10 | [Flash Card](https://openwebui.com/posts/flash_card_65a2ea8f) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 11 | [Deep Dive](https://openwebui.com/posts/deep_dive_c0b846e4) | action |  |  |  |  |  | 2026-01-08 |
|
||||
| 12 | [导出为Word增强版](https://openwebui.com/posts/导出为_word_支持公式流程图表格和代码块_8a6306c0) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 13 | [📂 Folder Memory – Auto-Evolving Project Context](https://openwebui.com/posts/folder_memory_auto_evolving_project_context_4a9875b2) | filter |  |  |  |  |  | 2026-01-20 |
|
||||
| 14 | [🧠 Smart Mind Map Tool: Auto-Generate Interactive Knowledge Graphs](https://openwebui.com/posts/smart_mind_map_tool_auto_generate_interactive_know_d25f4e3d) | tool |  |  |  |  |  | 2026-03-05 |
|
||||
| 15 | [GitHub Copilot SDK Files Filter](https://openwebui.com/posts/github_copilot_sdk_files_filter_403a62ee) | filter |  |  |  |  |  | 2026-03-04 |
|
||||
| 15 | [GitHub Copilot SDK Files Filter](https://openwebui.com/posts/github_copilot_sdk_files_filter_403a62ee) | filter |  |  |  |  |  | 2026-03-03 |
|
||||
| 16 | [智能信息图](https://openwebui.com/posts/智能信息图_e04a48ff) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 17 | [思维导图](https://openwebui.com/posts/智能生成交互式思维导图帮助用户可视化知识_8d4b097b) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 18 | [异步上下文压缩](https://openwebui.com/posts/异步上下文压缩_5c0617cb) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 19 | [闪记卡 (Flash Card)](https://openwebui.com/posts/闪记卡生成插件_4a31eac3) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 20 | [精读](https://openwebui.com/posts/精读_99830b0f) | action |  |  |  |  |  | 2026-01-08 |
|
||||
| 19 | [精读](https://openwebui.com/posts/精读_99830b0f) | action |  |  |  |  |  | 2026-01-08 |
|
||||
| 20 | [闪记卡 (Flash Card)](https://openwebui.com/posts/闪记卡生成插件_4a31eac3) | action |  |  |  |  |  | 2026-02-13 |
|
||||
| 21 | [An Unconventional Use of Open Terminal ⚡](https://openwebui.com/posts/an_unconventional_use_of_open_terminal_35498f8f) | action |  |  |  |  |  | 2026-03-07 |
|
||||
| 22 | [🚀 GitHub Copilot SDK Pipe v0.9.0: Skills & RichUI](https://openwebui.com/posts/github_copilot_sdk_pipe_v090_copilot_sdk_skills_co_99a42452) | pipe |  |  |  |  |  | 2026-02-28 |
|
||||
| 23 | [🚀 GitHub Copilot SDK Pipe v0.7.0: Skills & Rich UI 🛠️](https://openwebui.com/posts/github_copilot_sdk_pipe_v070_native_tool_ui_zero_c_4af38131) | pipe |  |  |  |  |  | 2026-02-28 |
|
||||
| 24 | [🚀 GitHub Copilot SDK Pipe: AI That Executes, Not Just Talks](https://openwebui.com/posts/github_copilot_sdk_for_openwebui_elevate_your_ai_t_a140f293) | pipe |  |  |  |  |  | 2026-02-10 |
|
||||
| 25 | [🚀 Open WebUI Prompt Plus: AI-Powered Prompt Manager](https://openwebui.com/posts/open_webui_prompt_plus_ai_powered_prompt_manager_s_15fa060e) | action |  |  |  |  |  | 2026-01-29 |
|
||||
| 25 | [🚀 Open WebUI Prompt Plus: AI-Powered Prompt Manager](https://openwebui.com/posts/open_webui_prompt_plus_ai_powered_prompt_manager_s_15fa060e) | action |  |  |  |  |  | 2026-01-28 |
|
||||
| 26 | [Review of Claude Haiku 4.5](https://openwebui.com/posts/review_of_claude_haiku_45_41b0db39) | review |  |  |  |  |  | 2026-01-14 |
|
||||
| 27 | [ 🛠️ Debug Open WebUI Plugins in Your Browser](https://openwebui.com/posts/debug_open_webui_plugins_in_your_browser_81bf7960) | action |  |  |  |  |  | 2026-01-10 |
|
||||
|
||||
@@ -1,13 +1,19 @@
|
||||
# Async Context Compression Filter
|
||||
|
||||
**Author:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **Version:** 1.4.1 | **Project:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **License:** MIT
|
||||
**Author:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **Version:** 1.5.0 | **Project:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **License:** MIT
|
||||
|
||||
This filter reduces token consumption in long conversations through intelligent summarization and message compression while keeping conversations coherent.
|
||||
|
||||
## What's new in 1.4.1
|
||||
## What's new in 1.5.0
|
||||
|
||||
- **Reverse-Unfolding Mechanism**: Accurately reconstructs the expanded native tool-calling sequence during the outlet phase to permanently fix coordinate drift and missing summaries for long tool-based conversations.
|
||||
- **Safer Tool Trimming**: Refactored `enable_tool_output_trimming` to strictly use atomic block groups for safe trimming, completely preventing JSON payload corruption.
|
||||
- **External Chat Reference Summaries**: Added support for referenced chat context blocks that can reuse cached summaries, inject small referenced chats directly, or generate summaries for larger referenced chats before injection.
|
||||
- **Fast Multilingual Token Estimation**: Added a new mixed-script token estimation pipeline so inlet/outlet preflight checks can avoid unnecessary exact token counts while staying much closer to real usage.
|
||||
- **Stronger Working-Memory Prompt**: Refined the XML summary prompt to better preserve actionable context across general chat, coding tasks, and tool-heavy conversations.
|
||||
- **Clearer Frontend Debug Logs**: Reworked browser-console logging into grouped structural snapshots that are easier to scan during debugging.
|
||||
- **Safer Tool Trimming Defaults**: Enabled native tool-output trimming by default and exposed a dedicated `tool_trim_threshold_chars` valve with a 600-character default.
|
||||
- **Safer Referenced-Chat Fallbacks**: If generating a referenced chat summary fails, the new reference-summary path now falls back to direct contextual injection instead of failing the whole chat.
|
||||
- **Correct Summary Budgeting**: `summary_model_max_context` now controls summary-input fitting, while `max_summary_tokens` remains an output-length cap.
|
||||
- **More Visible Summary Failures**: Important background summary failures now surface in the browser console (`F12`) and as a status hint even when `show_debug_log` is off.
|
||||
|
||||
---
|
||||
|
||||
@@ -19,15 +25,85 @@ This filter reduces token consumption in long conversations through intelligent
|
||||
- ✅ Persistent storage via Open WebUI's shared database connection (PostgreSQL, SQLite, etc.).
|
||||
- ✅ Flexible retention policy to keep the first and last N messages.
|
||||
- ✅ Smart injection of historical summaries back into the context.
|
||||
- ✅ External chat reference summarization with cached-summary reuse, direct injection for small chats, and generated summaries for larger chats.
|
||||
- ✅ Structure-aware trimming that preserves document structure (headers, intro, conclusion).
|
||||
- ✅ Native tool output trimming for cleaner context when using function calling.
|
||||
- ✅ Real-time context usage monitoring with warning notifications (>90%).
|
||||
- ✅ Detailed token logging for precise debugging and optimization.
|
||||
- ✅ Fast multilingual token estimation plus exact token fallback for precise debugging and optimization.
|
||||
- ✅ **Smart Model Matching**: Automatically inherits configuration from base models for custom presets.
|
||||
- ⚠ **Multimodal Support**: Images are preserved but their tokens are **NOT** calculated. Please adjust thresholds accordingly.
|
||||
|
||||
---
|
||||
|
||||
## What This Fixes
|
||||
|
||||
- **Problem 1: A referenced chat could break the current request.**
|
||||
Before, if the filter needed to summarize a referenced chat and that LLM call failed, the current chat could fail with it. Now it degrades gracefully and injects direct context instead.
|
||||
- **Problem 2: Some referenced chats were being cut too aggressively.**
|
||||
Before, the output limit (`max_summary_tokens`) could be treated like the input window, which made large referenced chats shrink earlier than necessary. Now input fitting uses the summary model's real context window (`summary_model_max_context` or model/global fallback).
|
||||
- **Problem 3: Some background summary failures were too easy to miss.**
|
||||
Before, a failure during background summary preparation could disappear quietly when frontend debug logging was off. Now important failures are forced to the browser console and also shown through a user-facing status message.
|
||||
|
||||
---
|
||||
|
||||
## Workflow Overview
|
||||
|
||||
This filter operates in two phases:
|
||||
|
||||
1. `inlet`: injects stored summaries, processes external chat references, and trims context when required before the request is sent to the model.
|
||||
2. `outlet`: runs asynchronously after the response is complete, decides whether a new summary should be generated, and persists it when appropriate.
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
A[Request enters inlet] --> B[Normalize tool IDs and optionally trim large tool outputs]
|
||||
B --> C{Referenced chats attached?}
|
||||
C -- No --> D[Load current chat summary if available]
|
||||
C -- Yes --> E[Inspect each referenced chat]
|
||||
|
||||
E --> F{Existing cached summary?}
|
||||
F -- Yes --> G[Reuse cached summary]
|
||||
F -- No --> H{Fits direct budget?}
|
||||
H -- Yes --> I[Inject full referenced chat text]
|
||||
H -- No --> J[Prepare referenced-chat summary input]
|
||||
|
||||
J --> K{Referenced-chat summary call succeeds?}
|
||||
K -- Yes --> L[Inject generated referenced summary]
|
||||
K -- No --> M[Fallback to direct contextual injection]
|
||||
|
||||
G --> D
|
||||
I --> D
|
||||
L --> D
|
||||
M --> D
|
||||
|
||||
D --> N[Build current-chat Head + Summary + Tail]
|
||||
N --> O{Over max_context_tokens?}
|
||||
O -- Yes --> P[Trim oldest atomic groups]
|
||||
O -- No --> Q[Send final context to the model]
|
||||
P --> Q
|
||||
|
||||
Q --> R[Model returns the reply]
|
||||
R --> S[Outlet rebuilds the full history]
|
||||
S --> T{Reached compression threshold?}
|
||||
T -- No --> U[Finish]
|
||||
T -- Yes --> V[Fit summary input to the summary model context]
|
||||
|
||||
V --> W{Background summary call succeeds?}
|
||||
W -- Yes --> X[Save new chat summary and update status]
|
||||
W -- No --> Y[Force browser-console error and show status hint]
|
||||
```
|
||||
|
||||
### Key Notes
|
||||
|
||||
- `inlet` only injects and trims context. It does not generate the main chat summary.
|
||||
- `outlet` performs summary generation asynchronously and does not block the current reply.
|
||||
- External chat references may come from an existing persisted summary, a small chat's full text, or a generated/truncated reference summary.
|
||||
- If a referenced-chat summary call fails, the filter falls back to direct context injection instead of failing the whole request.
|
||||
- `summary_model_max_context` controls summary-input fitting. `max_summary_tokens` only controls how long the generated summary may be.
|
||||
- Important background summary failures are surfaced to the browser console (`F12`) and the chat status area.
|
||||
- External reference messages are protected during trimming so they are not discarded first.
|
||||
|
||||
---
|
||||
|
||||
## Installation & Configuration
|
||||
|
||||
### 1) Database (automatic)
|
||||
@@ -51,11 +127,12 @@ This filter reduces token consumption in long conversations through intelligent
|
||||
| `keep_first` | `1` | Always keep the first N messages (protects system prompts). |
|
||||
| `keep_last` | `6` | Always keep the last N messages to preserve recent context. |
|
||||
| `summary_model` | `None` | Model for summaries. Strongly recommended to set a fast, economical model (e.g., `gemini-2.5-flash`, `deepseek-v3`). Falls back to the current chat model when empty. |
|
||||
| `summary_model_max_context` | `0` | Max context tokens for the summary model. If 0, falls back to `model_thresholds` or global `max_context_tokens`. |
|
||||
| `max_summary_tokens` | `16384` | Maximum tokens for the generated summary. |
|
||||
| `summary_temperature` | `0.3` | Randomness for summary generation. Lower is more deterministic. |
|
||||
| `summary_model_max_context` | `0` | Input context window used to fit summary requests. If `0`, falls back to `model_thresholds` or global `max_context_tokens`. |
|
||||
| `max_summary_tokens` | `16384` | Maximum output length for the generated summary. This is not the summary-input context limit. |
|
||||
| `summary_temperature` | `0.1` | Randomness for summary generation. Lower is more deterministic. |
|
||||
| `model_thresholds` | `{}` | Per-model overrides for `compression_threshold_tokens` and `max_context_tokens` (useful for mixed models). |
|
||||
| `enable_tool_output_trimming` | `false` | When enabled and `function_calling: "native"` is active, trims verbose tool outputs to extract only the final answer. |
|
||||
| `enable_tool_output_trimming` | `true` | When enabled for `function_calling: "native"`, trims oversized native tool outputs while keeping the tool-call chain intact. |
|
||||
| `tool_trim_threshold_chars` | `600` | Trim native tool output blocks once their total content length reaches this threshold. |
|
||||
| `debug_mode` | `false` | Log verbose debug info. Set to `false` in production. |
|
||||
| `show_debug_log` | `false` | Print debug logs to browser console (F12). Useful for frontend debugging. |
|
||||
| `show_token_usage_status` | `true` | Show token usage status notification in the chat interface. |
|
||||
@@ -71,8 +148,12 @@ If this plugin has been useful, a star on [OpenWebUI Extensions](https://github.
|
||||
|
||||
- **Initial system prompt is lost**: Keep `keep_first` greater than 0 to protect the initial message.
|
||||
- **Compression effect is weak**: Raise `compression_threshold_tokens` or lower `keep_first` / `keep_last` to allow more aggressive compression.
|
||||
- **A referenced chat summary fails**: The current request should continue with a direct-context fallback. Check the browser console (`F12`) if you need the upstream failure details.
|
||||
- **A background summary silently seems to do nothing**: Important failures now surface in chat status and the browser console (`F12`).
|
||||
- **Submit an Issue**: If you encounter any problems, please submit an issue on GitHub: [OpenWebUI Extensions Issues](https://github.com/Fu-Jie/openwebui-extensions/issues)
|
||||
|
||||
## Changelog
|
||||
|
||||
See [`v1.5.0` Release Notes](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/v1.5.0.md) for the release-specific summary.
