modelscope · wangxingjun778 · Jun 18, 2026 · Jun 18, 2026 · gemini-code-assist · Jun 18, 2026
diff --git a/content/_index.md b/content/_index.md
@@ -20,7 +20,7 @@ sections:
         text: Read the Technical Report
         url: blog/technical-deep-dive/
       announcement:
-        text: "Sirchmunk v0.0.6post1 — OpenClaw Skill, SSE Streaming API & Multi-turn Conversation"
+        text: "Sirchmunk v0.0.8 — Knowledge Compile (Beta), DEEP Mode Generalization & I/O Optimization"
         link:
           text: "View all releases"
           url: "https://github.com/modelscope/sirchmunk/releases"

diff --git a/content/_index.zh.md b/content/_index.zh.md
@@ -20,7 +20,7 @@ sections:
         text: 阅读技术报告
         url: blog/technical-deep-dive/
       announcement:
-        text: "Sirchmunk v0.0.6post1 — OpenClaw 技能支持、SSE 流式 API 与多轮对话"
+        text: "Sirchmunk v0.0.8 — 知识编译（Beta）、DEEP 模式泛化增强与 I/O 优化"
         link:
           text: "查看所有版本"
           url: "https://github.com/modelscope/sirchmunk/releases"

diff --git a/content/docs/guide/architecture.md b/content/docs/guide/architecture.md
@@ -18,6 +18,8 @@ Sirchmunk's architecture is organized into cleanly separated layers, following t
 | **EvidenceProcessor** | Evidence processing based on the Monte Carlo Importance Sampling         |
 | **GrepRetriever**     | High-performance _indexless_ file search with parallel processing        |
 | **OpenAIChat**        | Unified LLM interface supporting streaming and usage tracking            |
+| **KnowledgeCompiler** | Offline document compilation into tree indices and knowledge clusters (Beta) |
+| **KnowledgeLint**     | Knowledge health checks and auto-repair                                  |
 | **MonitorTracker**    | Real-time system and application metrics collection                      |
 
 ## Multi-Phase Search Pipeline
@@ -166,6 +168,13 @@ All persistent data is stored in the configured `SIRCHMUNK_WORK_PATH` (default:
     │   └── chat_history.db
     ├── knowledge/            # Knowledge clusters (Parquet)
     │   └── knowledge_clusters.parquet
+    ├── compile/              # Compile artifacts (Beta)
+    │   ├── manifest.json     # File manifest with hashes
+    │   ├── document_catalog.json
+    │   ├── summary_index.json
+    │   ├── trees/            # Hierarchical tree indices
+    │   ├── table_digests/    # Table extraction digests
+    │   └── xlsx_digests/     # Spreadsheet digests
     └── settings/             # User settings (DuckDB)
         └── settings.db
 ```

diff --git a/content/docs/guide/architecture.zh.md b/content/docs/guide/architecture.zh.md
@@ -18,6 +18,8 @@ Sirchmunk 采用清晰分离的层次化架构，遵循**关注点分离**原则
 | **EvidenceProcessor**   | 基于蒙特卡洛重要性采样的证据处理                                       |
 | **GrepRetriever**       | 高性能 _无索引_ 文件检索，支持并行处理                                 |
 | **OpenAIChat**          | 统一 LLM 接口，支持流式与用量统计                                       |
+| **KnowledgeCompiler**   | 离线文档编译为树索引和知识簇（Beta）                                   |
+| **KnowledgeLint**       | 知识健康检查与自动修复                                                 |
 | **MonitorTracker**      | 实时系统与应用指标采集                                                 |
 
 ## 多阶段搜索管线
@@ -166,6 +168,13 @@ KnowledgeCluster 是一个丰富标注的对象，完整记录了单次搜索周
     │   └── chat_history.db
     ├── knowledge/            # 知识簇（Parquet）
     │   └── knowledge_clusters.parquet
+    ├── compile/              # 编译产物（Beta）
+    │   ├── manifest.json     # 文件清单与哈希
+    │   ├── document_catalog.json
+    │   ├── summary_index.json
+    │   ├── trees/            # 层次化树索引
+    │   ├── table_digests/    # 表格提取摘要
+    │   └── xlsx_digests/     # 电子表格摘要
     └── settings/             # 用户设置（DuckDB）
         └── settings.db
 ```

diff --git a/content/docs/guide/cli.md b/content/docs/guide/cli.md
@@ -58,7 +58,7 @@ sirchmunk search "How does authentication work?"
 # Search in specific paths
 sirchmunk search "find all API endpoints" ./src ./docs
 
