update CHANGELOG.md

Author: iamwjun

CHANGELOG.md

@@ -1,6 +1,14 @@
 # Changelog
 
-<!-- last-commit: f7db5e7 -->
+<!-- last-commit: 5c6e6ad -->
+
+## v0.1.11 - 2026-05-08
+
+| Commit | Description |
+|--------|-------------|
+| `5c6e6ad` | feat: support /server |
+| `acf74b8` | update CHANGELOG.md |
+
 
 ## v0.1.10 - 2026-04-30
 

Cargo.lock

@@ -1,6 +1,6 @@
 [package]
 name = "localcoder"
-version = "0.1.10"
+version = "0.1.11"
 edition = "2024"
 description = "oxink."
 license = "MIT"

Cargo.toml

@@ -11,6 +11,7 @@ Localcoder is a Claude-like command-line AI assistant implemented in Rust. The c
 - ✅ Ollama-backed chat with streaming responses and one-shot mode
 - ✅ Tool calling runtime with file, search, Bash, web, and LSP tools
 - ✅ Interactive REPL with model switching, session resume, config UI, and output styles
+- ✅ Local server mode with HTTP and WebSocket entrypoints
 - ✅ Context compaction, git workflows, memory extraction, plan mode, and skills
 - ✅ Lightweight runtime with fast startup and low memory usage
 
@@ -20,7 +21,7 @@ Localcoder is a Claude-like command-line AI assistant implemented in Rust. The c
 
 ## 📊 Implementation Status
 
-The staged roadmap in [`docs/P00-plan.md`](./docs/P00-plan.md) is mostly implemented. Current status: **15 / 20 stages completed**.
+The staged roadmap in [`docs/P00-plan.md`](./docs/P00-plan.md) is mostly implemented. Current status: **16 / 21 stages completed**.
 
 | Stage | Area | Status | Deliverable |
 |------|------|------|------|
@@ -44,6 +45,7 @@ The staged roadmap in [`docs/P00-plan.md`](./docs/P00-plan.md) is mostly impleme
 | S17 | MCP integration | ❌ | MCP client and transport support are not implemented yet |
 | S18 | Output styles | ✅ | Output style loading and `/output-style` |
 | S19 | LSP integration | ✅ | Language-server-backed code navigation via `Lsp` |
+| S20 | Server mode | ✅ | Axum-based local HTTP and WebSocket server via `/server` |
 
 ---
 
@@ -93,7 +95,7 @@ On startup, Localcoder automatically checks for a settings file:
 
 - It first looks for `.localcoder/settings.json` in the current directory
 - If that file does not exist, it falls back to `$HOME/.localcoder/settings.json`
-- If neither exists, it creates a default config in the current directory
+- If neither exists, it creates a default config in `$HOME/.localcoder/settings.json`
 
 The default config format is:
 