|
||||
|
||||
See the full history on GitHub: [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions)
|
||||
|
||||
@@ -1,15 +1,21 @@
|
||||
# 异步上下文压缩过滤器
|
||||
|
||||
**作者:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **版本:** 1.4.1 | **项目:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **许可证:** MIT
|
||||
**作者:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **版本:** 1.5.0 | **项目:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **许可证:** MIT
|
||||
|
||||
> **重要提示**：为了确保所有过滤器的可维护性和易用性，每个过滤器都应附带清晰、完整的文档，以确保其功能、配置和使用方法得到充分说明。
|
||||
|
||||
本过滤器通过智能摘要和消息压缩技术，在保持对话连贯性的同时，显著降低长对话的 Token 消耗。
|
||||
|
||||
## 1.4.1 版本更新
|
||||
## 1.5.0 版本更新
|
||||
|
||||
- **逆向展开机制**: 引入 `_unfold_messages` 机制以在 `outlet` 阶段精确对齐坐标系，彻底解决了由于前端视图折叠导致长轮次工具调用对话出现进度漂移或跳过生成摘要的问题。
|
||||
- **更安全的工具内容裁剪**: 重构了 `enable_tool_output_trimming`，现在严格使用原子级分组进行安全的原生工具内容裁剪，替代了激进的正则表达式匹配，防止 JSON 载荷损坏。
|
||||
- **外部聊天引用摘要**: 新增对引用聊天上下文的摘要支持。现在可以复用缓存摘要、直接注入较小引用聊天，或先为较大的引用聊天生成摘要再注入。
|
||||
- **快速多语言 Token 预估**: 新增混合脚本 Token 预估链路，使 inlet / outlet 的预检可以减少不必要的精确计数，同时比旧的粗略字符比值更接近真实用量。
|
||||
- **更稳健的工作记忆提示词**: 重写 XML 摘要提示词，增强普通聊天、编码任务和连续工具调用场景下的关键信息保留能力。
|
||||
- **更清晰的前端调试日志**: 浏览器控制台日志改为分组化、结构化展示，排查上下文压缩行为更直观。
|
||||
- **更安全的工具裁剪默认值**: 原生工具输出裁剪默认开启，并新增 `tool_trim_threshold_chars` 配置项，默认阈值为 600 字符。
|
||||
- **更稳妥的引用聊天回退**: 当新的引用聊天摘要路径生成失败时，不再拖垮当前请求，而是自动回退为直接注入上下文。
|
||||
- **更准确的摘要预算**: `summary_model_max_context` 现在只负责摘要输入窗口，`max_summary_tokens` 继续只负责摘要输出长度。
|
||||
- **更容易发现摘要失败**: 重要的后台摘要失败现在会强制显示到浏览器控制台 (`F12`)，并同步给出状态提示。
|
||||
|
||||
---
|
||||
|
||||
@@ -21,14 +27,84 @@
|
||||
- ✅ **持久化存储**: 复用 Open WebUI 共享数据库连接，自动支持 PostgreSQL/SQLite 等。
|
||||
- ✅ **灵活保留策略**: 可配置保留对话头部和尾部消息，确保关键信息连贯。
|
||||
- ✅ **智能注入**: 将历史摘要智能注入到新上下文中。
|
||||
- ✅ **外部聊天引用摘要**: 支持复用缓存摘要、小聊天直接注入，以及大聊天先摘要后注入。
|
||||
- ✅ **结构感知裁剪**: 智能折叠过长消息，保留文档骨架（标题、首尾）。
|
||||
- ✅ **原生工具输出裁剪**: 支持裁剪冗长的工具调用输出。
|
||||
- ✅ **实时监控**: 实时监控上下文使用情况，超过 90% 发出警告。
|
||||
- ✅ **详细日志**: 提供精确的 Token 统计日志，便于调试。
|
||||
- ✅ **快速预估 + 精确回退**: 提供更快的多语言 Token 预估，并在必要时回退到精确统计，便于调试。
|
||||
- ✅ **智能模型匹配**: 自定义模型自动继承基础模型的阈值配置。
|
||||
- ⚠ **多模态支持**: 图片内容会被保留，但其 Token **不参与计算**。请相应调整阈值。
|
||||
|
||||
详细的工作原理和流程请参考 [工作流程指南](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/WORKFLOW_GUIDE_CN.md)。
|
||||
详细的工作原理和更长说明仍可参考 [工作流程指南](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/WORKFLOW_GUIDE_CN.md)。
|
||||
|
||||
---
|
||||
|
||||
## 这次解决了什么问题（通俗版）
|
||||
|
||||
- **问题 1：引用别的聊天时，摘要失败可能把当前对话一起弄挂。**
|
||||
以前如果过滤器需要先帮被引用聊天做摘要，而这一步的 LLM 调用失败了，当前请求也可能直接失败。现在改成了“能摘要就摘要，失败就退回直接塞上下文”，当前对话不会被一起拖死。
|
||||
- **问题 2：有些被引用聊天被截得太早，信息丢得太多。**
|
||||
以前有一段逻辑把 `max_summary_tokens` 这种“输出长度限制”误当成了“输入上下文窗口”，结果大一点的引用聊天会被过早截断。现在改成按摘要模型真实的输入窗口来算，能保留更多有用内容。
|
||||
- **问题 3：后台摘要失败时，用户不容易知道发生了什么。**
|
||||
以前在 `show_debug_log=false` 时，有些后台失败只会留在内部日志里。现在关键失败会强制打到浏览器控制台，并在聊天状态里提醒去看 `F12`。
|
||||
|
||||
---
|
||||
|
||||
## 工作流总览
|
||||
|
||||
该过滤器分为两个阶段：
|
||||
|
||||
1. `inlet`：在请求发送给模型前执行，负责注入已有摘要、处理外部聊天引用、并在必要时裁剪上下文。
|
||||
2. `outlet`：在模型回复完成后异步执行，负责判断是否需要生成新摘要，并在合适时写入数据库。
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
A[请求进入 inlet] --> B[规范化工具 ID 并按需裁剪超长工具输出]
|
||||
B --> C{是否附带引用聊天?}
|
||||
C -- 否 --> D[如果有当前聊天摘要就先加载]
|
||||
C -- 是 --> E[逐个检查被引用聊天]
|
||||
|
||||
E --> F{已有缓存摘要?}
|
||||
F -- 是 --> G[直接复用缓存摘要]
|
||||
F -- 否 --> H{能直接放进当前预算?}
|
||||
H -- 是 --> I[直接注入完整引用聊天文本]
|
||||
H -- 否 --> J[准备引用聊天的摘要输入]
|
||||
|
||||
J --> K{引用聊天摘要调用成功?}
|
||||
K -- 是 --> L[注入生成后的引用摘要]
|
||||
K -- 否 --> M[回退为直接注入上下文]
|
||||
|
||||
G --> D
|
||||
I --> D
|
||||
L --> D
|
||||
M --> D
|
||||
|
||||
D --> N[为当前聊天构造 Head + Summary + Tail]
|
||||
N --> O{是否超过 max_context_tokens?}
|
||||
O -- 是 --> P[从最旧 atomic groups 开始裁剪]
|
||||
O -- 否 --> Q[把最终上下文发给模型]
|
||||
P --> Q
|
||||
|
||||
Q --> R[模型返回当前回复]
|
||||
R --> S[Outlet 重建完整历史]
|
||||
S --> T{达到压缩阈值了吗?}
|
||||
T -- 否 --> U[结束]
|
||||
T -- 是 --> V[把摘要输入压到摘要模型可接受的上下文窗口]
|
||||
|
||||
V --> W{后台摘要调用成功?}
|
||||
W -- 是 --> X[保存新摘要并更新状态]
|
||||
W -- 否 --> Y[强制输出浏览器控制台错误并提示用户查看]
|
||||
```
|
||||
|
||||
### 关键说明
|
||||
|
||||
- `inlet` 只负责注入和裁剪上下文，不负责生成当前聊天的主摘要。
|
||||
- `outlet` 异步生成摘要，不会阻塞当前回复。
|
||||
- 外部聊天引用可以来自已有持久化摘要、小聊天的完整文本，或动态生成/截断后的引用摘要。
|
||||
- 如果引用聊天摘要失败，会自动回退为直接注入上下文，而不是让当前请求失败。
|
||||
- `summary_model_max_context` 控制摘要输入窗口；`max_summary_tokens` 只控制生成摘要的输出长度。
|
||||
- 重要的后台摘要失败会显示到浏览器控制台 (`F12`) 和聊天状态提示里。
|
||||
- 外部引用消息在裁剪阶段会被特殊保护，避免被最先删除。
|
||||
|
||||
---
|
||||
|
||||
@@ -64,8 +140,8 @@
|
||||
| 参数 | 默认值 | 描述 |
|
||||
| :-------------------- | :------ | :------------------------------------------------------------------------------------------------------------------------------------------ |
|
||||
| `summary_model` | `None` | 用于生成摘要的模型 ID。**强烈建议**配置快速、经济、上下文窗口大的模型（如 `gemini-2.5-flash`、`deepseek-v3`）。留空则尝试复用当前对话模型。 |
|
||||
| `summary_model_max_context` | `0` | 摘要模型的最大上下文 Token 数。如果为 0，则回退到 `model_thresholds` 或全局 `max_context_tokens`。 |
|
||||
| `max_summary_tokens` | `16384` | 生成摘要时允许的最大 Token 数。 |
|
||||
| `summary_model_max_context` | `0` | 摘要请求可使用的输入上下文窗口。如果为 0，则回退到 `model_thresholds` 或全局 `max_context_tokens`。 |
|
||||
| `max_summary_tokens` | `16384` | 生成摘要时允许的最大输出 Token 数。它不是摘要输入窗口上限。 |
|
||||
| `summary_temperature` | `0.1` | 控制摘要生成的随机性，较低的值结果更稳定。 |
|
||||
|
||||
### 高级配置
|
||||
@@ -93,7 +169,8 @@
|
||||
|
||||
| 参数 | 默认值 | 描述 |
|
||||
| :----------------------------- | :------- | :-------------------------------------------------------------------------------------------------------------------------------------- |
|
||||
| `enable_tool_output_trimming` | `false` | 启用时，若 `function_calling: "native"` 激活，将裁剪冗长的工具输出以仅提取最终答案。 |
|
||||
| `enable_tool_output_trimming` | `true` | 启用后（仅在 `function_calling: "native"` 下生效）会裁剪过大的本机工具输出，保留工具调用链结构并以简短占位替换冗长内容。 |
|
||||
| `tool_trim_threshold_chars` | `600` | 当本机工具输出累计字符数达到该值时触发裁剪，适用于包含长文本或表格的工具结果。 |
|
||||
| `debug_mode` | `false` | 是否在 Open WebUI 的控制台日志中打印详细的调试信息。生产环境默认且建议设为 `false`。 |
|
||||
| `show_debug_log` | `false` | 是否在浏览器控制台 (F12) 打印调试日志。便于前端调试。 |
|
||||
| `show_token_usage_status` | `true` | 是否在对话结束时显示 Token 使用情况的状态通知。 |
|
||||
@@ -109,8 +186,12 @@
|
||||
|
||||
- **初始系统提示丢失**：将 `keep_first` 设置为大于 0。
|
||||
- **压缩效果不明显**：提高 `compression_threshold_tokens`，或降低 `keep_first` / `keep_last` 以增强压缩力度。
|
||||
- **引用聊天摘要失败**：当前请求现在应该会继续执行，并回退为直接注入上下文。如果要看上游失败原因，请打开浏览器控制台 (`F12`)。
|
||||
- **后台摘要看起来“没反应”**：重要失败现在会同时出现在状态提示和浏览器控制台 (`F12`) 中。
|
||||
- **提交 Issue**: 如果遇到任何问题，请在 GitHub 上提交 Issue：[OpenWebUI Extensions Issues](https://github.com/Fu-Jie/openwebui-extensions/issues)
|
||||
|
||||
## 更新日志
|
||||
|
||||
请查看 [`v1.5.0` 版本发布说明](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/v1.5.0_CN.md) 获取本次版本的独立发布摘要。
|
||||
|
||||
完整历史请查看 GitHub 项目： [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions)
|
||||
|
||||
@@ -20,9 +20,9 @@ Filters act as middleware in the message pipeline:
|
||||
|
||||
---
|
||||
|
||||
Reduces token consumption in long conversations through intelligent summarization while maintaining coherence.
|
||||
Reduces token consumption in long conversations with safer summary fallbacks and clearer failure visibility.
|
||||
|
||||
**Version:** 1.4.1
|
||||
**Version:** 1.5.0
|
||||
|
||||
[:octicons-arrow-right-24: Documentation](async-context-compression.md)
|
||||
|
||||
|
||||
@@ -20,11 +20,11 @@ Filter 充当消息管线中的中间件：
|
||||
|
||||
---
|
||||
|
||||
通过智能总结减少长对话的 token 消耗，同时保持连贯性。
|
||||
通过更稳健的摘要回退和更清晰的失败提示，降低长对话的 token 消耗并保持连贯性。
|
||||
|
||||
**版本：** 1.4.1
|
||||
**版本：** 1.5.0
|
||||
|
||||
[:octicons-arrow-right-24: 查看文档](async-context-compression.md)
|
||||
[:octicons-arrow-right-24: 查看文档](async-context-compression.zh.md)
|
||||
|
||||
- :material-text-box-plus:{ .lg .middle } **Context Enhancement**
|
||||
|
||||
|
||||
139
docs/plugins/tools/batch-install-plugins-tool.md
Normal file
139
docs/plugins/tools/batch-install-plugins-tool.md
Normal file
@@ -0,0 +1,139 @@
|
||||
# Batch Install Plugins from GitHub
|
||||
|
||||
**Author:** [Fu-Jie](https://github.com/Fu-Jie) | **Version:** 1.0.0 | **Project:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **License:** MIT
|
||||
|
||||
---
|
||||
|
||||
One-click batch install plugins from GitHub repositories to your OpenWebUI instance.
|
||||
|
||||
## Key Features
|
||||
|
||||
- **One-Click Install**: Install all plugins with a single command
|
||||
- **Auto-Update**: Automatically updates previously installed plugins
|
||||
- **GitHub Support**: Install plugins from any GitHub repository
|
||||
- **Multi-Type Support**: Supports Pipe, Action, Filter, and Tool plugins
|
||||
- **Confirmation**: Shows plugin list before installing, allows selective installation
|
||||
- **i18n**: Supports 11 languages
|
||||
|
||||
## Flow
|
||||
|
||||
```
|
||||
User Input
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Discover Plugins from GitHub │
|
||||
│ (fetch file tree + parse .py) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Filter by Type & Keywords │
|
||||
│ (tool/filter/pipe/action) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Show Confirmation Dialog │
|
||||
│ (list plugins + exclude hint) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
├── [Cancel] → End
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Install to OpenWebUI │
|
||||
│ (update or create each plugin) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
Done
|
||||
```
|
||||
|
||||
## How to Use
|
||||
|
||||
1. Open OpenWebUI and go to **Workspace > Tools**
|
||||
2. Install **Batch Install Plugins from GitHub** from the official marketplace
|
||||
3. Enable this tool for your model/chat
|
||||
4. Ask the model to install plugins
|
||||
|
||||
## Usage Examples
|
||||
|
||||
```
|
||||
"Install all plugins"
|
||||
"Install all plugins from github.com/username/repo"
|
||||
"Install only pipe plugins"
|
||||
"Install action and filter plugins"
|
||||
"Install all plugins, exclude_keywords=copilot"
|
||||
```
|
||||
|
||||
## Popular Plugin Repositories
|
||||
|
||||
Here are some popular repositories with many plugins you can install:
|
||||
|
||||
### Community Collections
|
||||
|
||||
```
|
||||
# Install all plugins from iChristGit's collection
|
||||
"Install all plugins from iChristGit/OpenWebui-Tools"
|
||||
|
||||
# Install all tools from Haervwe's tools collection
|
||||
"Install all plugins from Haervwe/open-webui-tools"
|
||||
|
||||
# Install all plugins from Classic298's repository
|
||||
"Install all plugins from Classic298/open-webui-plugins"
|
||||
|
||||
# Install all functions from suurt8ll's collection
|
||||
"Install all plugins from suurt8ll/open_webui_functions"
|
||||
|
||||
# Install only specific types (e.g., only tools)
|
||||
"Install only tool plugins from iChristGit/OpenWebui-Tools"
|
||||
|
||||
# Exclude certain keywords while installing
|
||||
"Install all plugins from Haervwe/open-webui-tools, exclude_keywords=test,deprecated"
|
||||
```
|
||||
|
||||
### Supported Repositories
|
||||
|
||||
- `Fu-Jie/openwebui-extensions` - Default, official plugin collection
|
||||
- `iChristGit/OpenWebui-Tools` - Comprehensive tool and plugin collection
|
||||
- `Haervwe/open-webui-tools` - Specialized tools and utilities
|
||||
- `Classic298/open-webui-plugins` - Various plugin implementations
|
||||
- `suurt8ll/open_webui_functions` - Function-based plugins
|
||||
|
||||
## Default Repository
|
||||
|
||||
When no repository is specified, defaults to `Fu-Jie/openwebui-extensions`.
|
||||
|
||||
## Plugin Detection Rules
|
||||
|
||||
### Fu-Jie/openwebui-extensions (Strict)
|
||||
|
||||
For the default repository, plugins must have:
|
||||
1. A `.py` file containing `class Tools:`, `class Filter:`, `class Pipe:`, or `class Action:`
|
||||
2. A docstring with `title:`, `description:`, and **`openwebui_id:`** fields
|
||||
3. Filename must not end with `_cn`
|
||||
|
||||
### Other GitHub Repositories
|
||||
|
||||
For other repositories:
|
||||
1. A `.py` file containing `class Tools:`, `class Filter:`, `class Pipe:`, or `class Action:`
|
||||
2. A docstring with `title:` and `description:` fields
|
||||
|
||||
## Configuration (Valves)
|
||||
|
||||
| Parameter | Default | Description |
|
||||
| --- | --- | --- |
|
||||
| `SKIP_KEYWORDS` | `test,verify,example,template,mock` | Comma-separated keywords to skip |
|
||||
| `TIMEOUT` | `20` | Request timeout in seconds |
|
||||
|
||||
## Confirmation Timeout
|
||||
|
||||
User confirmation dialogs have a default timeout of **2 minutes (120 seconds)**, allowing sufficient time for users to:
|
||||
- Read and review the plugin list
|
||||
- Make installation decisions
|
||||
- Handle network delays
|
||||
|
||||
## Support
|
||||
|
||||
If this plugin has been useful, a star on [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) is a big motivation for me. Thank you for the support.
|
||||
139
docs/plugins/tools/batch-install-plugins-tool.zh.md
Normal file
139
docs/plugins/tools/batch-install-plugins-tool.zh.md
Normal file
@@ -0,0 +1,139 @@
|
||||
# Batch Install Plugins from GitHub - 从 GitHub 批量安装插件
|
||||
|
||||
**作者:** [Fu-Jie](https://github.com/Fu-Jie) | **版本:** 1.0.0 | **项目:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **许可:** MIT
|
||||
|
||||
---
|
||||
|
||||
一键从 GitHub 仓库批量安装插件到你的 OpenWebUI 实例。
|
||||
|
||||
## ✨ 主要特性
|
||||
|
||||
- **一键安装**: 一条命令安装所有插件
|
||||
- **自动更新**: 自动更新之前已安装的插件
|
||||
- **GitHub 支持**: 支持从任何 GitHub 仓库安装插件
|
||||
- **多类型支持**: 支持 Pipe、Action、Filter 和 Tool 插件
|
||||
- **确认机制**: 安装前显示插件列表，允许选择性安装
|
||||
- **国际化**: 支持 11 种语言
|
||||
|
||||
## 工作流
|
||||
|
||||
```
|
||||
用户输入
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 从 GitHub 发现插件 │
|
||||
│ (获取文件树 + 解析 .py) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 按类型和关键词过滤 │
|
||||
│ (tool/filter/pipe/action) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 显示确认对话框 │
|
||||
│ (插件列表 + 排除提示) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
├── [取消] → 结束
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 安装到 OpenWebUI │
|
||||
│ (更新或创建每个插件) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
完成
|
||||
```
|
||||
|
||||
## 🚀 使用方法
|
||||
|
||||
1. 打开 OpenWebUI，进入 **工作区 > 工具**
|
||||
2. 从官方市场安装 **Batch Install Plugins from GitHub**
|
||||
3. 为你的模型/聊天启用此工具
|
||||
4. 让模型安装插件
|
||||
|
||||
## 使用示例
|
||||
|
||||
```
|
||||
"安装所有插件"
|
||||
"从 github.com/username/repo 安装所有插件"
|
||||
"仅安装 pipe 插件"
|
||||
"安装 action 和 filter 插件"
|
||||
"安装所有插件，exclude_keywords=copilot"
|
||||
```
|
||||
|
||||
## 热门插件仓库
|
||||
|
||||
这些是包含大量插件的热门仓库，你可以从中安装插件：
|
||||
|
||||
### 社区合集
|
||||
|
||||
```
|
||||
# 从 iChristGit 的集合安装所有插件
|
||||
"从 iChristGit/OpenWebui-Tools 安装所有插件"
|
||||
|
||||
# 从 Haervwe 的工具集合只安装工具
|
||||
"从 Haervwe/open-webui-tools 安装所有插件"
|
||||
|
||||
# 从 Classic298 的仓库安装所有插件
|
||||
"从 Classic298/open-webui-plugins 安装所有插件"
|
||||
|
||||
# 从 suurt8ll 的集合安装所有函数
|
||||
"从 suurt8ll/open_webui_functions 安装所有插件"
|
||||
|
||||
# 仅安装特定类型的插件（比如只安装工具）
|
||||
"从 iChristGit/OpenWebui-Tools 仅安装 tool 插件"
|
||||
|
||||
# 安装时排除特定关键词
|
||||
"从 Haervwe/open-webui-tools 安装所有插件，exclude_keywords=test,deprecated"
|
||||
```
|
||||
|
||||
### 支持的仓库
|
||||
|
||||
- `Fu-Jie/openwebui-extensions` - 默认的官方插件集合
|
||||
- `iChristGit/OpenWebui-Tools` - 全面的工具和插件集合
|
||||
- `Haervwe/open-webui-tools` - 专业的工具和实用程序
|
||||
- `Classic298/open-webui-plugins` - 各种插件实现
|
||||
- `suurt8ll/open_webui_functions` - 基于函数的插件
|
||||
|
||||
## 默认仓库
|
||||
|
||||
未指定仓库时，默认使用 `Fu-Jie/openwebui-extensions`。
|
||||
|
||||
## 插件检测规则
|
||||
|
||||
### Fu-Jie/openwebui-extensions（严格模式）
|
||||
|
||||
对于默认仓库，插件必须有：
|
||||
1. 包含 `class Tools:`、`class Filter:`、`class Pipe:` 或 `class Action:` 的 `.py` 文件
|
||||
2. 包含 `title:`、`description:` 和 **`openwebui_id:`** 字段的文档字符串
|
||||
3. 文件名不能以 `_cn` 结尾
|
||||
|
||||
### 其他 GitHub 仓库
|
||||
|
||||
对于其他仓库：
|
||||
1. 包含 `class Tools:`、`class Filter:`、`class Pipe:` 或 `class Action:` 的 `.py` 文件
|
||||
2. 包含 `title:` 和 `description:` 字段的文档字符串
|
||||
|
||||
## 配置 (Valves)
|
||||
|
||||
| 参数 | 默认值 | 描述 |
|
||||
| --- | --- | --- |
|
||||
| `SKIP_KEYWORDS` | `test,verify,example,template,mock` | 要跳过的关键词，用逗号分隔 |
|
||||
| `TIMEOUT` | `20` | 请求超时时间（秒） |
|
||||
|
||||
## 确认超时时间
|
||||
|
||||
用户确认对话框的默认超时时间为 **2 分钟（120 秒）**，为用户提供充足的时间来：
|
||||
- 阅读和查看插件列表
|
||||
- 做出安装决定
|
||||
- 处理网络延迟
|
||||
|
||||
## 支持
|
||||
|
||||
如果这个插件对你有帮助，欢迎到 [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) 点个 Star，这将是我持续改进的动力，感谢支持。
|
||||
@@ -4,5 +4,6 @@ OpenWebUI native Tool plugins that can be used across models.