-# DEEP mode: comprehensive analysis with Monte Carlo sampling
+# DEEP mode: comprehensive agentic retrieval analysis
 sirchmunk search "database architecture" --mode DEEP
 
 # Quick filename search (no LLM required)
@@ -103,6 +103,45 @@ sirchmunk mcp serve
 sirchmunk mcp serve --transport http --port 3000
 ```
 
+### `sirchmunk compile` (Beta)
+
+Pre-process document collections into hierarchical tree indices and knowledge clusters. This is an **optional** step — search works without it, but compile artifacts can significantly boost retrieval precision for large document sets.
+
+```bash
+# Compile documents (incremental by default)
+sirchmunk compile --paths /path/to/documents
+
+# Full recompile (ignore cache)
+sirchmunk compile --paths /path/to/documents --full
+
+# Shallow mode (skip tree indexing, faster)
+sirchmunk compile --paths /path/to/documents --shallow
+
+# Check compile status
+sirchmunk compile --paths /path/to/documents --status
+
+# Run knowledge health checks
+sirchmunk compile --lint --work-path ~/.sirchmunk
+
+# Auto-fix lint issues
+sirchmunk compile --lint --fix --work-path ~/.sirchmunk
+```
+
+| Option | Description |
+|--------|-------------|
+| `--paths` | Directories or files to compile (required) |
+| `--full` | Force full recompile, ignoring incremental cache |
+| `--shallow` | Skip tree indexing, use direct LLM summarization only (faster) |
+| `--max-files` | Max files to process (triggers importance sampling for large sets) |
+| `--concurrency` | Max parallel file compilations (default: 3) |
+| `--status` | Show compile status instead of running compile |
+| `--lint` | Run knowledge health checks |
+| `--fix` | Auto-fix lint issues (use with `--lint`) |
+| `--work-path` | Working directory (default: `~/.sirchmunk`) |
+
+> [!NOTE]
+> Compile artifacts are automatically detected by the search pipeline — no additional configuration is needed after compilation. When no compile artifacts exist, search falls back to the standard retrieval pipeline.
+
 ### `sirchmunk version`
 
 Display version information.
@@ -119,6 +158,7 @@ sirchmunk mcp version
 | `sirchmunk init` | Initialize working directory, .env, and MCP config |
 | `sirchmunk serve` | Start the backend API server |
 | `sirchmunk search` | Perform search queries |
+| `sirchmunk compile` | Compile documents into knowledge indices **(Beta)** |
 | `sirchmunk web init` | Build WebUI frontend (requires Node.js 18+) |
 | `sirchmunk web serve` | Start API + WebUI (single port) |
 | `sirchmunk web serve --dev` | Start API + Next.js dev server (hot-reload) |

diff --git a/content/docs/guide/cli.zh.md b/content/docs/guide/cli.zh.md
@@ -58,7 +58,7 @@ sirchmunk search "How does authentication work?"
 # 在指定路径中搜索
 sirchmunk search "find all API endpoints" ./src ./docs
 
-# DEEP 模式：蒙特卡洛证据采样全面分析
+# DEEP 模式：智能体检索全面分析
 sirchmunk search "数据库架构" --mode DEEP
 
 # 快速文件名搜索（无需 LLM）
@@ -103,6 +103,45 @@ sirchmunk mcp serve
 sirchmunk mcp serve --transport http --port 3000
 ```
 
+### `sirchmunk compile`（Beta）
+
+将文档集预处理为层次化树索引和知识簇。这是一个**可选**步骤 — 无需编译即可搜索，但编译产物可显著提升大型文档集的检索精度。
+
+```bash
+# 编译文档（默认增量模式）
+sirchmunk compile --paths /path/to/documents
+
+# 全量重新编译（忽略缓存）
+sirchmunk compile --paths /path/to/documents --full
+
+# 浅层模式（跳过树索引，更快）
+sirchmunk compile --paths /path/to/documents --shallow
+
+# 查看编译状态
+sirchmunk compile --paths /path/to/documents --status
+
+# 运行知识健康检查
+sirchmunk compile --lint --work-path ~/.sirchmunk
+
+# 自动修复检查问题
+sirchmunk compile --lint --fix --work-path ~/.sirchmunk
+```
+
+| 选项 | 描述 |
+|------|------|
+| `--paths` | 要编译的目录或文件（必填） |
+| `--full` | 强制全量重编译，忽略增量缓存 |
+| `--shallow` | 跳过树索引，仅使用 LLM 直接摘要（更快） |
+| `--max-files` | 最大处理文件数（超出时触发重要性采样） |
+| `--concurrency` | 最大并行编译数（默认：3） |
+| `--status` | 显示编译状态而非执行编译 |
+| `--lint` | 运行知识健康检查 |
+| `--fix` | 自动修复检查问题（需配合 `--lint`） |
+| `--work-path` | 工作目录（默认：`~/.sirchmunk`） |
+
+> [!NOTE]
+> 编译产物会被搜索管线自动检测 — 编译完成后无需额外配置。当不存在编译产物时，搜索会回退到标准检索管线。
+
 ### `sirchmunk version`
 