@@ -150,8 +152,77 @@ localcoder --continue
 
 # Resume a specific session
 localcoder --resume s1712345678-12345
+
+# Start the local server in the foreground
+localcoder -- "/server"
+
+# Start the local server on a custom address
+localcoder -- "/server 127.0.0.1:4000"
+```
+
+---
+
+## 🌐 Server Mode
+
+Localcoder can also run as a local HTTP and WebSocket server. The default bind address is `127.0.0.1:3000`.
+
+You can start it in either mode:
+
+```bash
+# Start in the REPL, but keep the REPL usable
+/server
+/server status
+/server stop
+
+# Start in one-shot mode and keep the process in the foreground
+localcoder -- "/server"
+localcoder -- "/server 127.0.0.1:4000"
+```
+
+Available routes:
+
+- `GET /healthz`
+- `POST /v1/message`
+- `GET /v1/ws`
+
+Example HTTP request:
+
+```bash
+curl -X POST http://127.0.0.1:3000/v1/message \
+  -H "content-type: application/json" \
+  -d '{
+    "message": "Explain the role of src/main.rs",
+    "session_id": "",
+    "output_style": "default"
+  }'
+```
+
+Example response:
+
+```json
+{
+  "session_id": "s1746690000000-12345-0",
+  "reply": "src/main.rs bootstraps configuration, registers tools, and decides between REPL and one-shot execution.",
+  "model": "qwen3.5:4b"
+}
 ```
 
+WebSocket messages are JSON-based and currently one request maps to one full agent execution:
+
+```json
+{
+  "type": "message",
+  "message": "Continue the previous turn and summarize main.rs",
+  "session_id": "s1746690000000-12345-0"
+}
+```
+
+The server is intentionally local-first:
+
+- It listens on `127.0.0.1` by default
+- There is no built-in auth or TLS yet
+- `wss` should be handled by a reverse proxy if needed
+
 ---
 
 ## 🛠️ Built-in Tools
@@ -189,6 +260,7 @@ localcoder -- "Fetch https://www.rust-lang.org/"
 | `/output-style [name]` | List or switch output styles |
 | `/web <query>` | Search the public web directly |
 | `/fetch <url>` | Fetch a public web page |
+| `/server [status\|stop\|host:port]` | Start, stop, or inspect the local HTTP/WebSocket server |
 | `/plan` | Show plan-mode status |
 | `/plan on` | Enable plan mode manually |
 | `/plan off` | Disable plan mode manually |
@@ -218,7 +290,7 @@ localcoder/
 ├── README.zh.md         # Chinese documentation
 ├── docs/                # Roadmap and stage-by-stage implementation notes
 │   ├── P00-plan.md      # Overall staged plan
-│   └── S00-S19*.md      # Detailed stage documents
+│   └── S00-S20*.md      # Detailed stage documents
 ├── examples/            # Example programs
 │   ├── basic.rs          # Basic API usage
 │   ├── streaming.rs      # Streaming responses
@@ -236,6 +308,8 @@ localcoder/
     ├── output_style.rs   # Output style loading and prompt injection
     ├── plan.rs           # Plan mode state and todo management
     ├── repl.rs           # Interactive REPL interface
+    ├── runtime.rs        # Shared runtime/bootstrap helpers
+    ├── server.rs         # Local HTTP and WebSocket server mode
     ├── session.rs        # JSONL session persistence
     ├── skills.rs         # SKILL.md loading and activation
     ├── tools/            # Built-in tools
@@ -251,6 +325,7 @@ localcoder/
 |------|----------|
 | Async runtime | tokio 1.40 |
 | HTTP client | reqwest 0.12 |
+| Local server | axum 0.8 |
 | JSON handling | serde + serde_json 1.0 |
 | Line editing | rustyline 14.0 |
 | Error handling | anyhow |

README.md

@@ -11,6 +11,7 @@ Localcoder 是一个基于 Rust 实现的 Claude-like 命令行 AI 助手，当
 - ✅ 基于 Ollama 的对话、流式响应和单次查询模式
 - ✅ 文件、搜索、Bash、Web、LSP 等工具调用运行时
 - ✅ 带模型切换、会话恢复、配置菜单和输出风格的交互式 REPL
+- ✅ 本地 Server 模式，支持 HTTP 和 WebSocket 入口
 - ✅ 上下文压缩、Git 工作流、记忆提取、计划模式和技能系统
 - ✅ 轻量级（启动快、内存占用低）
 
@@ -20,7 +21,7 @@ Localcoder 是一个基于 Rust 实现的 Claude-like 命令行 AI 助手，当
 
 ## 📊 实现状态
 
-[`docs/P00-plan.md`](./docs/P00-plan.md) 中的阶段路线图大部分已经落地。当前进度：**20 个阶段中已完成 15 个**。
+[`docs/P00-plan.md`](./docs/P00-plan.md) 中的阶段路线图大部分已经落地。当前进度：**21 个阶段中已完成 16 个**。
 
 | 阶段 | 模块 | 状态 | 核心交付物 |
 |------|------|------|------|
@@ -44,6 +45,7 @@ Localcoder 是一个基于 Rust 实现的 Claude-like 命令行 AI 助手，当
 | S17 | MCP 集成 | ❌ | MCP 客户端和多传输支持尚未实现 |
 | S18 | 输出样式 | ✅ | 输出样式加载和 `/output-style` |
 | S19 | LSP 集成 | ✅ | 基于语言服务器的代码导航 `Lsp` |
+| S20 | Server 模式 | ✅ | 基于 Axum 的本地 HTTP / WebSocket 服务与 `/server` |
 
 ---
 
@@ -93,19 +95,45 @@ localcoder
 
 - 优先读取当前目录的 `.localcoder/settings.json`
 - 如果当前目录没有，则读取 `$HOME/.localcoder/settings.json`
-- 如果两处都没有，则在当前目录自动创建默认配置
+- 如果两处都没有，则在 `$HOME/.localcoder/settings.json` 自动创建默认配置
 
 默认配置格式如下：
 