|
||||
|
||||
## Available Tool Plugins
|
||||
|
||||
- [Batch Install Plugins from GitHub](batch-install-plugins-tool.md) (v1.0.0) - One-click batch install plugins from GitHub repositories with confirmation and multi-language support.
|
||||
- [OpenWebUI Skills Manager Tool](openwebui-skills-manager-tool.md) (v0.3.0) - Simple native skill management (`list/show/install/create/update/delete`).
|
||||
- [Smart Mind Map Tool](smart-mind-map-tool.md) (v1.0.0) - Intelligently analyzes text content and proactively generates interactive mind maps to help users structure and visualize knowledge.
|
||||
|
||||
@@ -4,5 +4,6 @@
|
||||
|
||||
## 可用 Tool 插件
|
||||
|
||||
- [Batch Install Plugins from GitHub](batch-install-plugins-tool.zh.md) (v1.0.0) - 一键从 GitHub 仓库批量安装插件，支持确认和多语言。
|
||||
- [OpenWebUI Skills 管理工具](openwebui-skills-manager-tool.zh.md) (v0.3.0) - 简化技能管理（`list/show/install/create/update/delete`）。
|
||||
- [智能思维导图工具 (Smart Mind Map Tool)](smart-mind-map-tool.zh.md) (v1.0.0) - 智能分析文本内容并主动生成交互式思维导图，帮助用户结构化与可视化知识。
|
||||
|
||||
@@ -1,14 +1,19 @@
|
||||
# Async Context Compression Filter
|
||||
|
||||
**Author:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **Version:** 1.4.2 | **Project:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **License:** MIT
|
||||
**Author:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **Version:** 1.5.0 | **Project:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **License:** MIT
|
||||
|
||||
This filter reduces token consumption in long conversations through intelligent summarization and message compression while keeping conversations coherent.
|
||||
|
||||
## What's new in 1.4.2
|
||||
## What's new in 1.5.0
|
||||
|
||||
- **Enhanced Summary Path Robustness**: Thread `__request__` context through entire summary generation pipeline for reliable authentication and provider handling.
|
||||
- **Improved Error Diagnostics**: LLM response validation failures now include complete response body in error logs for transparent troubleshooting.
|
||||
- **Smart Previous Summary Loading**: Automatically load and merge previous summaries from DB when not present in outlet payload, enabling incremental state merging across summary generations.
|
||||
- **External Chat Reference Summaries**: Added support for referenced chat context blocks that can reuse cached summaries, inject small referenced chats directly, or generate summaries for larger referenced chats before injection.
|
||||
- **Fast Multilingual Token Estimation**: Added a new mixed-script token estimation pipeline so inlet/outlet preflight checks can avoid unnecessary exact token counts while staying much closer to real usage.
|
||||
- **Stronger Working-Memory Prompt**: Refined the XML summary prompt to better preserve actionable context across general chat, coding tasks, and tool-heavy conversations.
|
||||
- **Clearer Frontend Debug Logs**: Reworked browser-console logging into grouped structural snapshots that are easier to scan during debugging.
|
||||
- **Safer Tool Trimming Defaults**: Enabled native tool-output trimming by default and exposed a dedicated `tool_trim_threshold_chars` valve with a 600-character default.
|
||||
- **Safer Referenced-Chat Fallbacks**: If generating a referenced chat summary fails, the new reference-summary path now falls back to direct contextual injection instead of failing the whole chat.
|
||||
- **Correct Summary Budgeting**: `summary_model_max_context` now controls summary-input fitting, while `max_summary_tokens` remains an output-length cap.
|
||||
- **More Visible Summary Failures**: Important background summary failures now surface in the browser console (`F12`) and as a status hint even when `show_debug_log` is off.
|
||||
|
||||
---
|
||||
|
||||
@@ -20,15 +25,85 @@ This filter reduces token consumption in long conversations through intelligent
|
||||
- ✅ Persistent storage via Open WebUI's shared database connection (PostgreSQL, SQLite, etc.).
|
||||
- ✅ Flexible retention policy to keep the first and last N messages.
|
||||
- ✅ Smart injection of historical summaries back into the context.
|
||||
- ✅ External chat reference summarization with cached-summary reuse, direct injection for small chats, and generated summaries for larger chats.
|
||||
- ✅ Structure-aware trimming that preserves document structure (headers, intro, conclusion).
|
||||
- ✅ Native tool output trimming for cleaner context when using function calling.
|
||||
- ✅ Real-time context usage monitoring with warning notifications (>90%).
|
||||
- ✅ Detailed token logging for precise debugging and optimization.
|
||||
- ✅ Fast multilingual token estimation plus exact token fallback for precise debugging and optimization.
|
||||
- ✅ **Smart Model Matching**: Automatically inherits configuration from base models for custom presets.
|
||||
- ⚠ **Multimodal Support**: Images are preserved but their tokens are **NOT** calculated. Please adjust thresholds accordingly.
|
||||
|
||||
---
|
||||
|
||||
## What This Fixes
|
||||
|
||||
- **Problem 1: A referenced chat could break the current request.**
|
||||
Before, if the filter needed to summarize a referenced chat and that LLM call failed, the current chat could fail with it. Now it degrades gracefully and injects direct context instead.
|
||||
- **Problem 2: Some referenced chats were being cut too aggressively.**
|
||||
Before, the output limit (`max_summary_tokens`) could be treated like the input window, which made large referenced chats shrink earlier than necessary. Now input fitting uses the summary model's real context window (`summary_model_max_context` or model/global fallback).
|
||||
- **Problem 3: Some background summary failures were too easy to miss.**
|
||||
Before, a failure during background summary preparation could disappear quietly when frontend debug logging was off. Now important failures are forced to the browser console and also shown through a user-facing status message.
|
||||
|
||||
---
|
||||
|
||||
## Workflow Overview
|
||||
|
||||
This filter operates in two phases:
|
||||
|
||||
1. `inlet`: injects stored summaries, processes external chat references, and trims context when required before the request is sent to the model.
|
||||
2. `outlet`: runs asynchronously after the response is complete, decides whether a new summary should be generated, and persists it when appropriate.
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
A[Request enters inlet] --> B[Normalize tool IDs and optionally trim large tool outputs]
|
||||
B --> C{Referenced chats attached?}
|
||||
C -- No --> D[Load current chat summary if available]
|
||||
C -- Yes --> E[Inspect each referenced chat]
|
||||
|
||||
E --> F{Existing cached summary?}
|
||||
F -- Yes --> G[Reuse cached summary]
|
||||
F -- No --> H{Fits direct budget?}
|
||||
H -- Yes --> I[Inject full referenced chat text]
|
||||
H -- No --> J[Prepare referenced-chat summary input]
|
||||
|
||||
J --> K{Referenced-chat summary call succeeds?}
|
||||
K -- Yes --> L[Inject generated referenced summary]
|
||||
K -- No --> M[Fallback to direct contextual injection]
|
||||
|
||||
G --> D
|
||||
I --> D
|
||||
L --> D
|
||||
M --> D
|
||||
|
||||
D --> N[Build current-chat Head + Summary + Tail]
|
||||
N --> O{Over max_context_tokens?}
|
||||
O -- Yes --> P[Trim oldest atomic groups]
|
||||
O -- No --> Q[Send final context to the model]
|
||||
P --> Q
|
||||
|
||||
Q --> R[Model returns the reply]
|
||||
R --> S[Outlet rebuilds the full history]
|
||||
S --> T{Reached compression threshold?}
|
||||
T -- No --> U[Finish]
|
||||
T -- Yes --> V[Fit summary input to the summary model context]
|
||||
|
||||
V --> W{Background summary call succeeds?}
|
||||
W -- Yes --> X[Save new chat summary and update status]
|
||||
W -- No --> Y[Force browser-console error and show status hint]
|
||||
```
|
||||
|
||||
### Key Notes
|
||||
|
||||
- `inlet` only injects and trims context. It does not generate the main chat summary.
|
||||
- `outlet` performs summary generation asynchronously and does not block the current reply.
|
||||
- External chat references may come from an existing persisted summary, a small chat's full text, or a generated/truncated reference summary.
|
||||
- If a referenced-chat summary call fails, the filter falls back to direct context injection instead of failing the whole request.
|
||||
- `summary_model_max_context` controls summary-input fitting. `max_summary_tokens` only controls how long the generated summary may be.
|
||||
- Important background summary failures are surfaced to the browser console (`F12`) and the chat status area.
|
||||
- External reference messages are protected during trimming so they are not discarded first.
|
||||
|
||||
---
|
||||
|
||||
## Installation & Configuration
|
||||
|
||||
### 1) Database (automatic)
|
||||
@@ -52,11 +127,12 @@ This filter reduces token consumption in long conversations through intelligent
|
||||
| `keep_first` | `1` | Always keep the first N messages (protects system prompts). |
|
||||
| `keep_last` | `6` | Always keep the last N messages to preserve recent context. |
|
||||
| `summary_model` | `None` | Model for summaries. Strongly recommended to set a fast, economical model (e.g., `gemini-2.5-flash`, `deepseek-v3`). Falls back to the current chat model when empty. |
|
||||
| `summary_model_max_context` | `0` | Max context tokens for the summary model. If 0, falls back to `model_thresholds` or global `max_context_tokens`. |
|
||||
| `max_summary_tokens` | `16384` | Maximum tokens for the generated summary. |
|
||||
| `summary_temperature` | `0.3` | Randomness for summary generation. Lower is more deterministic. |
|
||||
| `summary_model_max_context` | `0` | Input context window used to fit summary requests. If `0`, falls back to `model_thresholds` or global `max_context_tokens`. |
|
||||
| `max_summary_tokens` | `16384` | Maximum output length for the generated summary. This is not the summary-input context limit. |
|
||||
| `summary_temperature` | `0.1` | Randomness for summary generation. Lower is more deterministic. |
|
||||
| `model_thresholds` | `{}` | Per-model overrides for `compression_threshold_tokens` and `max_context_tokens` (useful for mixed models). |
|
||||
| `enable_tool_output_trimming` | `false` | When enabled and `function_calling: "native"` is active, trims verbose tool outputs to extract only the final answer. |
|
||||
| `enable_tool_output_trimming` | `true` | When enabled for `function_calling: "native"`, trims oversized native tool outputs while keeping the tool-call chain intact. |
|
||||
| `tool_trim_threshold_chars` | `600` | Trim native tool output blocks once their total content length reaches this threshold. |
|
||||
| `debug_mode` | `false` | Log verbose debug info. Set to `false` in production. |
|
||||
| `show_debug_log` | `false` | Print debug logs to browser console (F12). Useful for frontend debugging. |
|
||||
| `show_token_usage_status` | `true` | Show token usage status notification in the chat interface. |
|
||||
@@ -72,8 +148,12 @@ If this plugin has been useful, a star on [OpenWebUI Extensions](https://github.
|
||||
|
||||
- **Initial system prompt is lost**: Keep `keep_first` greater than 0 to protect the initial message.
|
||||
- **Compression effect is weak**: Raise `compression_threshold_tokens` or lower `keep_first` / `keep_last` to allow more aggressive compression.
|
||||
- **A referenced chat summary fails**: The current request should continue with a direct-context fallback. Check the browser console (`F12`) if you need the upstream failure details.
|
||||
- **A background summary silently seems to do nothing**: Important failures now surface in chat status and the browser console (`F12`).
|
||||
- **Submit an Issue**: If you encounter any problems, please submit an issue on GitHub: [OpenWebUI Extensions Issues](https://github.com/Fu-Jie/openwebui-extensions/issues)
|
||||
|
||||
## Changelog
|
||||
|
||||
See [`v1.5.0` Release Notes](./v1.5.0.md) for the release-specific summary.
|
||||
|
||||
See the full history on GitHub: [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions)
|
||||
|
||||
@@ -1,16 +1,21 @@
|
||||
# 异步上下文压缩过滤器
|
||||
|
||||
**作者:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **版本:** 1.4.2 | **项目:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **许可证:** MIT
|
||||
**作者:** [Fu-Jie](https://github.com/Fu-Jie/openwebui-extensions) | **版本:** 1.5.0 | **项目:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) | **许可证:** MIT
|
||||
|
||||
> **重要提示**：为了确保所有过滤器的可维护性和易用性，每个过滤器都应附带清晰、完整的文档，以确保其功能、配置和使用方法得到充分说明。
|
||||
|
||||
本过滤器通过智能摘要和消息压缩技术，在保持对话连贯性的同时，显著降低长对话的 Token 消耗。
|
||||
|
||||
## 1.4.2 版本更新
|
||||
## 1.5.0 版本更新
|
||||
|
||||
- **强化摘要路径健壮性**: 在整个摘要生成管道中透传 `__request__` 上下文，确保认证和提供商处理的可靠性。
|
||||
- **改进错误诊断**: LLM 响应校验失败时，错误日志现在包含完整的响应体，便于透明的故障排除。
|
||||
- **智能旧摘要加载**: 当 outlet payload 中缺失摘要消息时，自动从 DB 加载并合并前一代摘要，实现增量式状态合并。
|
||||
- **外部聊天引用摘要**: 新增对引用聊天上下文的摘要支持。现在可以复用缓存摘要、直接注入较小引用聊天，或先为较大的引用聊天生成摘要再注入。
|
||||
- **快速多语言 Token 预估**: 新增混合脚本 Token 预估链路，使 inlet / outlet 的预检可以减少不必要的精确计数，同时比旧的粗略字符比值更接近真实用量。
|
||||
- **更稳健的工作记忆提示词**: 重写 XML 摘要提示词，增强普通聊天、编码任务和连续工具调用场景下的关键信息保留能力。
|
||||
- **更清晰的前端调试日志**: 浏览器控制台日志改为分组化、结构化展示，排查上下文压缩行为更直观。
|
||||
- **更安全的工具裁剪默认值**: 原生工具输出裁剪默认开启，并新增 `tool_trim_threshold_chars` 配置项，默认阈值为 600 字符。
|
||||
- **更稳妥的引用聊天回退**: 当新的引用聊天摘要路径生成失败时，不再拖垮当前请求，而是自动回退为直接注入上下文。
|
||||
- **更准确的摘要预算**: `summary_model_max_context` 现在只负责摘要输入窗口，`max_summary_tokens` 继续只负责摘要输出长度。
|
||||
- **更容易发现摘要失败**: 重要的后台摘要失败现在会强制显示到浏览器控制台 (`F12`)，并同步给出状态提示。
|
||||
|
||||
---
|
||||
|
||||
@@ -22,14 +27,84 @@
|
||||
- ✅ **持久化存储**: 复用 Open WebUI 共享数据库连接，自动支持 PostgreSQL/SQLite 等。
|
||||
- ✅ **灵活保留策略**: 可配置保留对话头部和尾部消息，确保关键信息连贯。
|
||||
- ✅ **智能注入**: 将历史摘要智能注入到新上下文中。
|
||||
- ✅ **外部聊天引用摘要**: 支持复用缓存摘要、小聊天直接注入，以及大聊天先摘要后注入。
|
||||
- ✅ **结构感知裁剪**: 智能折叠过长消息，保留文档骨架（标题、首尾）。
|
||||
- ✅ **原生工具输出裁剪**: 支持裁剪冗长的工具调用输出。
|
||||
- ✅ **实时监控**: 实时监控上下文使用情况，超过 90% 发出警告。
|
||||
- ✅ **详细日志**: 提供精确的 Token 统计日志，便于调试。
|
||||
- ✅ **快速预估 + 精确回退**: 提供更快的多语言 Token 预估，并在必要时回退到精确统计，便于调试。
|
||||
- ✅ **智能模型匹配**: 自定义模型自动继承基础模型的阈值配置。
|
||||
- ⚠ **多模态支持**: 图片内容会被保留，但其 Token **不参与计算**。请相应调整阈值。
|
||||
|
||||
详细的工作原理和流程请参考 [工作流程指南](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/WORKFLOW_GUIDE_CN.md)。
|
||||
详细的工作原理和更长说明仍可参考 [工作流程指南](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/WORKFLOW_GUIDE_CN.md)。
|
||||
|
||||
---
|
||||
|
||||
## 这次解决了什么问题（通俗版）
|
||||
|
||||
- **问题 1：引用别的聊天时，摘要失败可能把当前对话一起弄挂。**
|
||||
以前如果过滤器需要先帮被引用聊天做摘要，而这一步的 LLM 调用失败了，当前请求也可能直接失败。现在改成了“能摘要就摘要，失败就退回直接塞上下文”，当前对话不会被一起拖死。
|
||||
- **问题 2：有些被引用聊天被截得太早，信息丢得太多。**
|
||||
以前有一段逻辑把 `max_summary_tokens` 这种“输出长度限制”误当成了“输入上下文窗口”，结果大一点的引用聊天会被过早截断。现在改成按摘要模型真实的输入窗口来算，能保留更多有用内容。
|
||||
- **问题 3：后台摘要失败时，用户不容易知道发生了什么。**
|
||||
以前在 `show_debug_log=false` 时，有些后台失败只会留在内部日志里。现在关键失败会强制打到浏览器控制台，并在聊天状态里提醒去看 `F12`。
|
||||
|
||||
---
|
||||
|
||||
## 工作流总览
|
||||
|
||||
该过滤器分为两个阶段：
|
||||
|
||||