 显示版本信息。
@@ -119,6 +158,7 @@ sirchmunk mcp version
 | `sirchmunk init` | 初始化工作目录、.env 和 MCP 配置 |
 | `sirchmunk serve` | 启动后端 API 服务器 |
 | `sirchmunk search` | 执行搜索查询 |
+| `sirchmunk compile` | 将文档编译为知识索引 **（Beta）** |
 | `sirchmunk web init` | 构建 WebUI 前端（需要 Node.js 18+） |
 | `sirchmunk web serve` | 启动 API + WebUI（单端口） |
 | `sirchmunk web serve --dev` | 启动 API + Next.js 开发服务器（热重载） |

diff --git a/content/docs/guide/configuration.md b/content/docs/guide/configuration.md
@@ -53,6 +53,13 @@ All persistent data is stored under `SIRCHMUNK_WORK_PATH`:
   │   │   └── chat_history.db
   │   ├── knowledge/            # Knowledge clusters (Parquet)
   │   │   └── knowledge_clusters.parquet
+  │   ├── compile/              # Compile artifacts (Beta)
+  │   │   ├── manifest.json     # File manifest with hashes
+  │   │   ├── document_catalog.json
+  │   │   ├── summary_index.json
+  │   │   ├── trees/            # Hierarchical tree indices
+  │   │   ├── table_digests/    # Table extraction digests
+  │   │   └── xlsx_digests/     # Spreadsheet digests
   │   └── settings/             # User settings (DuckDB)
   │       └── settings.db
   ├── .env                      # Environment configuration
@@ -67,7 +74,7 @@ When invoking search (via SDK, CLI, or API), the following parameters are availa
 |-----------|------|---------|-------------|
 | `query` | `string` | *required* | Search query or question |
 | `paths` | `string \| string[]` | *optional* | Directories or files to search; falls back to `SIRCHMUNK_SEARCH_PATHS`, then cwd |
-| `mode` | `string` | `FAST` | `FAST` (greedy, 2-5s), `DEEP` (Monte Carlo, 10-30s), or `FILENAME_ONLY` |
+| `mode` | `string` | `FAST` | `FAST` (greedy, 2-5s), `DEEP` (agentic retrieval, 10-30s), or `FILENAME_ONLY` |
 | `max_depth` | `int` | `null` | Maximum directory depth |
 | `top_k_files` | `int` | `null` | Number of top files to return |
 | `enable_dir_scan` | `bool` | `true` | Enable directory scanning |

diff --git a/content/docs/guide/configuration.zh.md b/content/docs/guide/configuration.zh.md
@@ -53,6 +53,13 @@ Sirchmunk 通过存储在 `.env` 文件中的环境变量进行配置。运行 `
   │   │   └── chat_history.db
   │   ├── knowledge/            # 知识簇（Parquet）
   │   │   └── knowledge_clusters.parquet
+  │   ├── compile/              # 编译产物（Beta）
+  │   │   ├── manifest.json     # 文件清单与哈希
+  │   │   ├── document_catalog.json
+  │   │   ├── summary_index.json
+  │   │   ├── trees/            # 层次化树索引
+  │   │   ├── table_digests/    # 表格提取摘要
+  │   │   └── xlsx_digests/     # 电子表格摘要
   │   └── settings/             # 用户设置（DuckDB）
   │       └── settings.db
   ├── .env                      # 环境配置
@@ -67,7 +74,7 @@ Sirchmunk 通过存储在 `.env` 文件中的环境变量进行配置。运行 `
 |------|------|--------|------|
 | `query` | `string` | *必填* | 搜索查询或问题 |
 | `paths` | `string \| string[]` | *可选* | 要搜索的目录或文件；未设置时依次回退到 `SIRCHMUNK_SEARCH_PATHS`、当前工作目录 |
-| `mode` | `string` | `FAST` | `FAST`（贪心搜索，2-5s）、`DEEP`（蒙特卡洛采样，10-30s）或 `FILENAME_ONLY` |
+| `mode` | `string` | `FAST` | `FAST`（贪心搜索，2-5s）、`DEEP`（智能体检索，10-30s）或 `FILENAME_ONLY` |
 | `max_depth` | `int` | `null` | 最大目录深度 |
 | `top_k_files` | `int` | `null` | 返回的文件数量 |
 | `enable_dir_scan` | `bool` | `true` | 是否启用目录扫描 |