 ```json
 {
-  "ollama": {
-    "url": "http://localhost:11434",
+  "llm": {
+    "type": "ollama",
+    "base_url": "http://localhost:11434",
     "model": "qwen3.5:4b"
   }
 }
 ```
 
+`llm.type` 用于选择提供商：`ollama`、`lmstudio`、`openai`。
+
+```json
+{
+  "llm": {
+    "type": "lmstudio",
+    "base_url": "http://localhost:1234",
+    "model": "qwen/qwen3-coder-30b"
+  }
+}
+```
+
+对于 OpenAI 兼容服务：
+
+```json
+{
+  "llm": {
+    "type": "openai",
+    "base_url": "https://api.openai.com/v1",
+    "api_key": "sk-...",
+    "model": "gpt-4o-mini"
+  }
+}
+```
+
 你也可以手动编辑这个文件，或在 REPL 中使用 `/model` 指令切换模型。
 
 ---
@@ -124,8 +152,77 @@ localcoder --continue
 
 # 恢复指定会话
 localcoder --resume s1712345678-12345
+
+# 前台启动本地服务
+localcoder -- "/server"
+
+# 使用自定义地址启动本地服务
+localcoder -- "/server 127.0.0.1:4000"
+```
+
+---
+
+## 🌐 Server 模式
+
+Localcoder 也可以作为本地 HTTP / WebSocket 服务运行。默认监听地址为 `127.0.0.1:3000`。
+
+可以通过两种方式启动：
+
+```bash
+# 在 REPL 中后台启动，同时继续使用 REPL
+/server
+/server status
+/server stop
+
+# 在 one-shot 模式下前台运行
+localcoder -- "/server"
+localcoder -- "/server 127.0.0.1:4000"
+```
+
+当前可用路由：
+
+- `GET /healthz`
+- `POST /v1/message`
+- `GET /v1/ws`
+
+HTTP 请求示例：
+
+```bash
+curl -X POST http://127.0.0.1:3000/v1/message \
+  -H "content-type: application/json" \
+  -d '{
+    "message": "解释一下 src/main.rs 的职责",
+    "session_id": "",
+    "output_style": "default"
+  }'
+```
+
+响应示例：
+
+```json
+{
+  "session_id": "s1746690000000-12345-0",
+  "reply": "src/main.rs 负责初始化配置、注册工具，并决定进入 REPL 还是 one-shot 执行流程。",
+  "model": "qwen3.5:4b"
+}
 ```
 
+WebSocket 消息同样使用 JSON；当前一条请求对应一次完整 agent 执行：
+
+```json
+{
+  "type": "message",
+  "message": "继续上一轮，并总结一下 main.rs",
+  "session_id": "s1746690000000-12345-0"
+}
+```
+
+当前服务模式的定位是本地优先：
+
+- 默认只监听 `127.0.0.1`
+- 暂时没有内建认证和 TLS
+- 如果需要 `wss`，建议通过反向代理处理
+
 ---
 
 ## 🛠️ 内置工具
@@ -163,6 +260,7 @@ localcoder -- "抓取 https://www.rust-lang.org/"
 | `/output-style [name]` | 列出或切换输出风格 |
 | `/web <query>` | 直接搜索公网内容 |
 | `/fetch <url>` | 抓取公开网页 |
+| `/server [status\|stop\|host:port]` | 启动、停止或查看本地 HTTP / WebSocket 服务 |
 | `/plan` | 查看计划模式状态 |
 | `/plan on` | 手动启用计划模式 |
 | `/plan off` | 手动关闭计划模式 |
@@ -192,7 +290,7 @@ localcoder/
 ├── README.zh.md         # 中文说明
 ├── docs/                # 路线图与分阶段实现文档
 │   ├── P00-plan.md      # 总体阶段计划
-│   └── S00-S19*.md      # 各阶段详细说明
+│   └── S00-S20*.md      # 各阶段详细说明
 ├── examples/            # 示例代码
 │   ├── basic.rs          # 基本 API 调用
 │   ├── streaming.rs      # 流式响应
@@ -210,6 +308,8 @@ localcoder/
     ├── output_style.rs   # 输出风格加载与 prompt 注入
     ├── plan.rs           # 计划模式状态与 todo 管理
     ├── repl.rs           # 交互式 REPL
+    ├── runtime.rs        # 共享运行时与启动辅助
+    ├── server.rs         # 本地 HTTP / WebSocket 服务模式
     ├── session.rs        # JSONL 会话持久化
     ├── skills.rs         # SKILL.md 加载与激活
     ├── tools/            # 内置工具
@@ -225,10 +325,11 @@ localcoder/
 |------|----------|
 | 异步运行时 | tokio 1.40 |
 | HTTP 客户端 | reqwest 0.12 |
+| 本地服务 | axum 0.8 |
 | JSON 处理 | serde + serde_json 1.0 |
 | 命令行编辑 | rustyline 14.0 |
 | 错误处理 | anyhow |
-| 终端彩色 | colored 2.1 |
+| 终端彩色 | oxink 0.1.1 |
 
 ---