1. `inlet`：在请求发送给模型前执行，负责注入已有摘要、处理外部聊天引用、并在必要时裁剪上下文。
|
||||
2. `outlet`：在模型回复完成后异步执行，负责判断是否需要生成新摘要，并在合适时写入数据库。
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
A[请求进入 inlet] --> B[规范化工具 ID 并按需裁剪超长工具输出]
|
||||
B --> C{是否附带引用聊天?}
|
||||
C -- 否 --> D[如果有当前聊天摘要就先加载]
|
||||
C -- 是 --> E[逐个检查被引用聊天]
|
||||
|
||||
E --> F{已有缓存摘要?}
|
||||
F -- 是 --> G[直接复用缓存摘要]
|
||||
F -- 否 --> H{能直接放进当前预算?}
|
||||
H -- 是 --> I[直接注入完整引用聊天文本]
|
||||
H -- 否 --> J[准备引用聊天的摘要输入]
|
||||
|
||||
J --> K{引用聊天摘要调用成功?}
|
||||
K -- 是 --> L[注入生成后的引用摘要]
|
||||
K -- 否 --> M[回退为直接注入上下文]
|
||||
|
||||
G --> D
|
||||
I --> D
|
||||
L --> D
|
||||
M --> D
|
||||
|
||||
D --> N[为当前聊天构造 Head + Summary + Tail]
|
||||
N --> O{是否超过 max_context_tokens?}
|
||||
O -- 是 --> P[从最旧 atomic groups 开始裁剪]
|
||||
O -- 否 --> Q[把最终上下文发给模型]
|
||||
P --> Q
|
||||
|
||||
Q --> R[模型返回当前回复]
|
||||
R --> S[Outlet 重建完整历史]
|
||||
S --> T{达到压缩阈值了吗?}
|
||||
T -- 否 --> U[结束]
|
||||
T -- 是 --> V[把摘要输入压到摘要模型可接受的上下文窗口]
|
||||
|
||||
V --> W{后台摘要调用成功?}
|
||||
W -- 是 --> X[保存新摘要并更新状态]
|
||||
W -- 否 --> Y[强制输出浏览器控制台错误并提示用户查看]
|
||||
```
|
||||
|
||||
### 关键说明
|
||||
|
||||
- `inlet` 只负责注入和裁剪上下文，不负责生成当前聊天的主摘要。
|
||||
- `outlet` 异步生成摘要，不会阻塞当前回复。
|
||||
- 外部聊天引用可以来自已有持久化摘要、小聊天的完整文本，或动态生成/截断后的引用摘要。
|
||||
- 如果引用聊天摘要失败，会自动回退为直接注入上下文，而不是让当前请求失败。
|
||||
- `summary_model_max_context` 控制摘要输入窗口；`max_summary_tokens` 只控制生成摘要的输出长度。
|
||||
- 重要的后台摘要失败会显示到浏览器控制台 (`F12`) 和聊天状态提示里。
|
||||
- 外部引用消息在裁剪阶段会被特殊保护，避免被最先删除。
|
||||
|
||||
---
|
||||
|
||||
@@ -65,8 +140,8 @@
|
||||
| 参数 | 默认值 | 描述 |
|
||||
| :-------------------- | :------ | :------------------------------------------------------------------------------------------------------------------------------------------ |
|
||||
| `summary_model` | `None` | 用于生成摘要的模型 ID。**强烈建议**配置快速、经济、上下文窗口大的模型（如 `gemini-2.5-flash`、`deepseek-v3`）。留空则尝试复用当前对话模型。 |
|
||||
| `summary_model_max_context` | `0` | 摘要模型的最大上下文 Token 数。如果为 0，则回退到 `model_thresholds` 或全局 `max_context_tokens`。 |
|
||||
| `max_summary_tokens` | `16384` | 生成摘要时允许的最大 Token 数。 |
|
||||
| `summary_model_max_context` | `0` | 摘要请求可使用的输入上下文窗口。如果为 0，则回退到 `model_thresholds` 或全局 `max_context_tokens`。 |
|
||||
| `max_summary_tokens` | `16384` | 生成摘要时允许的最大输出 Token 数。它不是摘要输入窗口上限。 |
|
||||
| `summary_temperature` | `0.1` | 控制摘要生成的随机性，较低的值结果更稳定。 |
|
||||
|
||||
### 高级配置
|
||||
@@ -94,7 +169,8 @@
|
||||
|
||||
| 参数 | 默认值 | 描述 |
|
||||
| :----------------------------- | :------- | :-------------------------------------------------------------------------------------------------------------------------------------- |
|
||||
| `enable_tool_output_trimming` | `false` | 启用时，若 `function_calling: "native"` 激活，将裁剪冗长的工具输出以仅提取最终答案。 |
|
||||
| `enable_tool_output_trimming` | `true` | 启用后（仅在 `function_calling: "native"` 下生效）会裁剪过大的本机工具输出，保留工具调用链结构并以简短占位替换冗长内容。 |
|
||||
| `tool_trim_threshold_chars` | `600` | 当本机工具输出累计字符数达到该值时触发裁剪，适用于包含长文本或表格的工具结果。 |
|
||||
| `debug_mode` | `false` | 是否在 Open WebUI 的控制台日志中打印详细的调试信息。生产环境默认且建议设为 `false`。 |
|
||||
| `show_debug_log` | `false` | 是否在浏览器控制台 (F12) 打印调试日志。便于前端调试。 |
|
||||
| `show_token_usage_status` | `true` | 是否在对话结束时显示 Token 使用情况的状态通知。 |
|
||||
@@ -110,8 +186,12 @@
|
||||
|
||||
- **初始系统提示丢失**：将 `keep_first` 设置为大于 0。
|
||||
- **压缩效果不明显**：提高 `compression_threshold_tokens`，或降低 `keep_first` / `keep_last` 以增强压缩力度。
|
||||
- **引用聊天摘要失败**：当前请求现在应该会继续执行，并回退为直接注入上下文。如果要看上游失败原因，请打开浏览器控制台 (`F12`)。
|
||||
- **后台摘要看起来“没反应”**：重要失败现在会同时出现在状态提示和浏览器控制台 (`F12`) 中。
|
||||
- **提交 Issue**: 如果遇到任何问题，请在 GitHub 上提交 Issue：[OpenWebUI Extensions Issues](https://github.com/Fu-Jie/openwebui-extensions/issues)
|
||||
|
||||
## 更新日志
|
||||
|
||||
请查看 [`v1.5.0` 版本发布说明](./v1.5.0_CN.md) 获取本次版本的独立发布摘要。
|
||||
|
||||
完整历史请查看 GitHub 项目： [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions)
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
270
plugins/filters/async-context-compression/community_post.md
Normal file
270
plugins/filters/async-context-compression/community_post.md
Normal file
@@ -0,0 +1,270 @@
|
||||
[](https://openwebui.com/posts/async_context_compression_b1655bc8)
|
||||
|
||||
# Async Context Compression: A Production-Scale Working-Memory Filter for OpenWebUI
|
||||
|
||||
Long chats do not just get expensive. They also get fragile.
|
||||
|
||||
Once a conversation grows large enough, you usually have to choose between two bad options:
|
||||
|
||||
- keep the full history and pay a heavy context cost
|
||||
- trim aggressively and risk losing continuity, tool state, or important prior decisions
|
||||
|
||||
`Async Context Compression` is built to avoid that tradeoff.
|
||||
|
||||
It is not a simple “summarize old messages” utility. It is a structure-aware, async, database-backed working-memory system for OpenWebUI that can compress long conversations while preserving conversational continuity, tool-calling integrity, and now, as of `v1.5.0`, referenced-chat context injection as well.
|
||||
|
||||
This plugin has now reached the point where it feels complete enough to be described as a serious, high-capability filter rather than a small convenience add-on.
|
||||
|
||||
**[📖 Full README](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/README.md)**
|
||||
**[📝 v1.5.0 Release Notes](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/v1.5.0.md)**
|
||||
|
||||
---
|
||||
|
||||
## Why This Plugin Exists
|
||||
|
||||
OpenWebUI conversations often contain much more than plain chat:
|
||||
|
||||
- long-running planning threads
|
||||
- coding sessions with repeated tool use
|
||||
- model-specific context limits
|
||||
- multimodal messages
|
||||
- external referenced chats
|
||||
- custom models with different context windows
|
||||
|
||||
A naive compression strategy is not enough in those environments.
|
||||
|
||||
If a filter only drops earlier messages based on length, it can:
|
||||
|
||||
- break native tool-calling chains
|
||||
- lose critical task state
|
||||
- destroy continuity in old chats
|
||||
- make debugging impossible
|
||||
- hide important provider-side failures
|
||||
|
||||
`Async Context Compression` is designed around a stronger premise:
|
||||
|
||||
> compress history without treating conversation structure as disposable
|
||||
|
||||
That means it tries to preserve what actually matters for the next turn:
|
||||
|
||||
- the current goal
|
||||
- durable user preferences
|
||||
- recent progress
|
||||
- tool outputs that still matter
|
||||
- error state
|
||||
- summary continuity
|
||||
- referenced context from other chats
|
||||
|
||||
---
|
||||
|
||||
## What Makes It Different
|
||||
|
||||
This plugin now combines several capabilities that are usually split across separate systems:
|
||||
|
||||
### 1. Asynchronous working-memory generation
|
||||
|
||||
The current reply is not blocked while the plugin generates a new summary in the background.
|
||||
|
||||
### 2. Persistent summary storage
|
||||
|
||||
Summaries are stored in OpenWebUI's shared database and reused across turns, instead of being regenerated from scratch every time.
|
||||
|
||||
### 3. Structure-aware trimming
|
||||
|
||||
The filter respects atomic message boundaries so native tool-calling history is not corrupted by compression.
|
||||
|
||||
### 4. External chat reference summarization
|
||||
|
||||
New in `v1.5.0`: referenced chats can now be reused as cached summaries, injected directly if small enough, or summarized before injection if too large.
|
||||
|
||||
### 5. Mixed-script token estimation
|
||||
|
||||
The plugin now uses a much stronger multilingual token estimation path before falling back to exact counting, which helps reduce unnecessary expensive token calculations while staying much closer to real usage.
|
||||
|
||||
### 6. Real failure visibility
|
||||
|
||||
Important background summary failures are surfaced to the browser console and status messages instead of disappearing silently.
|
||||
|
||||
---
|
||||
|
||||
## Workflow Overview
|
||||
|
||||
This is the current high-level flow:
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
A[Request enters inlet] --> B[Normalize tool IDs and optionally trim large tool outputs]
|
||||
B --> C{Referenced chats attached?}
|
||||
C -- No --> D[Load current chat summary if available]
|
||||
C -- Yes --> E[Inspect each referenced chat]
|
||||
|
||||
E --> F{Existing cached summary?}
|
||||
F -- Yes --> G[Reuse cached summary]
|
||||
F -- No --> H{Fits direct budget?}
|
||||
H -- Yes --> I[Inject full referenced chat text]
|
||||
H -- No --> J[Prepare referenced-chat summary input]
|
||||
|
||||
J --> K{Referenced-chat summary call succeeds?}
|
||||
K -- Yes --> L[Inject generated referenced summary]
|
||||
K -- No --> M[Fallback to direct contextual injection]
|
||||
|
||||
G --> D
|
||||
I --> D
|
||||
L --> D
|
||||
M --> D
|
||||
|
||||
D --> N[Build current-chat Head + Summary + Tail]
|
||||
N --> O{Over max_context_tokens?}
|
||||
O -- Yes --> P[Trim oldest atomic groups]
|
||||
O -- No --> Q[Send final context to the model]
|
||||
P --> Q
|
||||
|
||||
Q --> R[Model returns the reply]
|
||||
R --> S[Outlet rebuilds the full history]
|
||||
S --> T{Reached compression threshold?}
|
||||
T -- No --> U[Finish]
|
||||
T -- Yes --> V[Fit summary input to the summary model context]
|
||||
|
||||
V --> W{Background summary call succeeds?}
|
||||
W -- Yes --> X[Save new chat summary and update status]
|
||||
W -- No --> Y[Force browser-console error and show status hint]
|
||||
```
|
||||
|
||||
This is why I consider the plugin “powerful” now: it is no longer solving a single problem. It is coordinating context reduction, summary persistence, tool safety, referenced-chat handling, and model-budget control inside one filter.
|
||||
|
||||
---
|
||||
|
||||
## New in v1.5.0
|
||||
|
||||
This release is important because it turns the plugin from “long-chat compression with strong tool safety” into something closer to a reusable context-management layer.
|
||||
|
||||
### External chat reference summaries
|
||||
|
||||
This is a new feature in `v1.5.0`, not just a small adjustment.
|
||||
|
||||
When a user references another chat:
|
||||
|
||||
- the plugin can reuse an existing cached summary
|
||||
- inject the full referenced chat if it is small enough
|
||||
- or generate a summary first if the referenced chat is too large
|
||||
|
||||
That means the filter can now carry relevant context across chats, not just across turns inside the same chat.
|
||||
|
||||
### Fast multilingual token estimation
|
||||
|
||||
Also new in `v1.5.0`.
|
||||
|
||||
The plugin no longer relies on a rough one-size-fits-all character ratio. It now estimates token usage with mixed-script heuristics that behave much better for:
|
||||
|
||||
- English
|
||||
- Chinese
|
||||
- Japanese
|
||||
- Korean
|
||||
- Cyrillic
|
||||
- Arabic
|
||||
- Thai
|
||||
- mixed-language conversations
|
||||
|
||||
This matters because the plugin makes context decisions constantly. Better estimation means fewer unnecessary exact counts and fewer bad preflight assumptions.
|
||||
|
||||
### Stronger final-prompt budgeting
|
||||
|
||||
The summary path now fits the **real final summary request**, not just an intermediate estimate. That includes:
|
||||
|
||||
- prompt wrapper
|
||||
- formatted conversation text
|
||||
- previous summary
|
||||
- reserved output budget
|
||||
- safety margin
|
||||
|
||||
This directly improves reliability in the large old-chat cases that are hardest to handle.
|
||||
|
||||
---
|
||||
|
||||
## Why It Feels Complete Now
|
||||
|
||||
I would describe the current plugin as “feature-complete for the main problem space,” because it now covers the major operational surfaces that matter in real usage:
|
||||
|
||||
- long plain-chat conversations
|
||||
- multi-step coding threads
|
||||
- native tool-calling conversations
|
||||
- persistent summaries
|
||||
- custom model thresholds
|
||||
- background async generation
|
||||
- external chat references
|
||||
- multilingual token estimation
|
||||
- failure surfacing for debugging
|
||||
|
||||
That does not mean it is finished forever. It means the plugin has crossed the line from a narrow experimental filter into a robust context-management system with enough breadth to support demanding OpenWebUI usage patterns.
|
||||
|
||||
---
|
||||
|
||||
## Scale and Engineering Depth
|
||||
|
||||
For people who care about implementation depth, this plugin is not small anymore.
|
||||
|
||||
Current code size:
|
||||
|
||||
- main plugin: **4,573 lines**
|
||||
- focused test file: **1,037 lines**
|
||||
- combined visible implementation + regression coverage: **5,610 lines**
|
||||
|
||||
Line count is not a quality metric by itself, but at this scale it does say something real:
|
||||
|
||||
- the plugin has grown well beyond a toy filter
|
||||
- the behavior surface is large enough to require explicit regression testing
|
||||
- the plugin now encodes a lot of edge-case handling that only shows up after repeated real-world usage
|
||||
|
||||
In other words: this is no longer “just summarize old messages.” It is a fairly serious stateful filter.
|
||||
|
||||
---
|
||||
|
||||
## Practical Benefits
|
||||
|
||||
If you use OpenWebUI heavily, the value is straightforward:
|
||||
|
||||
- lower token consumption in long chats
|
||||
- better continuity across long-running sessions
|
||||
- safer native tool-calling history
|
||||
- fewer broken conversations after compression
|
||||
- more stable summary generation on large histories
|
||||
- better visibility when the provider rejects a summary request
|
||||
- useful reuse of context from referenced chats
|
||||
|
||||
This plugin is especially valuable if you:
|
||||
|
||||
- regularly work in long coding chats
|
||||
- use models with strict context budgets
|
||||
- rely on native tool calling
|
||||
- revisit old project chats
|
||||
- want summaries to behave like working memory, not like lossy notes
|
||||
|
||||
---
|
||||
|
||||
## Installation
|
||||
|
||||
- OpenWebUI Community: <https://openwebui.com/posts/async_context_compression_b1655bc8>
|
||||
- Source: <https://github.com/Fu-Jie/openwebui-extensions/tree/main/plugins/filters/async-context-compression>
|
||||
|
||||
If you want the full valve list, deployment notes, and troubleshooting details, the README is the best reference.
|
||||
|
||||
---
|
||||
|
||||
## Final Note
|
||||
|
||||
Do I think this plugin is powerful?
|
||||
|
||||
Yes, genuinely.
|
||||
|
||||
Not because it is large, but because it now solves the right combination of problems at once:
|
||||
|
||||
- cost control
|
||||
- continuity
|
||||
- structural safety
|
||||
- async persistence
|
||||
- cross-chat reuse
|
||||
- operational debuggability
|
||||
|
||||
That combination is what makes it feel strong.
|
||||
|
||||
If you have been looking for a serious long-conversation memory/compression filter for OpenWebUI, `Async Context Compression` is now in that category.