diff --git a/content/docs/guide/docker.md b/content/docs/guide/docker.md
@@ -9,14 +9,14 @@ Pre-built Docker images are available on Alibaba Cloud Container Registry for qu
 
 | Region | Image |
 |---|---|
-| US West | `modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6` |
-| China Beijing | `modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6` |
+| US West | `modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7` |
+| China Beijing | `modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7` |
 
 ## Quick Start
 
 ```bash
 # Pull the image
-docker pull modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6
+docker pull modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7
 
 # Start the service
 docker run -d \
@@ -35,7 +35,7 @@ docker run -d \
   -e SIRCHMUNK_SEARCH_PATHS=/mnt/docs \
   -v /path/to/your_work_path:/data/sirchmunk \
   -v /path/to/your/docs:/mnt/docs:ro \
-  modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6
+  modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7
 ```
 
 Open http://localhost:8584 to access the WebUI, or call the API directly:

diff --git a/content/docs/guide/docker.zh.md b/content/docs/guide/docker.zh.md
@@ -9,14 +9,14 @@ weight: 8
 
 | 区域 | 镜像 |
 |---|---|
-| 美西 | `modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6` |
-| 北京 | `modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6` |
+| 美西 | `modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7` |
+| 北京 | `modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7` |
 
 ## 快速开始
 
 ```bash
 # 拉取镜像（根据地理位置选择最近的 Registry）
-docker pull modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6
+docker pull modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7
 
 # 启动服务
 docker run -d \
@@ -35,7 +35,7 @@ docker run -d \
   -e SIRCHMUNK_SEARCH_PATHS=/mnt/docs \
   -v /path/to/your_work_path:/data/sirchmunk \
   -v /path/to/your/docs:/mnt/docs:ro \
-  modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.6
+  modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/sirchmunk:ubuntu22.04-py312-0.0.7
 ```
 
 打开 http://localhost:8584 访问 WebUI，或直接调用 API：

diff --git a/content/docs/guide/mcp.md b/content/docs/guide/mcp.md
@@ -89,7 +89,7 @@ List all stored knowledge clusters.
 | Mode | Description | LLM Required |
 |------|-------------|:------------:|
 | **FAST** | Greedy search with 2-level keyword cascade and early stopping (2-5s, ~10x faster than DEEP) | Yes |
-| **DEEP** | Full multi-phase analysis with Monte Carlo evidence sampling (10-30s) | Yes |
+| **DEEP** | Full multi-phase agentic retrieval analysis with self-correction (10-30s) | Yes |
 | **FILENAME_ONLY** | Filename-based search without content analysis | No |
 
 ## OpenClaw Integration

diff --git a/content/docs/guide/mcp.zh.md b/content/docs/guide/mcp.zh.md
@@ -89,7 +89,7 @@ sirchmunk mcp serve --transport http --port 3000
 | 模式 | 描述 | 需要 LLM |
 |------|------|:--------:|
 | **FAST** | 贪心搜索，两级关键词级联 + early stopping（2-5s，速度约为 DEEP 的 10 倍） | 是 |
-| **DEEP** | 蒙特卡洛证据采样完整多阶段分析（10-30s） | 是 |
+| **DEEP** | 智能体检索全面分析，支持自纠正（10-30s） | 是 |
 | **FILENAME_ONLY** | 文件名搜索，无内容分析 | 否 |
 