|
||||
282
plugins/filters/async-context-compression/community_post_CN.md
Normal file
282
plugins/filters/async-context-compression/community_post_CN.md
Normal file
@@ -0,0 +1,282 @@
|
||||
[](https://openwebui.com/posts/async_context_compression_b1655bc8)
|
||||
|
||||
# Async Context Compression：一个面向生产场景的 OpenWebUI 工作记忆过滤器
|
||||
|
||||
长对话的问题，从来不只是“贵”。
|
||||
|
||||
当聊天足够长时，通常只剩下两个都不太好的选择：
|
||||
|
||||
- 保留完整历史，继续承担很高的上下文成本
|
||||
- 粗暴裁剪旧消息，但冒着丢失上下文、工具状态和关键决策的风险
|
||||
|
||||
`Async Context Compression` 的目标，就是尽量避免这个二选一。
|
||||
|
||||
它不是一个简单的“把老消息总结一下”的小工具，而是一个带有结构感知、异步摘要、数据库持久化能力的 OpenWebUI 工作记忆系统。它的任务不是单纯缩短上下文，而是在压缩长对话的同时，尽量保留：
|
||||
|
||||
- 对话连续性
|
||||
- 工具调用状态完整性
|
||||
- 历史摘要进度
|
||||
- 跨聊天引用上下文
|
||||
- 出错时的可诊断性
|
||||
|
||||
到 `v1.5.0` 这个阶段，我认为它已经不再只是一个“方便的小过滤器”，而是一个足够完整、足够强、也足够有工程深度的上下文管理插件。
|
||||
|
||||
**[📖 完整 README](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/README_CN.md)**
|
||||
**[📝 v1.5.0 发布说明](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/v1.5.0_CN.md)**
|
||||
|
||||
---
|
||||
|
||||
## 为什么会有这个插件
|
||||
|
||||
OpenWebUI 里的真实对话，通常并不只是“用户问一句，模型答一句”。
|
||||
|
||||
它常常还包含：
|
||||
|
||||
- 很长的项目型对话
|
||||
- 多轮编码与调试
|
||||
- 原生工具调用
|
||||
- 多模态消息
|
||||
- 不同模型上下文窗口差异
|
||||
- 其他聊天的引用上下文
|
||||
|
||||
在这种环境里，单纯靠“按长度裁掉旧消息”其实不够。
|
||||
|
||||
如果一个过滤器只会按长度或索引裁剪消息，它很容易：
|
||||
|
||||
- 把原生 tool-calling 历史裁坏
|
||||
- 丢掉仍然会影响下一轮回复的关键信息
|
||||
- 在老聊天里破坏连续性
|
||||
- 出问题时几乎无法排查
|
||||
- 把上游 provider 报错伪装成模糊的内部错误
|
||||
|
||||
`Async Context Compression` 的核心思路更强一些：
|
||||
|
||||
> 可以压缩历史，但不能把“对话结构”当成无关紧要的东西一起压掉
|
||||
|
||||
它真正想保留的是下一轮最需要的状态：
|
||||
|
||||
- 当前目标
|
||||
- 持久偏好
|
||||
- 最近进展
|
||||
- 仍然有效的工具结果
|
||||
- 错误状态
|
||||
- 已有摘要的连续性
|
||||
- 来自其他聊天的相关上下文
|
||||
|
||||
---
|
||||
|
||||
## 它和普通摘要插件有什么不同
|
||||
|
||||
现在这个插件，实际上已经把几个通常要分散在不同系统里的能力组合到了一起：
|
||||
|
||||
### 1. 异步工作记忆生成
|
||||
|
||||
用户当前这次回复不会被后台摘要阻塞。
|
||||
|
||||
### 2. 持久化摘要存储
|
||||
|
||||
摘要会写入 OpenWebUI 共享数据库，并在后续轮次中复用，而不是每次都从头重算。
|
||||
|
||||
### 3. 结构感知裁剪
|
||||
|
||||
裁剪逻辑会尊重原子消息边界，避免把原生 tool-calling 历史裁坏。
|
||||
|
||||
### 4. 外部聊天引用摘要
|
||||
|
||||
这是 `v1.5.0` 新增的重要能力：被引用聊天现在可以直接复用缓存摘要、在小体量时直接注入、或者在过大时先生成摘要再注入。
|
||||
|
||||
### 5. 多语言 Token 预估
|
||||
|
||||
插件现在具备更强的多脚本文本 Token 预估逻辑，在很多情况下可以减少不必要的精确计数，同时明显比旧的粗略字符比值更贴近真实用量。
|
||||
|
||||
### 6. 失败可见性
|
||||
|
||||
关键的后台摘要失败现在会出现在浏览器控制台和状态提示里，不再悄悄消失。
|
||||
|
||||
---
|
||||
|
||||
## 工作流总览
|
||||
|
||||
下面是当前的高层流程：
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
A[Request enters inlet] --> B[Normalize tool IDs and optionally trim large tool outputs]
|
||||
B --> C{Referenced chats attached?}
|
||||
C -- No --> D[Load current chat summary if available]
|
||||
C -- Yes --> E[Inspect each referenced chat]
|
||||
|
||||
E --> F{Existing cached summary?}
|
||||
F -- Yes --> G[Reuse cached summary]
|
||||
F -- No --> H{Fits direct budget?}
|
||||
H -- Yes --> I[Inject full referenced chat text]
|
||||
H -- No --> J[Prepare referenced-chat summary input]
|
||||
|
||||
J --> K{Referenced-chat summary call succeeds?}
|
||||
K -- Yes --> L[Inject generated referenced summary]
|
||||
K -- No --> M[Fallback to direct contextual injection]
|
||||
|
||||
G --> D
|
||||
I --> D
|
||||
L --> D
|
||||
M --> D
|
||||
|
||||
D --> N[Build current-chat Head + Summary + Tail]
|
||||
N --> O{Over max_context_tokens?}
|
||||
O -- Yes --> P[Trim oldest atomic groups]
|
||||
O -- No --> Q[Send final context to the model]
|
||||
P --> Q
|
||||
|
||||
Q --> R[Model returns the reply]
|
||||
R --> S[Outlet rebuilds the full history]
|
||||
S --> T{Reached compression threshold?}
|
||||
T -- No --> U[Finish]
|
||||
T -- Yes --> V[Fit summary input to the summary model context]
|
||||
|
||||
V --> W{Background summary call succeeds?}
|
||||
W -- Yes --> X[Save new chat summary and update status]
|
||||
W -- No --> Y[Force browser-console error and show status hint]
|
||||
```
|
||||
|
||||
这也是为什么我会觉得它现在“强”：它已经不再只解决一个问题，而是在一个过滤器里同时协调：
|
||||
|
||||
- 上下文压缩
|
||||
- 历史摘要复用
|
||||
- 工具调用安全性
|
||||
- 被引用聊天上下文
|
||||
- 模型预算控制
|
||||
|
||||
---
|
||||
|
||||
## v1.5.0 为什么重要
|
||||
|
||||
这个版本的重要性在于，它把插件从“长对话压缩器”推进成了一个更接近“上下文管理层”的东西。
|
||||
|
||||
### 外部聊天引用摘要
|
||||
|
||||
这是 `v1.5.0` 的新功能，不是小修小补。
|
||||
|
||||
当用户引用另一个聊天时，插件现在可以：
|
||||
|
||||
- 直接复用已有缓存摘要
|
||||
- 如果聊天足够小，直接把完整内容注入
|
||||
- 如果聊天太大，先生成摘要再注入
|
||||
|
||||
这意味着它现在不仅能跨“轮次”保留上下文，也能开始跨“聊天”携带相关上下文。
|
||||
|
||||
### 快速多语言 Token 预估
|
||||
|
||||
这同样是 `v1.5.0` 的新能力。
|
||||
|
||||
插件不再依赖简单粗暴的统一字符比值，而是改用更适合混合语言文本的估算方式，尤其对下面这些场景更有意义：
|
||||
|
||||
- 英文
|
||||
- 中文
|
||||
- 日文
|
||||
- 韩文
|
||||
- 西里尔字符
|
||||
- 阿拉伯语
|
||||
- 泰语
|
||||
- 中英混合或多语言混合对话
|
||||
|
||||
这很重要，因为上下文管理类插件会不断做预算判断。预估更准，就意味着更少无意义的精确计算，也更不容易在预检阶段做出错误判断。
|
||||
|
||||
### 更强的最终请求预算控制
|
||||
|
||||
现在的摘要路径会去拟合“真实最终 summary request”，而不是只看一个中间估算值。它会把这些内容都算进去：
|
||||
|
||||
- prompt 包装
|
||||
- 格式化后的对话文本
|
||||
- previous summary
|
||||
- 预留输出预算
|
||||
- 安全余量
|
||||
|
||||
这对老聊天、大聊天和最难处理的边界情况特别关键。
|
||||
|
||||
---
|
||||
|
||||
## 为什么我觉得它现在已经足够完整
|
||||
|
||||
如果把“问题空间”列出来，我会说这个插件现在对主要场景已经覆盖得比较完整了：
|
||||
|
||||
- 很长的普通聊天
|
||||
- 多轮编码与调试对话
|
||||
- 原生工具调用
|
||||
- 历史摘要持久化
|
||||
- 自定义模型阈值
|
||||
- 异步后台摘要
|
||||
- 外部聊天引用
|
||||
- 多语言 Token 预估
|
||||
- 调试可见性
|
||||
|
||||
这并不代表它永远不会再迭代，而是说它已经越过了“窄功能实验品”的阶段，进入了一个更像“通用上下文管理系统”的形态。
|
||||
|
||||
---
|
||||
|
||||
## 代码规模与工程深度
|
||||
|
||||
如果你关心实现深度，这个插件现在已经不小了。
|
||||
|
||||
当前代码规模：
|
||||
|
||||
- 主插件文件：**4,573 行**
|
||||
- 聚焦测试文件：**1,037 行**
|
||||
- 可见实现 + 回归测试合计：**5,610 行**
|
||||
|
||||
代码行数本身不等于质量，但在这个量级上，它至少说明了几件真实的事：
|
||||
|
||||
- 这已经不是一个玩具级过滤器
|
||||
- 这个插件的行为面足够大，必须靠专门回归测试兜住
|
||||
- 它已经积累了很多只有在真实使用中才会暴露出来的边界处理逻辑
|
||||
|
||||
也就是说，它现在做的事情，已经明显不是“把老消息总结一下”那么简单。
|
||||
|
||||
---
|
||||
|
||||
## 实际价值
|
||||
|
||||
如果你是 OpenWebUI 的重度用户，这个插件的价值其实很直接：
|
||||
|
||||
- 长聊天更省 Token
|
||||
- 长会话连续性更好
|
||||
- 原生 tool-calling 更安全
|
||||
- 压缩后更不容易把会话搞坏
|
||||
- 大历史摘要生成更稳定
|
||||
- provider 拒绝摘要请求时更容易看到真错误
|
||||
- 能复用其他聊天里的有效上下文
|
||||
|
||||
尤其适合这些用户：
|
||||
|
||||
- 经常做长时间编码聊天
|
||||
- 使用上下文窗口比较紧的模型
|
||||
- 依赖原生工具调用
|
||||
- 经常回看旧项目聊天
|
||||
- 希望摘要更像“工作记忆”而不是“丢失细节的简要笔记”
|
||||
|
||||
---
|
||||
|
||||
## 安装
|
||||
|
||||
- OpenWebUI 社区：<https://openwebui.com/posts/async_context_compression_b1655bc8>
|
||||
- 源码目录：<https://github.com/Fu-Jie/openwebui-extensions/tree/main/plugins/filters/async-context-compression>
|
||||
|
||||
如果你想看完整的 valves、部署说明和故障排查，README 仍然是最完整的参考入口。
|
||||
|
||||
---
|
||||
|
||||
## 最后一句
|
||||
|
||||
你问我这个插件是不是强大。
|
||||
|
||||
我的答案是：**是，确实强，而且现在已经不是“看起来强”，而是“问题空间覆盖得比较完整”的那种强。**
|
||||
|
||||
不是因为它代码多，而是因为它现在同时解决的是一组真正相关的问题：
|
||||
|
||||
- 成本控制
|
||||
- 连续性
|
||||
- 结构安全
|
||||
- 异步持久化
|
||||
- 跨聊天上下文复用
|
||||
- 出错时的可诊断性
|
||||
|
||||
正是这几个东西一起成立，才让它现在像一个真正成熟的长对话上下文管理插件。
|
||||
@@ -18,6 +18,63 @@ def _ensure_module(name: str) -> types.ModuleType:
|
||||
return module
|
||||
|
||||
|
||||
def _install_dependency_stubs() -> None:
|
||||
pydantic_module = _ensure_module("pydantic")
|
||||
sqlalchemy_module = _ensure_module("sqlalchemy")
|
||||
sqlalchemy_orm_module = _ensure_module("sqlalchemy.orm")
|
||||
sqlalchemy_engine_module = _ensure_module("sqlalchemy.engine")
|
||||
|
||||
class DummyBaseModel:
|
||||
def __init__(self, **kwargs):
|
||||
annotations = getattr(self.__class__, "__annotations__", {})
|
||||
for field_name in annotations:
|
||||
if field_name in kwargs:
|
||||
value = kwargs[field_name]
|
||||
else:
|
||||
value = getattr(self.__class__, field_name, None)
|
||||
setattr(self, field_name, value)
|
||||
|
||||
def dummy_field(default=None, **kwargs):
|
||||
return default
|
||||
|
||||
class DummyMetadata:
|
||||
def create_all(self, *args, **kwargs):
|
||||
return None
|
||||
|
||||
def dummy_declarative_base():
|
||||
class DummyBase:
|
||||
metadata = DummyMetadata()
|
||||
|
||||
return DummyBase
|
||||
|
||||
def dummy_sessionmaker(*args, **kwargs):
|
||||
return lambda: None
|
||||
|
||||
class DummyEngine:
|
||||
pass
|
||||
|
||||
def dummy_column(*args, **kwargs):
|
||||
return None
|
||||
|
||||
def dummy_type(*args, **kwargs):
|
||||
return None
|
||||
|
||||
def dummy_inspect(*args, **kwargs):
|
||||
return types.SimpleNamespace(has_table=lambda *a, **k: False)
|
||||
|
||||
pydantic_module.BaseModel = DummyBaseModel
|
||||
pydantic_module.Field = dummy_field
|
||||
sqlalchemy_module.Column = dummy_column
|
||||
sqlalchemy_module.String = dummy_type
|
||||
sqlalchemy_module.Text = dummy_type
|
||||
sqlalchemy_module.DateTime = dummy_type
|
||||
sqlalchemy_module.Integer = dummy_type
|
||||
sqlalchemy_module.inspect = dummy_inspect
|
||||
sqlalchemy_orm_module.declarative_base = dummy_declarative_base
|
||||
sqlalchemy_orm_module.sessionmaker = dummy_sessionmaker
|
||||
sqlalchemy_engine_module.Engine = DummyEngine
|
||||
|
||||
|
||||
def _install_openwebui_stubs() -> None:
|
||||
_ensure_module("open_webui")
|
||||
_ensure_module("open_webui.utils")
|
||||
@@ -47,7 +104,8 @@ def _install_openwebui_stubs() -> None:
|
||||
return None
|
||||
|
||||
class DummyRequest:
|
||||
pass
|
||||
def __init__(self, *args, **kwargs):
|
||||
pass
|
||||
|
||||
chat_module.generate_chat_completion = generate_chat_completion
|
||||
users_module.Users = DummyUsers
|
||||
@@ -57,6 +115,7 @@ def _install_openwebui_stubs() -> None:
|
||||
fastapi_requests.Request = DummyRequest
|
||||
|
||||
|
||||
_install_dependency_stubs()
|
||||
_install_openwebui_stubs()
|
||||
spec = importlib.util.spec_from_file_location(MODULE_NAME, PLUGIN_PATH)
|
||||
module = importlib.util.module_from_spec(spec)
|
||||
@@ -189,9 +248,12 @@ class TestAsyncContextCompression(unittest.TestCase):
|
||||
{"role": "assistant", "content": "Final answer"},
|
||||
]
|
||||
|
||||
trimmed_count = self.filter._trim_native_tool_outputs(messages, "en-US")
|
||||
trimmed_count, trim_debug = self.filter._trim_native_tool_outputs(
|
||||
messages, "en-US"
|
||||
)
|
||||
|
||||
self.assertEqual(trimmed_count, 1)
|
||||
self.assertIsNone(trim_debug)
|
||||
self.assertEqual(messages[1]["content"], "... [Content collapsed] ...")
|
||||
self.assertTrue(messages[1]["metadata"]["is_trimmed"])
|
||||
self.assertTrue(messages[2]["metadata"]["tool_outputs_trimmed"])
|
||||
@@ -213,9 +275,12 @@ class TestAsyncContextCompression(unittest.TestCase):
|
||||
}
|
||||
]
|
||||
|
||||
trimmed_count = self.filter._trim_native_tool_outputs(messages, "en-US")
|
||||
trimmed_count, trim_debug = self.filter._trim_native_tool_outputs(
|
||||
messages, "en-US"
|
||||
)
|
||||
|
||||
self.assertEqual(trimmed_count, 1)
|
||||
self.assertIsNone(trim_debug)
|
||||
self.assertIn(
|
||||
'result=""... [Content collapsed] ...""',
|
||||
messages[0]["content"],
|
||||
@@ -258,9 +323,12 @@ class TestAsyncContextCompression(unittest.TestCase):
|
||||
{"role": "tool", "content": "x" * 1600},
|
||||
]
|
||||
|
||||
trimmed_count = self.filter._trim_native_tool_outputs(messages, "en-US")
|
||||
trimmed_count, trim_debug = self.filter._trim_native_tool_outputs(
|
||||
messages, "en-US"
|
||||
)
|
||||
|
||||
self.assertEqual(trimmed_count, 1)
|
||||
self.assertIsNone(trim_debug)
|
||||
self.assertEqual(messages[1]["content"], "... [Content collapsed] ...")
|
||||
self.assertTrue(messages[1]["metadata"]["is_trimmed"])
|
||||
|
||||
@@ -391,11 +459,55 @@ class TestAsyncContextCompression(unittest.TestCase):
|
||||
|
||||
self.assertTrue(create_task_called)
|
||||
|
||||
def test_summary_save_progress_matches_truncated_input(self):
|
||||
def test_estimate_messages_tokens_counts_output_text_parts(self):
|
||||
messages = [
|
||||
{
|
||||
"role": "assistant",
|
||||
"content": [{"type": "output_text", "text": "abcd" * 25}],
|
||||
}
|
||||
]
|
||||
|
||||
self.assertEqual(
|
||||
self.filter._estimate_messages_tokens(messages),
|
||||
module._estimate_text_tokens("abcd" * 25),
|
||||
)
|
||||
|
||||
def test_unfold_messages_keeps_plain_assistant_output_when_expand_is_not_richer(self):
|
||||
misc_module = _ensure_module("open_webui.utils.misc")
|
||||
misc_module.convert_output_to_messages = lambda output, raw=True: [
|
||||
{
|
||||
"role": "assistant",
|
||||
"content": [{"type": "output_text", "text": "Plain reply"}],
|
||||
}
|
||||
]
|
||||
|
||||
messages = [
|
||||
{
|
||||
"id": "assistant-1",
|
||||
"role": "assistant",
|
||||
"content": "Plain reply",
|
||||
"output": [
|
||||
{
|
||||
"type": "message",
|
||||
"role": "assistant",
|
||||
"content": [{"type": "output_text", "text": "Plain reply"}],
|
||||
}
|
||||
],
|
||||
}
|
||||
]
|
||||
|
||||
unfolded = self.filter._unfold_messages(messages)
|
||||
|
||||
self.assertEqual(len(unfolded), 1)
|
||||
self.assertEqual(unfolded[0]["id"], "assistant-1")
|
||||
self.assertEqual(unfolded[0]["content"], "Plain reply")
|
||||
self.assertNotIn("output", unfolded[0])
|
||||
|
||||
def test_summary_save_progress_matches_final_prompt_shrink(self):
|
||||
self.filter.valves.keep_first = 1
|
||||
self.filter.valves.keep_last = 1
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
self.filter.valves.summary_model_max_context = 0
|
||||
self.filter.valves.summary_model_max_context = 1200
|
||||
|
||||
captured = {}
|
||||
events = []
|
||||
@@ -404,12 +516,14 @@ class TestAsyncContextCompression(unittest.TestCase):
|
||||
events.append(event)
|
||||
|
||||
async def mock_summary_llm(
|
||||
previous_summary,
|
||||
new_conversation_text,
|
||||
body,
|
||||
user_data,
|
||||
__event_call__,
|
||||
__event_call__=None,
|
||||
__request__=None,
|
||||
previous_summary=None,
|
||||
):
|
||||
captured["conversation_text"] = new_conversation_text
|
||||
return "new summary"
|
||||
|
||||
def mock_save_summary(chat_id, summary, compressed_count):
|
||||
@@ -424,17 +538,22 @@ class TestAsyncContextCompression(unittest.TestCase):
|
||||
self.filter._call_summary_llm = mock_summary_llm
|
||||
self.filter._save_summary = mock_save_summary
|
||||
self.filter._get_model_thresholds = lambda model_id: {
|
||||
"max_context_tokens": 3500
|
||||
"max_context_tokens": 1200
|
||||
}
|
||||
self.filter._calculate_messages_tokens = lambda messages: len(messages) * 1000
|
||||
self.filter._count_tokens = lambda text: 1000
|
||||
self.filter._format_messages_for_summary = lambda messages: "\n".join(
|
||||
msg["content"] for msg in messages
|
||||
)
|
||||
self.filter._build_summary_prompt = (
|
||||
lambda conversation_text, previous_summary=None: conversation_text
|
||||
)
|
||||
self.filter._count_tokens = lambda text: len(text)
|
||||
|
||||
messages = [
|
||||
{"role": "system", "content": "System prompt"},
|
||||
{"role": "user", "content": "Question 1"},
|
||||
{"role": "assistant", "content": "Answer 1"},
|
||||
{"role": "user", "content": "Question 2"},
|
||||
{"role": "assistant", "content": "Answer 2"},
|
||||
{"role": "user", "content": "Q" * 100},
|
||||
{"role": "assistant", "content": "A" * 100},
|
||||
{"role": "user", "content": "B" * 100},
|
||||
{"role": "assistant", "content": "C" * 100},
|
||||
{"role": "user", "content": "Question 3"},
|
||||
]
|
||||
|
||||
@@ -453,9 +572,466 @@ class TestAsyncContextCompression(unittest.TestCase):
|
||||
|
||||
self.assertEqual(captured["chat_id"], "chat-1")
|
||||
self.assertEqual(captured["summary"], "new summary")
|
||||
self.assertEqual(captured["compressed_count"], 2)
|
||||
self.assertEqual(captured["compressed_count"], 3)
|
||||
self.assertEqual(captured["conversation_text"], f"{'Q' * 100}\n{'A' * 100}")
|
||||
self.assertTrue(any(event["type"] == "status" for event in events))
|
||||
|
||||
def test_generate_summary_async_drops_previous_summary_when_prompt_still_oversized(self):
|
||||
self.filter.valves.keep_first = 1
|
||||
self.filter.valves.keep_last = 1
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
self.filter.valves.summary_model_max_context = 1200
|
||||
|
||||
captured = {}
|
||||
|
||||
async def mock_summary_llm(
|
||||
new_conversation_text,
|
||||
body,
|
||||
user_data,
|
||||
__event_call__=None,
|
||||
__request__=None,
|
||||
previous_summary=None,
|
||||
):
|
||||
captured["conversation_text"] = new_conversation_text
|
||||
captured["previous_summary"] = previous_summary
|
||||
return "new summary"
|
||||
|
||||
async def noop_log(*args, **kwargs):
|
||||
return None
|
||||
|
||||
self.filter._log = noop_log
|
||||
self.filter._call_summary_llm = mock_summary_llm
|
||||
self.filter._save_summary = lambda *args: None
|
||||
self.filter._get_model_thresholds = lambda model_id: {
|
||||
"max_context_tokens": 1200
|
||||
}
|
||||
self.filter._format_messages_for_summary = lambda messages: "\n".join(
|
||||
msg["content"] for msg in messages
|
||||
)
|
||||
self.filter._build_summary_prompt = (
|
||||
lambda conversation_text, previous_summary=None: (
|
||||
(previous_summary or "") + "\n" + conversation_text
|
||||
)
|
||||
)
|
||||
self.filter._count_tokens = lambda text: len(text)
|
||||
self.filter._load_summary = lambda chat_id, body: "P" * 220
|
||||
|
||||
messages = [
|
||||
{"role": "system", "content": "System prompt"},
|
||||
{"role": "user", "content": "Q" * 60},
|
||||
{"role": "assistant", "content": "Answer 1"},
|
||||
{"role": "user", "content": "Question 2"},
|
||||
]
|
||||
|
||||
asyncio.run(
|
||||
self.filter._generate_summary_async(
|
||||
messages=messages,
|
||||
chat_id="chat-1",
|
||||
body={"model": "fake-summary-model"},
|
||||
user_data={"id": "user-1"},
|
||||
target_compressed_count=2,
|
||||
lang="en-US",
|
||||
__event_emitter__=None,
|
||||
__event_call__=None,
|
||||
)
|
||||
)
|
||||
|
||||
self.assertEqual(captured["conversation_text"], "Q" * 60)
|
||||
self.assertIsNone(captured["previous_summary"])
|
||||
|
||||
def test_call_summary_llm_surfaces_provider_error_dict(self):
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
self.filter.valves.show_debug_log = False
|
||||
|
||||
async def fake_generate_chat_completion(request, payload, user):
|
||||
return {"error": {"message": "context too long", "code": 400}}
|
||||
|
||||
async def noop_log(*args, **kwargs):
|
||||
return None
|
||||
|
||||
frontend_calls = []
|
||||
|
||||
async def fake_event_call(payload):
|
||||
frontend_calls.append(payload)
|
||||
return True
|
||||
|
||||
original_generate = module.generate_chat_completion
|
||||
original_get_user = getattr(module.Users, "get_user_by_id", None)
|
||||
|
||||
module.generate_chat_completion = fake_generate_chat_completion
|
||||
module.Users.get_user_by_id = staticmethod(
|
||||
lambda user_id: types.SimpleNamespace(email="user@example.com")
|
||||
)
|
||||
self.filter._log = noop_log
|
||||
self.filter._get_model_thresholds = lambda model_id: {
|
||||
"max_context_tokens": 8192
|
||||
}
|
||||
self.filter._build_summary_prompt = (
|
||||
lambda conversation_text, previous_summary=None: conversation_text
|
||||
)
|
||||
|
||||
try:
|
||||
with self.assertRaises(Exception) as exc_info:
|
||||
asyncio.run(
|
||||
self.filter._call_summary_llm(
|
||||
"conversation",
|
||||
{"model": "fake-summary-model"},
|
||||
{"id": "user-1"},
|
||||
__event_call__=fake_event_call,
|
||||
)
|
||||
)
|
||||
finally:
|
||||
module.generate_chat_completion = original_generate
|
||||
if original_get_user is None:
|
||||
delattr(module.Users, "get_user_by_id")
|
||||
else:
|
||||
module.Users.get_user_by_id = original_get_user
|
||||
|
||||
self.assertIn("Upstream provider error: context too long", str(exc_info.exception))
|
||||
self.assertNotIn(
|
||||
"LLM response format incorrect or empty", str(exc_info.exception)
|
||||
)
|
||||
self.assertTrue(frontend_calls)
|
||||
self.assertEqual(frontend_calls[0]["type"], "execute")
|
||||
self.assertIn("console.error", frontend_calls[0]["data"]["code"])
|
||||
self.assertIn("context too long", frontend_calls[0]["data"]["code"])
|
||||
|
||||
def test_generate_summary_async_status_guides_user_to_browser_console(self):
|
||||
self.filter.valves.keep_first = 1
|
||||
self.filter.valves.keep_last = 1
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
self.filter.valves.summary_model_max_context = 1200
|
||||
self.filter.valves.show_debug_log = False
|
||||
|
||||
events = []
|
||||
frontend_calls = []
|
||||
|
||||
async def fake_summary_llm(*args, **kwargs):
|
||||
raise Exception("boom details")
|
||||
|
||||
async def fake_emitter(event):
|
||||
events.append(event)
|
||||
|
||||
async def fake_event_call(payload):
|
||||
frontend_calls.append(payload)
|
||||
return True
|
||||
|
||||
async def noop_log(*args, **kwargs):
|
||||
return None
|
||||
|
||||
self.filter._log = noop_log
|
||||
self.filter._call_summary_llm = fake_summary_llm
|
||||
self.filter._get_model_thresholds = lambda model_id: {
|
||||
"max_context_tokens": 1200
|
||||
}
|
||||
self.filter._format_messages_for_summary = lambda messages: "\n".join(
|
||||
msg["content"] for msg in messages
|
||||
)
|
||||
self.filter._build_summary_prompt = (
|
||||
lambda conversation_text, previous_summary=None: conversation_text
|
||||
)
|
||||
self.filter._count_tokens = lambda text: len(text)
|
||||
|
||||
messages = [
|
||||
{"role": "system", "content": "System prompt"},
|
||||
{"role": "user", "content": "Q" * 40},
|
||||
{"role": "assistant", "content": "A" * 40},
|
||||
{"role": "user", "content": "Question 2"},
|
||||
]
|
||||
|
||||
asyncio.run(
|
||||
self.filter._generate_summary_async(
|
||||
messages=messages,
|
||||
chat_id="chat-1",
|
||||
body={"model": "fake-summary-model"},
|
||||
user_data={"id": "user-1"},
|
||||
target_compressed_count=2,
|
||||
lang="en-US",
|
||||
__event_emitter__=fake_emitter,
|
||||
__event_call__=fake_event_call,
|
||||
)
|
||||
)
|
||||
|
||||
self.assertTrue(frontend_calls)
|
||||
self.assertIn("console.error", frontend_calls[0]["data"]["code"])
|
||||
self.assertIn("boom details", frontend_calls[0]["data"]["code"])
|
||||
status_descriptions = [
|
||||
event["data"]["description"]
|
||||
for event in events
|
||||
if event.get("type") == "status"
|
||||
]
|
||||
self.assertTrue(
|
||||
any("Check browser console (F12) for details" in text for text in status_descriptions)
|
||||
)
|
||||
|
||||
def test_check_and_generate_summary_async_forces_frontend_and_status_on_pre_summary_error(
|
||||
self,
|
||||
):
|
||||
self.filter.valves.show_debug_log = False
|
||||
|
||||
events = []
|
||||
frontend_calls = []
|
||||
|
||||
async def fake_emitter(event):
|
||||
events.append(event)
|
||||
|
||||
async def fake_event_call(payload):
|
||||
frontend_calls.append(payload)
|
||||
return True
|
||||
|
||||
async def noop_log(*args, **kwargs):
|
||||
return None
|
||||
|
||||
def fail_estimate(_messages):
|
||||
raise Exception("pre summary boom")
|
||||
|
||||
self.filter._log = noop_log
|
||||
self.filter._estimate_messages_tokens = fail_estimate
|
||||
self.filter._get_model_thresholds = lambda model_id: {
|
||||
"compression_threshold_tokens": 100,
|
||||
"max_context_tokens": 1000,
|
||||
}
|
||||
|
||||
asyncio.run(
|
||||
self.filter._check_and_generate_summary_async(
|
||||
chat_id="chat-1",
|
||||
model="fake-model",
|
||||
body={"messages": [{"role": "user", "content": "Hello"}]},
|
||||
user_data={"id": "user-1"},
|
||||
target_compressed_count=1,
|
||||
lang="en-US",
|
||||
__event_emitter__=fake_emitter,
|
||||
__event_call__=fake_event_call,
|
||||
)
|
||||
)
|
||||
|
||||
self.assertTrue(frontend_calls)
|
||||
self.assertIn("console.error", frontend_calls[0]["data"]["code"])
|
||||
self.assertIn("pre summary boom", frontend_calls[0]["data"]["code"])
|
||||
status_descriptions = [
|
||||
event["data"]["description"]
|
||||
for event in events
|
||||
if event.get("type") == "status"
|
||||
]
|
||||
self.assertTrue(
|
||||
any("Check browser console (F12) for details" in text for text in status_descriptions)
|
||||
)
|
||||
|
||||
def test_external_reference_message_detection_matches_injected_marker(self):
|
||||
message = {
|
||||
"role": "assistant",
|
||||
"content": "External refs",
|
||||
"metadata": {
|
||||
"is_summary": True,
|
||||
"is_external_references": True,
|
||||
"source": "external_references",
|
||||
},
|
||||
}
|
||||
|
||||
self.assertTrue(self.filter._is_external_reference_message(message))
|
||||
|
||||
def test_handle_external_chat_references_falls_back_when_summary_llm_errors(self):
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
self.filter.valves.max_summary_tokens = 4096
|
||||
|
||||
async def fake_summary_llm(*args, **kwargs):
|
||||
raise Exception("reference summary failed")
|
||||
|
||||
self.filter._call_summary_llm = fake_summary_llm
|
||||
self.filter._load_summary_record = lambda chat_id: None
|
||||
self.filter._load_full_chat_messages = lambda chat_id: [
|
||||
{"role": "user", "content": "Referenced question"},
|
||||
{"role": "assistant", "content": "Referenced answer"},
|
||||
]
|
||||
self.filter._format_messages_for_summary = (
|
||||
lambda messages: "Referenced conversation body"
|
||||
)
|
||||
self.filter._get_model_thresholds = lambda model_id: {
|
||||
"max_context_tokens": 5001
|
||||
}
|
||||
self.filter._estimate_messages_tokens = lambda messages: 5000
|
||||
|
||||
body = {
|
||||
"model": "main-model",
|
||||
"messages": [{"role": "user", "content": "Current prompt"}],
|
||||
"metadata": {
|
||||
"files": [
|
||||
{
|
||||
"type": "chat",
|
||||
"id": "chat-ref-1",
|
||||
"name": "Referenced Chat",
|
||||
}
|
||||
]
|
||||
},
|
||||
}
|
||||
|
||||
result = asyncio.run(
|
||||
self.filter._handle_external_chat_references(
|
||||
body,
|
||||
user_data={"id": "user-1"},
|
||||
)
|
||||
)
|
||||
|
||||
self.assertIn("__external_references__", result)
|
||||
self.assertIn(
|
||||
"Referenced conversation body",
|
||||
result["__external_references__"]["content"],
|
||||
)
|
||||
|
||||
def test_generate_referenced_summaries_background_uses_model_context_window_fallback(
|
||||
self,
|
||||
):
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
self.filter.valves.summary_model_max_context = 0
|
||||
self.filter.valves.max_summary_tokens = 64
|
||||
|
||||
captured = {}
|
||||
truncate_calls = []
|
||||
|
||||
async def fake_summary_llm(
|
||||
new_conversation_text,
|
||||
body,
|
||||
user_data,
|
||||
__event_call__=None,
|
||||
__request__=None,
|
||||
previous_summary=None,
|
||||
):
|
||||
captured["conversation_text"] = new_conversation_text
|
||||
return "cached summary"
|
||||
|
||||
async def noop_log(*args, **kwargs):
|
||||
return None
|
||||
|
||||
self.filter._call_summary_llm = fake_summary_llm
|
||||
self.filter._log = noop_log
|
||||
self.filter._save_summary = lambda *args: None
|
||||
self.filter._get_model_thresholds = lambda model_id: {
|
||||
"max_context_tokens": 5000
|
||||
}
|
||||
self.filter._truncate_messages_for_summary = (
|
||||
lambda messages, max_tokens: truncate_calls.append(max_tokens) or "truncated"
|
||||
)
|
||||
|
||||
conversation_text = "x" * 600
|
||||
|
||||
asyncio.run(
|
||||
self.filter._generate_referenced_summaries_background(
|
||||
[
|
||||
{
|
||||
"chat_id": "chat-ref-ctx",
|
||||
"title": "Referenced Chat",
|
||||
"conversation_text": conversation_text,
|
||||
"covers_full_history": True,
|
||||
"covered_message_count": 1,
|
||||
}
|
||||
],
|
||||
user_data={"id": "user-1"},
|
||||
)
|
||||
)
|
||||
|
||||
self.assertEqual(captured["conversation_text"], conversation_text)
|
||||
self.assertEqual(truncate_calls, [])
|
||||
|
||||
def test_generate_referenced_summaries_background_uses_summary_llm_signature(self):
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
|
||||
captured = {}
|
||||
|
||||
async def fake_summary_llm(
|
||||
new_conversation_text,
|
||||
body,
|
||||
user_data,
|
||||
__event_call__=None,
|
||||
__request__=None,
|
||||
previous_summary=None,
|
||||
):
|
||||
captured["conversation_text"] = new_conversation_text
|
||||
captured["body"] = body
|
||||
captured["user_data"] = user_data
|
||||
captured["request"] = __request__
|
||||
captured["previous_summary"] = previous_summary
|
||||
return "cached reference summary"
|
||||
|
||||
def fake_save_summary(chat_id, summary, compressed_count):
|
||||
captured["saved"] = (chat_id, summary, compressed_count)
|
||||
|
||||
async def noop_log(*args, **kwargs):
|
||||
return None
|
||||
|
||||
self.filter._call_summary_llm = fake_summary_llm
|
||||
self.filter._save_summary = fake_save_summary
|
||||
self.filter._log = noop_log
|
||||
|
||||
request = object()
|
||||
|
||||
asyncio.run(
|
||||
self.filter._generate_referenced_summaries_background(
|
||||
[
|
||||
{
|
||||
"chat_id": "chat-ref-1",
|
||||
"title": "Referenced Chat",
|
||||
"conversation_text": "Full referenced conversation",
|
||||
"covers_full_history": True,
|
||||
"covered_message_count": 3,
|
||||
}
|
||||
],
|
||||
user_data={"id": "user-1"},
|
||||
__request__=request,
|
||||
)
|
||||
)
|
||||
|
||||
self.assertEqual(captured["conversation_text"], "Full referenced conversation")
|
||||
self.assertEqual(captured["body"]["model"], "fake-summary-model")
|
||||
self.assertEqual(captured["user_data"], {"id": "user-1"})
|
||||
self.assertIs(captured["request"], request)
|
||||
self.assertIsNone(captured["previous_summary"])
|
||||
self.assertEqual(
|
||||
captured["saved"], ("chat-ref-1", "cached reference summary", 3)
|
||||
)
|
||||
|
||||
def test_generate_referenced_summaries_background_skips_progress_save_for_truncation(self):
|
||||
self.filter.valves.summary_model = "fake-summary-model"
|
||||
self.filter.valves.summary_model_max_context = 100
|
||||
|
||||
saved_calls = []
|
||||
captured = {}
|
||||
|
||||
async def fake_summary_llm(
|
||||
new_conversation_text,
|
||||
body,
|
||||
user_data,
|
||||
__event_call__=None,
|
||||
__request__=None,
|
||||
previous_summary=None,
|
||||
):
|
||||
captured["conversation_text"] = new_conversation_text
|
||||
return "ephemeral summary"
|
||||
|
||||
async def noop_log(*args, **kwargs):
|
||||
return None
|
||||
|
||||
self.filter._call_summary_llm = fake_summary_llm
|
||||
self.filter._save_summary = lambda *args: saved_calls.append(args)
|
||||
self.filter._log = noop_log
|
||||
self.filter._load_full_chat_messages = lambda chat_id: [
|
||||
{"role": "user", "content": "msg 1"},
|
||||
{"role": "assistant", "content": "msg 2"},
|
||||
]
|
||||
self.filter._format_messages_for_summary = lambda messages: "x" * 600
|
||||
self.filter._truncate_messages_for_summary = (
|
||||
lambda messages, max_tokens: "tail only"
|
||||
)
|
||||
|
||||
asyncio.run(
|
||||
self.filter._generate_referenced_summaries_background(
|
||||
[{"chat_id": "chat-ref-2", "title": "Large Referenced Chat"}],
|
||||
user_data={"id": "user-1"},
|
||||
)
|
||||
)
|
||||
|
||||
self.assertEqual(captured["conversation_text"], "tail only")
|
||||
self.assertEqual(saved_calls, [])
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
|
||||
27
plugins/filters/async-context-compression/v1.5.0.md
Normal file
27
plugins/filters/async-context-compression/v1.5.0.md
Normal file
@@ -0,0 +1,27 @@
|
||||
[](https://openwebui.com/f/fujie/async_context_compression)
|
||||
|
||||
## Overview
|
||||
|
||||
Compared with the previous git version (`1.4.2`), this release introduces two major new capabilities: external chat reference summarization and a much stronger multilingual token-estimation pipeline. It also improves the reliability of the surrounding summary workflow, especially when provider-side failures occur.