 ## OpenClaw 集成

diff --git a/content/docs/guide/project-structure.md b/content/docs/guide/project-structure.md
@@ -16,7 +16,7 @@ sirchmunk/
 │   │   │   └── components/     # History, monitor, settings storage
 │   │   ├── cli/                # CLI entry point and web launcher
 │   │   ├── insight/            # Text insight extraction
-│   │   ├── learnings/          # Evidence processing, knowledge base
+│   │   ├── learnings/          # Evidence processing, knowledge base, compiler
 │   │   ├── llm/                # LLM interface (OpenAI-compatible)
 │   │   ├── retrieve/           # Indexless retrieval engine
 │   │   ├── scan/               # Directory and file scanners
@@ -57,9 +57,11 @@ The search pipeline coordinator:
 - **SearchContext** (`schema/search_context.py`) — Budget, state, and audit management
 
 ### Intelligence Layer
-Evidence extraction and knowledge synthesis:
+Evidence extraction, knowledge synthesis, and offline compilation:
 - **EvidenceProcessor** (`learnings/evidence_processor.py`) — Monte Carlo sampling
 - **KnowledgeBase** (`learnings/knowledge_base.py`) — Knowledge cluster management
+- **KnowledgeCompiler** (`learnings/compiler.py`) — Offline document compilation into tree indices and knowledge clusters (Beta)
+- **KnowledgeLint** (`learnings/lint.py`) — Knowledge health checks and auto-repair
 - **ReActAgent** (`agentic/react_agent.py`) — Autonomous exploration
 - **OpenAIChat** (`llm/openai_chat.py`) — Unified LLM interface
 
@@ -79,4 +81,6 @@ Persistence and caching:
 | **DirScanner** | `scan/dir_scanner.py` | Directory structure analysis |
 | **ReActAgent** | `agentic/react_agent.py` | Budget-bounded autonomous exploration |
 | **OpenAIChat** | `llm/openai_chat.py` | Unified LLM interface with streaming and usage tracking |
+| **KnowledgeCompiler** | `learnings/compiler.py` | Offline document compilation into tree indices (Beta) |
+| **KnowledgeLint** | `learnings/lint.py` | Knowledge health checks and auto-repair |
 | **MonitorTracker** | `api/components/monitor_tracker.py` | Real-time system metrics |
diff --git a/content/docs/guide/project-structure.zh.md b/content/docs/guide/project-structure.zh.md
@@ -16,7 +16,7 @@ sirchmunk/
 │   │   │   └── components/     # 历史、监控、设置存储
 │   │   ├── cli/                # CLI 入口和 Web 启动器
 │   │   ├── insight/            # 文本洞察提取
-│   │   ├── learnings/          # 证据处理、知识库
+│   │   ├── learnings/          # 证据处理、知识库、编译器
 │   │   ├── llm/                # LLM 接口（OpenAI 兼容）
 │   │   ├── retrieve/           # 无索引检索引擎
 │   │   ├── scan/               # 目录和文件扫描器
@@ -57,9 +57,11 @@ Sirchmunk 遵循严格的**关注点分离**模式，分为四个独立层次：
 - **SearchContext** (`schema/search_context.py`) — 预算、状态和审计管理
 
 ### 智能层
-证据提取和知识合成：
+证据提取、知识合成与离线编译：
 - **EvidenceProcessor** (`learnings/evidence_processor.py`) — 蒙特卡洛采样
 - **KnowledgeBase** (`learnings/knowledge_base.py`) — 知识簇管理
+- **KnowledgeCompiler** (`learnings/compiler.py`) — 离线文档编译为树索引和知识簇（Beta）
+- **KnowledgeLint** (`learnings/lint.py`) — 知识健康检查与自动修复
 - **ReActAgent** (`agentic/react_agent.py`) — 自主探索
 - **OpenAIChat** (`llm/openai_chat.py`) — 统一 LLM 接口
 
@@ -79,4 +81,6 @@ Sirchmunk 遵循严格的**关注点分离**模式，分为四个独立层次：
 | **DirScanner** | `scan/dir_scanner.py` | 目录结构分析 |
 | **ReActAgent** | `agentic/react_agent.py` | 预算约束下的自主探索 |
 | **OpenAIChat** | `llm/openai_chat.py` | 统一 LLM 接口（支持流式和使用量跟踪） |
+| **KnowledgeCompiler** | `learnings/compiler.py` | 离线文档编译为树索引（Beta） |
+| **KnowledgeLint** | `learnings/lint.py` | 知识健康检查与自动修复 |
 | **MonitorTracker** | `api/components/monitor_tracker.py` | 实时系统指标 |