|
||||
|
||||
**[📖 README](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/README.md)**
|
||||
|
||||
## New Features
|
||||
|
||||
- **External Chat Reference Summaries**: Add support for referenced chat context injection that can reuse cached summaries, inject small referenced chats directly, or generate summaries for larger referenced chats before injection.
|
||||
- **Fast Multilingual Token Estimation**: Replace the old rough `len(text)//4` fallback with a new mixed-script estimation pipeline so preflight decisions stay much closer to actual usage across English, Chinese, Japanese, Korean, Arabic, Cyrillic, Thai, and mixed content.
|
||||
- **Stronger Working-Memory Prompt**: Refined the XML summary prompt so generated working memory preserves more actionable state across general chat, coding tasks, and tool-heavy conversations.
|
||||
- **Clearer Frontend Debug Logs**: Reworked browser-console debug output into grouped structural snapshots that make inlet/outlet state easier to inspect.
|
||||
- **Safer Tool Trimming Defaults**: Enabled native tool-output trimming by default and exposed `tool_trim_threshold_chars` with a 600-character threshold.
|
||||
|
||||
## Bug Fixes
|
||||
|
||||
- **Referenced-Chat Fallback Reliability**: If the new referenced-chat summary path fails, the active request now falls back to direct contextual injection instead of failing the whole chat.
|
||||
- **Correct Summary Budgeting**: Fixed referenced-chat summary preparation so `summary_model_max_context` controls summary-input fitting, while `max_summary_tokens` remains an output cap.
|
||||
- **Visible Background Failures**: Important background summary failures now surface to the browser console and chat status even when `show_debug_log` is disabled.
|
||||
- **Provider Error Surfacing**: Improved summary-call error extraction so non-standard upstream provider error payloads are reported more clearly.
|
||||
|
||||
## Release Notes
|
||||
|
||||
- Bilingual plugin README files and mirrored docs pages were refreshed for the `1.5.0` release.
|
||||
- This release is aimed at reducing silent failure modes and making summary behavior easier to reason about during debugging.
|
||||
27
plugins/filters/async-context-compression/v1.5.0_CN.md
Normal file
27
plugins/filters/async-context-compression/v1.5.0_CN.md
Normal file
@@ -0,0 +1,27 @@
|
||||
[](https://openwebui.com/f/fujie/async_context_compression)
|
||||
|
||||
## 概述
|
||||
|
||||
相较上一个 git 版本（`1.4.2`），本次发布新增了两个重要能力：外部聊天引用摘要，以及更强的多语言 Token 预估链路。同时也补强了围绕这些新能力的摘要流程稳定性，特别是上游提供商报错时的回退与可见性。
|
||||
|
||||
**[📖 README](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/filters/async-context-compression/README_CN.md)**
|
||||
|
||||
## 新功能
|
||||
|
||||
- **外部聊天引用摘要**：新增引用聊天上下文注入能力。现在可以复用缓存摘要、直接注入较小引用聊天，或先为较大的引用聊天生成摘要再注入。
|
||||
- **快速多语言 Token 预估**：用新的混合脚本估算链路替代旧的 `len(text)//4` 粗略回退，使预检在英文、中文、日文、韩文、阿拉伯文、西里尔文、泰文及混合内容下都更接近真实用量。
|
||||
- **更稳健的工作记忆提示词**：重写 XML 摘要提示词，让生成出的 working memory 在普通聊天、编码任务和密集工具调用场景下保留更多可操作上下文。
|
||||
- **更清晰的前端调试日志**：浏览器控制台调试输出改为分组化、结构化展示，更容易观察 inlet / outlet 的真实状态。
|
||||
- **更安全的工具裁剪默认值**：原生工具输出裁剪默认开启，并新增 `tool_trim_threshold_chars`，默认阈值为 600 字符。
|
||||
|
||||
## 问题修复
|
||||
|
||||
- **引用聊天回退更稳妥**：当新的引用聊天摘要路径失败时，当前请求会自动回退为直接注入上下文，而不是整个对话一起失败。
|
||||
- **摘要预算计算更准确**：修复引用聊天摘要准备逻辑，明确由 `summary_model_max_context` 控制摘要输入窗口，而 `max_summary_tokens` 只控制摘要输出长度。
|
||||
- **后台失败更容易发现**：即使关闭 `show_debug_log`，关键后台摘要失败现在也会显示到浏览器控制台和聊天状态提示中。
|
||||
- **提供商错误信息更清晰**：改进摘要调用的错误提取逻辑，让非标准上游错误载荷也能更准确地显示出来。
|
||||
|
||||
## 发布说明
|
||||
|
||||
- 已同步更新中英插件 README 与 docs 镜像页，确保 `1.5.0` 发布说明一致。
|
||||
- 本次版本的目标，是减少“静默失败”这类难排查问题，并让摘要行为在调试时更容易理解。
|
||||
137
plugins/tools/batch-install-plugins/README.md
Normal file
137
plugins/tools/batch-install-plugins/README.md
Normal file
@@ -0,0 +1,137 @@
|
||||
# Batch Install Plugins from GitHub
|
||||
|
||||
**Author:** [Fu-Jie](https://github.com/Fu-Jie) | **Version:** 1.0.0 | **Project:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions)
|
||||
|
||||
One-click batch install plugins from GitHub repositories to your OpenWebUI instance.
|
||||
|
||||
## Key Features
|
||||
|
||||
- **One-Click Install**: Install all plugins with a single command
|
||||
- **Auto-Update**: Automatically updates previously installed plugins
|
||||
- **GitHub Support**: Install plugins from any GitHub repository
|
||||
- **Multi-Type Support**: Supports Pipe, Action, Filter, and Tool plugins
|
||||
- **Confirmation**: Shows plugin list before installing, allows selective installation
|
||||
- **i18n**: Supports 11 languages
|
||||
|
||||
## Flow
|
||||
|
||||
```
|
||||
User Input
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Discover Plugins from GitHub │
|
||||
│ (fetch file tree + parse .py) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Filter by Type & Keywords │
|
||||
│ (tool/filter/pipe/action) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Show Confirmation Dialog │
|
||||
│ (list plugins + exclude hint) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
├── [Cancel] → End
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ Install to OpenWebUI │
|
||||
│ (update or create each plugin) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
Done
|
||||
```
|
||||
|
||||
## How to Use
|
||||
|
||||
1. Open OpenWebUI and go to **Workspace > Tools**
|
||||
2. Install **Batch Install Plugins from GitHub** from the official marketplace
|
||||
3. Enable this tool for your model/chat
|
||||
4. Ask the model to install plugins
|
||||
|
||||
## Usage Examples
|
||||
|
||||
```
|
||||
"Install all plugins"
|
||||
"Install all plugins from github.com/username/repo"
|
||||
"Install only pipe plugins"
|
||||
"Install action and filter plugins"
|
||||
"Install all plugins, exclude_keywords=copilot"
|
||||
```
|
||||
|
||||
## Popular Plugin Repositories
|
||||
|
||||
Here are some popular repositories with many plugins you can install:
|
||||
|
||||
### Community Collections
|
||||
|
||||
```
|
||||
# Install all plugins from iChristGit's collection
|
||||
"Install all plugins from iChristGit/OpenWebui-Tools"
|
||||
|
||||
# Install all tools from Haervwe's tools collection
|
||||
"Install all plugins from Haervwe/open-webui-tools"
|
||||
|
||||
# Install all plugins from Classic298's repository
|
||||
"Install all plugins from Classic298/open-webui-plugins"
|
||||
|
||||
# Install all functions from suurt8ll's collection
|
||||
"Install all plugins from suurt8ll/open_webui_functions"
|
||||
|
||||
# Install only specific types (e.g., only tools)
|
||||
"Install only tool plugins from iChristGit/OpenWebui-Tools"
|
||||
|
||||
# Exclude certain keywords while installing
|
||||
"Install all plugins from Haervwe/open-webui-tools, exclude_keywords=test,deprecated"
|
||||
```
|
||||
|
||||
### Supported Repositories
|
||||
|
||||
- `Fu-Jie/openwebui-extensions` - Default, official plugin collection
|
||||
- `iChristGit/OpenWebui-Tools` - Comprehensive tool and plugin collection
|
||||
- `Haervwe/open-webui-tools` - Specialized tools and utilities
|
||||
- `Classic298/open-webui-plugins` - Various plugin implementations
|
||||
- `suurt8ll/open_webui_functions` - Function-based plugins
|
||||
|
||||
## Default Repository
|
||||
|
||||
When no repository is specified, defaults to `Fu-Jie/openwebui-extensions`.
|
||||
|
||||
## Plugin Detection Rules
|
||||
|
||||
### Fu-Jie/openwebui-extensions (Strict)
|
||||
|
||||
For the default repository, plugins must have:
|
||||
1. A `.py` file containing `class Tools:`, `class Filter:`, `class Pipe:`, or `class Action:`
|
||||
2. A docstring with `title:`, `description:`, and **`openwebui_id:`** fields
|
||||
3. Filename must not end with `_cn`
|
||||
|
||||
### Other GitHub Repositories
|
||||
|
||||
For other repositories:
|
||||
1. A `.py` file containing `class Tools:`, `class Filter:`, `class Pipe:`, or `class Action:`
|
||||
2. A docstring with `title:` and `description:` fields
|
||||
|
||||
## Configuration (Valves)
|
||||
|
||||
| Parameter | Default | Description |
|
||||
| --- | --- | --- |
|
||||
| `SKIP_KEYWORDS` | `test,verify,example,template,mock` | Comma-separated keywords to skip |
|
||||
| `TIMEOUT` | `20` | Request timeout in seconds |
|
||||
|
||||
## Confirmation Timeout
|
||||
|
||||
User confirmation dialogs have a default timeout of **2 minutes (120 seconds)**, allowing sufficient time for users to:
|
||||
- Read and review the plugin list
|
||||
- Make installation decisions
|
||||
- Handle network delays
|
||||
|
||||
## Support
|
||||
|
||||
If this plugin has been useful, a star on [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) is a big motivation for me. Thank you for the support.
|
||||
137
plugins/tools/batch-install-plugins/README_CN.md
Normal file
137
plugins/tools/batch-install-plugins/README_CN.md
Normal file
@@ -0,0 +1,137 @@
|
||||
# Batch Install Plugins from GitHub
|
||||
|
||||
**作者:** [Fu-Jie](https://github.com/Fu-Jie) | **版本:** 1.0.0 | **项目:** [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions)
|
||||
|
||||
一键将 GitHub 仓库中的插件批量安装到你的 OpenWebUI 实例。
|
||||
|
||||
## 主要功能
|
||||
|
||||
- 一键安装：单个命令安装所有插件
|
||||
- 自动更新：自动更新之前安装过的插件
|
||||
- GitHub 支持：从任意 GitHub 仓库安装插件
|
||||
- 多类型支持：支持 Pipe、Action、Filter 和 Tool 插件
|
||||
- 安装确认：安装前显示插件列表，支持选择性安装
|
||||
- 国际化：支持 11 种语言
|
||||
|
||||
## 流程
|
||||
|
||||
```
|
||||
用户输入
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 从 GitHub 发现插件 │
|
||||
│ (获取文件树 + 解析 .py 文件) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 按类型和关键词过滤 │
|
||||
│ (tool/filter/pipe/action) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 显示确认对话框 │
|
||||
│ (插件列表 + 排除提示) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
├── [取消] → 结束
|
||||
│
|
||||
▼
|
||||
┌─────────────────────────────────────┐
|
||||
│ 安装到 OpenWebUI │
|
||||
│ (更新或创建每个插件) │
|
||||
└─────────────────────────────────────┘
|
||||
│
|
||||
▼
|
||||
完成
|
||||
```
|
||||
|
||||
## 使用方法
|
||||
|
||||
1. 打开 OpenWebUI，进入 **Workspace > Tools**
|
||||
2. 从官方市场安装 **Batch Install Plugins from GitHub**
|
||||
3. 为你的模型/对话启用此工具
|
||||
4. 让模型调用工具方法
|
||||
|
||||
## 使用示例
|
||||
|
||||
```
|
||||
"安装所有插件"
|
||||
"从 github.com/username/repo 安装所有插件"
|
||||
"只安装 pipe 插件"
|
||||
"安装 action 和 filter 插件"
|
||||
"安装所有插件, exclude_keywords=copilot"
|
||||
```
|
||||
|
||||
## 热门插件仓库
|
||||
|
||||
这些是包含大量插件的热门仓库，你可以从中安装插件：
|
||||
|
||||
### 社区合集
|
||||
|
||||
```
|
||||
# 从 iChristGit 的集合安装所有插件
|
||||
"从 iChristGit/OpenWebui-Tools 安装所有插件"
|
||||
|
||||
# 从 Haervwe 的工具集合只安装工具
|
||||
"从 Haervwe/open-webui-tools 安装所有插件"
|
||||
|
||||
# 从 Classic298 的仓库安装所有插件
|
||||
"从 Classic298/open-webui-plugins 安装所有插件"
|
||||
|
||||
# 从 suurt8ll 的集合安装所有函数
|
||||
"从 suurt8ll/open_webui_functions 安装所有插件"
|
||||
|
||||
# 只安装特定类型的插件（比如只安装工具）
|
||||
"从 iChristGit/OpenWebui-Tools 只安装 tool 插件"
|
||||
|
||||
# 安装时排除特定关键词
|
||||
"从 Haervwe/open-webui-tools 安装所有插件, exclude_keywords=test,deprecated"
|
||||
```
|
||||
|
||||
### 支持的仓库
|
||||
|
||||
- `Fu-Jie/openwebui-extensions` - 默认的官方插件集合
|
||||
- `iChristGit/OpenWebui-Tools` - 全面的工具和插件集合
|
||||
- `Haervwe/open-webui-tools` - 专业的工具和实用程序
|
||||
- `Classic298/open-webui-plugins` - 各种插件实现
|
||||
- `suurt8ll/open_webui_functions` - 基于函数的插件
|
||||
|
||||
## 默认仓库
|
||||
|
||||
未指定仓库时，默认为 `Fu-Jie/openwebui-extensions`。
|
||||
|
||||
## 插件检测规则
|
||||
|
||||
### Fu-Jie/openwebui-extensions（严格模式）
|
||||
|
||||
默认仓库的插件必须满足：
|
||||
1. 包含 `class Tools:`、`class Filter:`、`class Pipe:` 或 `class Action:` 的 `.py` 文件
|
||||
2. Docstring 中包含 `title:`、`description:` 和 **`openwebui_id:`** 字段
|
||||
3. 文件名不能以 `_cn` 结尾
|
||||
|
||||
### 其他 GitHub 仓库
|
||||
|
||||
其他仓库的插件必须满足：
|
||||
1. 包含 `class Tools:`、`class Filter:`、`class Pipe:` 或 `class Action:` 的 `.py` 文件
|
||||
2. Docstring 中包含 `title:` 和 `description:` 字段
|
||||
|
||||
## 配置（Valves）
|
||||
|
||||
| 参数 | 默认值 | 描述 |
|
||||
| --- | --- | --- |
|
||||
| `SKIP_KEYWORDS` | `test,verify,example,template,mock` | 逗号分隔的跳过关键词 |
|
||||
| `TIMEOUT` | `20` | 请求超时时间（秒）|
|
||||
|
||||
## 确认超时时间
|
||||
|
||||
用户确认对话框的默认超时时间为 **2 分钟（120 秒）**，为用户提供充足的时间来：
|
||||
- 阅读和查看插件列表
|
||||
- 做出安装决定
|
||||
- 处理网络延迟
|
||||
|
||||
## 支持
|
||||
|
||||
如果这个插件对你有帮助，欢迎到 [OpenWebUI Extensions](https://github.com/Fu-Jie/openwebui-extensions) 点个 Star，这将是我持续改进的动力，感谢支持。
|
||||
1262
plugins/tools/batch-install-plugins/batch_install_plugins.py
Normal file
1262
plugins/tools/batch-install-plugins/batch_install_plugins.py
Normal file
File diff suppressed because it is too large
Load Diff
67
plugins/tools/batch-install-plugins/v1.0.0.md
Normal file
67
plugins/tools/batch-install-plugins/v1.0.0.md
Normal file
@@ -0,0 +1,67 @@
|
||||
[](https://openwebui.com/t/fujie/batch_install_plugins)
|
||||
|
||||
## Overview
|
||||
|
||||
Batch Install Plugins from GitHub is a new tool for OpenWebUI that enables one-click installation of multiple plugins directly from GitHub repositories. This initial release includes comprehensive features for discovering, filtering, and installing plugins with user confirmation, extensive multi-language support, and robust debugging capabilities for container deployments.
|
||||
|
||||
**[📖 README](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/tools/batch-install-plugins/README.md)**
|
||||
|
||||
## Features
|
||||
|
||||
- **One-Click Installation**: Install all plugins from a repository with a single command
|
||||
- **Smart Plugin Discovery**: Parse Python files to extract metadata and validate plugins automatically
|
||||
- **Multi-Type Support**: Support Pipe, Action, Filter, and Tool plugins in a single operation
|
||||
- **Confirmation Dialog**: Display plugin list before installation for user review and approval
|
||||
- **Selective Installation**: Exclude specific plugins using keyword-based filtering
|
||||
- **Smart Fallback**: Container deployments auto-retry with localhost:8080 if primary connection fails
|
||||
- **Enhanced Debugging**: Rich frontend JavaScript and backend Python logs for troubleshooting
|
||||
- **Extended Timeout**: 120-second confirmation window for thoughtful decision-making
|
||||
- **Async Architecture**: Non-blocking I/O operations for better performance
|
||||
- **Full Internationalization**: Complete support for 11 languages with proper fallback maps
|
||||
- **Auto-Update**: Automatically updates previously installed plugins
|
||||
- **Self-Exclusion**: Automatically excludes the tool itself from batch operations
|
||||
|
||||
## Technical Highlights
|
||||
|
||||
- **httpx Integration**: Modern async HTTP client for reliable, non-blocking requests
|
||||
- **Event Emitter Support**: Proper handling of OpenWebUI event injection with fallbacks
|
||||
- **Timeout Protection**: Wrapped frontend execution with timeout guards to prevent hanging
|
||||
- **Filtered List Consistency**: Uses single source of truth for confirmation and installation
|
||||
- **Error Localization**: All error messages are user-facing and properly localized across languages
|
||||
- **Deployment Resilience**: Intelligent base URL resolution handles domain, localhost, and containerized environments
|
||||
|
||||
## Supported Repositories
|
||||
|
||||
- **Default**: Fu-Jie/openwebui-extensions (strict validation)
|
||||
- **Custom**: Any GitHub repository with Python plugin files
|
||||
|
||||
## Testing
|
||||
|
||||
Comprehensive regression tests included:
|
||||
- Filtered installation list consistency
|
||||
- Missing event emitter handling
|
||||
- Confirmation timeout verification
|
||||
- Full failure scenarios
|
||||
- Localization completeness
|
||||
- Connection error debug logging and smart fallback
|
||||
|
||||
All 6 tests pass successfully.
|
||||
|
||||
## Documentation
|
||||
|
||||
- English README with flow diagrams and usage examples
|
||||
- Chinese README (README_CN.md) with complete translations
|
||||
- Mirrored documentation for official docs site
|
||||
- Plugin index entries in both English and Chinese
|
||||
|
||||
## Compatibility
|
||||
|
||||
- OpenWebUI: 0.2.x - 0.8.x
|
||||
- Python: 3.9+
|
||||
- Dependencies: httpx (async HTTP client), pydantic (type validation)
|
||||
|
||||
## Release Notes
|
||||
|
||||
- This initial v1.0.0 release includes complete plugin infrastructure with smart deployment handling.
|
||||
- The plugin is designed to handle diverse deployment scenarios (domain, localhost, containerized) with minimal configuration.
|
||||
|
||||
67
plugins/tools/batch-install-plugins/v1.0.0_CN.md
Normal file
67
plugins/tools/batch-install-plugins/v1.0.0_CN.md
Normal file
@@ -0,0 +1,67 @@
|
||||
[](https://openwebui.com/t/fujie/batch_install_plugins)
|
||||
|
||||
## 概述
|
||||
|
||||
从 GitHub 批量安装插件是一款全新的 OpenWebUI 工具，支持直接从 GitHub 仓库一键安装多个插件。此首个发布版本包含了全面的插件发现、过滤和安装功能，支持用户确认流程、广泛的多语言支持，以及针对容器部署的健壮调试能力。
|
||||
|
||||
**[📖 README](https://github.com/Fu-Jie/openwebui-extensions/blob/main/plugins/tools/batch-install-plugins/README_CN.md)**
|
||||
|
||||
## 主要功能
|
||||
|
||||
- **一键安装**：通过单个命令安装仓库中的所有插件
|
||||
- **智能插件发现**：解析 Python 文件提取元数据并自动验证插件
|
||||
- **多类型支持**：在单个操作中支持 Pipe、Action、Filter 和 Tool 插件
|
||||
- **确认对话框**：安装前显示插件列表供用户审查和批准
|
||||
- **选择性安装**：通过基于关键词的过滤排除特定插件
|
||||
- **智能降级**：容器环境中主 URL 连接失败时自动重试 localhost:8080
|
||||
- **增强调试**：前端 JavaScript 和后端 Python 富日志输出，便于排查问题
|
||||
- **延长超时**：120 秒确认窗口，给用户充分的思考时间
|
||||
- **异步架构**：非阻塞 I/O 操作，性能更优
|
||||
- **完整国际化**：支持 11 种语言，包含适当的回退机制
|
||||
- **自动更新**：自动更新之前安装过的插件
|
||||
- **自排除机制**：自动排除工具自身，避免在批量操作中重复安装
|
||||
|
||||
## 技术亮点
|
||||
|
||||
- **httpx 集成**：现代化的异步 HTTP 客户端，请求更可靠且非阻塞
|
||||
- **事件注入支持**：正确处理 OpenWebUI 事件注入，提供回退支持
|
||||
- **超时保护**：前端执行周围包装了超时保护，防止进程挂起
|
||||
- **过滤列表一致性**：确认和安装使用同一份过滤列表
|
||||
- **错误本地化**：所有错误消息都是面向用户的，已正确本地化到各语言
|
||||
- **部署弹性**：智能 Base URL 解析处理域名、localhost 和容器化环境
|
||||
|
||||
## 支持的仓库
|
||||
|
||||
- **默认**：Fu-Jie/openwebui-extensions（严格验证）
|
||||
- **自定义**：任意 GitHub 仓库中的 Python 插件文件
|
||||
|
||||
## 测试覆盖
|
||||
|
||||
包含全面的回归测试：
|
||||
- 过滤安装列表一致性
|
||||
- 缺少事件注入器时的处理
|
||||
- 确认超时验证
|
||||
- 完全失败场景
|
||||
- 本地化完整性
|
||||
- 连接错误调试日志和智能降级
|
||||
|
||||
所有 6 个测试均通过。
|
||||
|
||||
## 文档
|
||||
|
||||
- 英文 README，包含流程图和使用示例
|
||||
- 中文 README (README_CN.md)，完整翻译
|
||||
- 官方文档站点的镜像文档
|
||||
- 英文和中文的插件索引条目
|
||||
|
||||
## 兼容性
|
||||
|
||||
- OpenWebUI：0.2.x - 0.8.x
|
||||
- Python：3.9+
|
||||
- 依赖：httpx（异步 HTTP 客户端）、pydantic（类型验证）
|
||||
|
||||
## 发布说明
|
||||
|
||||
- 本首发 v1.0.0 版本包含完整的插件基础设施和智能部署处理能力。
|
||||
- 该插件设计用于处理多种部署场景（域名、localhost、容器化），配置最少。
|
||||
|
||||
302
tests/plugins/tools/test_batch_install_plugins.py
Normal file
302
tests/plugins/tools/test_batch_install_plugins.py
Normal file
@@ -0,0 +1,302 @@
|
||||
import asyncio
|
||||
import importlib.util
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
import httpx
|
||||
import pytest
|
||||
|
||||
|
||||
MODULE_PATH = (
|
||||
Path(__file__).resolve().parents[3]
|
||||
/ "plugins"
|
||||
/ "tools"
|
||||
/ "batch-install-plugins"
|
||||
/ "batch_install_plugins.py"
|
||||
)
|
||||
SPEC = importlib.util.spec_from_file_location("batch_install_plugins", MODULE_PATH)
|
||||
batch_install_plugins = importlib.util.module_from_spec(SPEC)
|
||||
assert SPEC.loader is not None
|
||||
sys.modules[SPEC.name] = batch_install_plugins
|
||||
SPEC.loader.exec_module(batch_install_plugins)
|
||||
|
||||
|
||||
def make_candidate(title: str, file_path: str, function_id: str):
|
||||
return batch_install_plugins.PluginCandidate(
|
||||
plugin_type="tool",
|
||||
file_path=file_path,
|
||||
metadata={"title": title, "description": f"{title} description"},
|
||||
content="class Tools:\n pass\n",
|
||||
function_id=function_id,
|
||||
)
|
||||
|
||||
|
||||
def make_request():
|
||||
class Request:
|
||||
base_url = "http://localhost:3000/"
|
||||
headers = {"Authorization": "Bearer token"}
|
||||
|
||||
return Request()
|
||||
|
||||
|
||||
class DummyResponse:
|
||||
def __init__(self, status_code: int, json_data=None, text: str = ""):
|
||||
self.status_code = status_code
|
||||
self._json_data = json_data
|
||||
self.text = text
|
||||
|
||||
def json(self):
|
||||
if self._json_data is None:
|
||||
raise ValueError("no json body")
|
||||
return self._json_data
|
||||
|
||||
|
||||
class FakeAsyncClient:
|
||||
posts = []
|
||||
responses = []
|
||||
|
||||
def __init__(self, *args, **kwargs):
|
||||
pass
|
||||
|
||||
async def __aenter__(self):
|
||||
return self
|
||||
|
||||
async def __aexit__(self, exc_type, exc, tb):
|
||||
return False
|
||||
|
||||
async def post(self, url, headers=None, json=None):
|
||||
type(self).posts.append((url, headers, json))
|
||||
if not type(self).responses:
|
||||
raise AssertionError("No fake response configured for POST request")
|
||||
response = type(self).responses.pop(0)
|
||||
if isinstance(response, Exception):
|
||||
raise response
|
||||
return response
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_install_all_plugins_only_installs_filtered_candidates(monkeypatch):
|
||||
keep = make_candidate("Keep Plugin", "plugins/tools/keep/keep.py", "keep_plugin")
|
||||
exclude = make_candidate(
|
||||
"Exclude Me",
|
||||
"plugins/tools/exclude-me/exclude_me.py",
|
||||
"exclude_me",
|
||||
)
|
||||
self_plugin = make_candidate(
|
||||
"Batch Install Plugins from GitHub",
|
||||
"plugins/tools/batch-install-plugins/batch_install_plugins.py",
|
||||
"batch_install_plugins",
|
||||
)
|
||||
|
||||
async def fake_discover_plugins(url, skip_keywords):
|
||||
return [keep, exclude, self_plugin], []
|
||||
|
||||
monkeypatch.setattr(batch_install_plugins, "discover_plugins", fake_discover_plugins)
|
||||
FakeAsyncClient.posts = []
|
||||
FakeAsyncClient.responses = [DummyResponse(404), DummyResponse(201)]
|
||||
monkeypatch.setattr(batch_install_plugins.httpx, "AsyncClient", FakeAsyncClient)
|
||||
|
||||
events = []
|
||||
captured = {}
|
||||
|
||||
async def event_call(payload):
|
||||
if payload["type"] == "confirmation":
|
||||
captured["message"] = payload["data"]["message"]
|
||||
elif payload["type"] == "execute":
|
||||
captured.setdefault("execute_codes", []).append(payload["data"]["code"])
|
||||
return True
|
||||
|
||||
async def emitter(event):
|
||||
events.append(event)
|
||||
|
||||
result = await batch_install_plugins.Tools().install_all_plugins(
|
||||
__user__={"id": "u1", "language": "en-US"},
|
||||
__event_call__=event_call,
|
||||
__request__=make_request(),
|
||||
__event_emitter__=emitter,
|
||||
repo=batch_install_plugins.DEFAULT_REPO,
|
||||
plugin_types=["tool"],
|
||||
exclude_keywords="exclude",
|
||||
)
|
||||
|
||||
assert "Created: Keep Plugin" in result
|
||||
assert "Exclude Me" not in result
|
||||
assert "1/1" in result
|
||||
assert captured["message"].count("[tool]") == 1
|
||||
assert "Keep Plugin" in captured["message"]
|
||||
assert "Exclude Me" not in captured["message"]
|
||||
assert "Batch Install Plugins from GitHub" not in captured["message"]
|
||||
assert "exclude, batch-install-plugins" in captured["message"]
|
||||
|
||||
urls = [url for url, _, _ in FakeAsyncClient.posts]
|
||||
assert urls == [
|
||||
"http://localhost:3000/api/v1/tools/id/keep_plugin/update",
|
||||
"http://localhost:3000/api/v1/tools/create",
|
||||
]
|
||||
assert any(
|
||||
"Starting OpenWebUI install requests" in code
|
||||
for code in captured.get("execute_codes", [])
|
||||
)
|
||||
assert events[-1]["type"] == "notification"
|
||||
assert events[-1]["data"]["type"] == "success"
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_install_all_plugins_supports_missing_event_emitter(monkeypatch):
|
||||
keep = make_candidate("Keep Plugin", "plugins/tools/keep/keep.py", "keep_plugin")
|
||||
|
||||
async def fake_discover_plugins(url, skip_keywords):
|
||||
return [keep], []
|
||||
|
||||
monkeypatch.setattr(batch_install_plugins, "discover_plugins", fake_discover_plugins)
|
||||
FakeAsyncClient.posts = []
|
||||
FakeAsyncClient.responses = [DummyResponse(404), DummyResponse(201)]
|
||||
monkeypatch.setattr(batch_install_plugins.httpx, "AsyncClient", FakeAsyncClient)
|
||||
|
||||
result = await batch_install_plugins.Tools().install_all_plugins(
|
||||
__user__={"id": "u1", "language": "en-US"},
|
||||
__request__=make_request(),
|
||||
repo="example/repo",
|
||||
plugin_types=["tool"],
|
||||
)
|
||||
|
||||
assert "Created: Keep Plugin" in result
|
||||
assert "1/1" in result
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_install_all_plugins_handles_confirmation_timeout(monkeypatch):
|
||||
keep = make_candidate("Keep Plugin", "plugins/tools/keep/keep.py", "keep_plugin")
|
||||
|
||||
async def fake_discover_plugins(url, skip_keywords):
|
||||
return [keep], []
|
||||
|
||||
async def fake_wait_for(awaitable, timeout):
|
||||
awaitable.close()
|
||||
raise asyncio.TimeoutError
|
||||
|
||||
monkeypatch.setattr(batch_install_plugins, "discover_plugins", fake_discover_plugins)
|
||||
monkeypatch.setattr(batch_install_plugins.asyncio, "wait_for", fake_wait_for)
|
||||
|
||||
events = []
|
||||
|
||||
async def event_call(payload):
|
||||
return True
|
||||
|
||||
async def emitter(event):
|
||||
events.append(event)
|
||||
|
||||
result = await batch_install_plugins.Tools().install_all_plugins(
|
||||
__user__={"id": "u1", "language": "en-US"},
|
||||
__event_call__=event_call,
|
||||
__request__=make_request(),
|
||||
__event_emitter__=emitter,
|
||||
repo="example/repo",
|
||||
plugin_types=["tool"],
|
||||
)
|
||||
|
||||
assert result == "Confirmation timed out or failed. Installation cancelled."
|
||||
assert events[-1]["type"] == "notification"
|
||||
assert events[-1]["data"]["type"] == "warning"
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_install_all_plugins_marks_total_failure_as_error(monkeypatch):
|
||||
keep = make_candidate("Keep Plugin", "plugins/tools/keep/keep.py", "keep_plugin")
|
||||
|
||||
async def fake_discover_plugins(url, skip_keywords):
|
||||
return [keep], []
|
||||
|
||||
monkeypatch.setattr(batch_install_plugins, "discover_plugins", fake_discover_plugins)
|
||||
FakeAsyncClient.posts = []
|
||||
FakeAsyncClient.responses = [
|
||||
DummyResponse(500, {"detail": "update failed"}, "update failed"),
|
||||
DummyResponse(500, {"detail": "create failed"}, "create failed"),
|
||||
]
|
||||
monkeypatch.setattr(batch_install_plugins.httpx, "AsyncClient", FakeAsyncClient)
|
||||
|
||||
events = []
|
||||
|
||||
async def emitter(event):
|
||||
events.append(event)
|
||||
|
||||
result = await batch_install_plugins.Tools().install_all_plugins(
|
||||
__user__={"id": "u1", "language": "en-US"},
|
||||
__request__=make_request(),
|
||||
__event_emitter__=emitter,
|
||||
repo="example/repo",
|
||||
plugin_types=["tool"],
|
||||
)
|
||||
|
||||
assert "Failed: Keep Plugin - status 500:" in result
|
||||
assert "0/1" in result
|
||||
assert events[-1]["type"] == "notification"
|
||||
assert events[-1]["data"]["type"] == "error"
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_install_all_plugins_localizes_timeout_errors(monkeypatch):
|
||||
keep = make_candidate("Keep Plugin", "plugins/tools/keep/keep.py", "keep_plugin")
|
||||
|
||||
async def fake_discover_plugins(url, skip_keywords):
|
||||
return [keep], []
|
||||
|
||||
monkeypatch.setattr(batch_install_plugins, "discover_plugins", fake_discover_plugins)
|
||||
FakeAsyncClient.posts = []
|
||||
FakeAsyncClient.responses = [httpx.TimeoutException("timed out")]
|
||||
monkeypatch.setattr(batch_install_plugins.httpx, "AsyncClient", FakeAsyncClient)
|
||||
|
||||
result = await batch_install_plugins.Tools().install_all_plugins(
|
||||
__user__={"id": "u1", "language": "zh-CN"},
|
||||
__request__=make_request(),
|
||||
repo="example/repo",
|
||||
plugin_types=["tool"],
|
||||
)
|
||||
|
||||
assert "失败：Keep Plugin - 请求超时" in result
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_install_all_plugins_emits_frontend_debug_logs_on_connect_error(
|
||||
monkeypatch,
|
||||
):
|
||||
keep = make_candidate("Keep Plugin", "plugins/tools/keep/keep.py", "keep_plugin")
|
||||
|
||||
async def fake_discover_plugins(url, skip_keywords):
|
||||
return [keep], []
|
||||
|
||||
monkeypatch.setattr(batch_install_plugins, "discover_plugins", fake_discover_plugins)
|
||||
FakeAsyncClient.posts = []
|
||||
# Both initial attempt and fallback retry should fail
|
||||
FakeAsyncClient.responses = [httpx.ConnectError("connect failed"), httpx.ConnectError("connect failed")]
|
||||
monkeypatch.setattr(batch_install_plugins.httpx, "AsyncClient", FakeAsyncClient)
|
||||
|
||||
execute_codes = []
|
||||
events = []
|
||||
|
||||
async def event_call(payload):
|
||||
if payload["type"] == "execute":
|
||||
execute_codes.append(payload["data"]["code"])
|
||||
return True
|
||||
if payload["type"] == "confirmation":
|
||||
return True
|
||||
raise AssertionError(f"Unexpected event_call payload type: {payload['type']}")
|
||||
|
||||
async def emitter(event):
|
||||
events.append(event)
|
||||
|
||||
result = await batch_install_plugins.Tools().install_all_plugins(
|
||||
__user__={"id": "u1", "language": "en-US"},
|
||||
__event_call__=event_call,
|
||||
__request__=make_request(),
|
||||
__event_emitter__=emitter,
|
||||
repo="example/repo",
|
||||
plugin_types=["tool"],
|
||||
)
|
||||
|
||||
assert result == "Cannot connect to OpenWebUI. Is it running?"
|
||||
assert any("OpenWebUI connection failed" in code for code in execute_codes)
|
||||
assert any("console.error" in code for code in execute_codes)
|
||||
assert any("http://localhost:3000" in code for code in execute_codes)
|
||||
assert events[-1]["type"] == "notification"
|
||||
assert events[-1]["data"]["type"] == "error"
|
||||
Reference in New Issue
Block a user