@arrislink/axon 1.0.2 → 1.0.4

package/README.md

@@ -195,29 +195,37 @@ safety:
 
 ## 🏗️ Architecture
 
+Axon is built on a modular architecture that separates requirement definition, task planning, and agentic execution.
+
+```mermaid
+graph TD
+    User([User CLI]) --> Commands[Axon Commands: init, spec, plan, work]
+    
+    subgraph "Axon Engine"
+        Commands --> Spec[OpenSpec Manager]
+        Spec --> Beads[Beads Engine: Graph Gen & Execution]
+        Beads --> Skills[Skills Library: Matching & Injection]
+        Skills --> Orch[Agent Orchestrator]
+    end
+    
+    subgraph "LLM Layer"
+        Orch --> LLMInt[Unified LLM Interface]
+        LLMInt --> OMO[OhMyOpenCode Registry]
+        OMO --> Providers[Providers: Anthropic, OpenAI, Antigrav, etc.]
+    end
 ```
-┌─────────────────────────────────────────────────────┐
-│              Axon (Orchestration Layer)              │
-│  ┌──────────┬──────────┬──────────┬──────────┐     │
-│  │  init    │  spec    │  plan    │  work    │     │
-│  └──────────┴──────────┴──────────┴──────────┘     │
-└─────────────────────────────────────────────────────┘
-                        ↓
-┌─────────────────────────────────────────────────────┐
-│         Unified LLM Interface (LLMClient)            │
-│  Auto-detects and uses OMO Provider System           │
-└─────────────────────────────────────────────────────┘
-                        ↓
-┌─────────────────────────────────────────────────────┐
-│      OhMyOpenCode Provider System (Optional)         │
-│  ┌──────────┬──────────┬──────────┬──────────┐     │
-│  │Anthropic │Antigrav  │  OpenAI  │  Google  │     │
-│  └──────────┴──────────┴──────────┴──────────┘     │
-│  • Unified config (~/.omo/providers.yaml)            │
-│  • Auto failover                                     │
-│  • Cost optimization                                 │
-└─────────────────────────────────────────────────────┘
-```
+
+### Core Components:
+
+1.  **OpenSpec Manager**: Parses and manages the specification (`.openspec/spec.md`). It ensures the AI always has a "Source of Truth" for requirements.
+2.  **Beads Engine**: 
+    *   **Planning**: Converts specifications into a Directed Acyclic Graph (DAG) of atomic tasks.
+    *   **Execution**: Manages task dependencies, state persistence (`graph.json`), and sequential execution.
+3.  **Skills Library**: A repository of reusable code patterns. It automatically matches relevant skills to tasks and injects them into the agent's context.
+4.  **Agent Orchestrator**: Coordinates specialized AI agents (like the general-purpose "Sisyphus") to execute specific beads.
+5.  **Unified LLM Interface**: A vendor-neutral abstraction layer that integrates with **OhMyOpenCode** to provide multi-provider failover and consistent API access.
+
+---
 
 ## 🧪 Development
 

package/README.zh-CN.md

@@ -193,31 +193,39 @@ safety:
   auto_pause_on_error: true
 ```
 
-## 🏗️ 架构
+## 🏗️ 架构设计
 
+Axon 基于模块化架构构建，将需求定义、任务规划和代理执行清晰分离。
+
+```mermaid
+graph TD
+    User([用户 CLI]) --> Commands[Axon 命令: init, spec, plan, work]
+    
+    subgraph "Axon 核心引擎"
+        Commands --> Spec[OpenSpec 管理器]
+        Spec --> Beads[Beads 引擎: 图生成与执行]
+        Beads --> Skills[技能库: 匹配与注入]
+        Skills --> Orch[任务编排器]
+    end
+    
+    subgraph "LLM 层"
+        Orch --> LLMInt[统一 LLM 接口]
+        LLMInt --> OMO[OhMyOpenCode 注册表]
+        OMO --> Providers[平台: Anthropic, OpenAI, Antigrav 等]
+    end
 ```
-┌─────────────────────────────────────────────────────┐
-│              Axon (编排层)                           │
-│  ┌──────────┬──────────┬──────────┬──────────┐     │
-│  │  init    │  spec    │  plan    │  work    │     │
-│  └──────────┴──────────┴──────────┴──────────┘     │
-└─────────────────────────────────────────────────────┘
-                        ↓
-┌─────────────────────────────────────────────────────┐
-│         统一 LLM 接口 (LLMClient)                    │
-│  自动检测并使用 OMO Provider 系统                    │
-└─────────────────────────────────────────────────────┘
-                        ↓
-┌─────────────────────────────────────────────────────┐
-│      OhMyOpenCode Provider 系统 (可选)               │
-│  ┌──────────┬──────────┬──────────┬──────────┐     │
-│  │Anthropic │Antigrav  │  OpenAI  │  Google  │     │
-│  └──────────┴──────────┴──────────┴──────────┘     │
-│  • 统一配置 (~/.omo/providers.yaml)                 │
-│  • 自动 failover                                    │
-│  • 成本优化                                         │
-└─────────────────────────────────────────────────────┘
-```
+
+### 核心组件说明：
+
+1.  **OpenSpec 管理器**: 解析并管理规格说明 (`.openspec/spec.md`)，确保 AI 始终拥有需求的“单一真理来源”。
+2.  **Beads 引擎**: 
+    *   **规划 (Planning)**: 将规格说明转换为由原子任务组成的有向无环图 (DAG)。
+    *   **执行 (Execution)**: 管理任务依赖、状态持久化 (`graph.json`) 以及顺序执行流程。
+3.  **技能库 (Skills Library)**: 可复用代码模式的仓库。系统会自动匹配相关技能并将其注入到 AI 智能体的上下文中。
+4.  **任务编排器 (Agent Orchestrator)**: 协调专门的 AI 代理（如通用型的 "Sisyphus"）来执行具体的任务珠子。
+5.  **统一 LLM 接口**: 厂商中立的抽象层。它集成了 **OhMyOpenCode**，提供多供应商故障转移和一致的 API 访问。
+
+---
 
 ## 🧪 开发
 

package/dist/index.js

@@ -16471,10 +16471,19 @@ class OpenCodeLLMClient {
             continue;
           try {
             const event = JSON.parse(line);
+            if (event.type === "config" && event.part?.model) {
+              metadata.model = event.part.model;
+            } else if (event.model && metadata.model === "unknown") {
+              metadata.model = event.model;
+            }
             if (event.type === "text" && event.part?.text) {
               const text = event.part.text;
               fullResponse += text;
               yield text;
+            } else if (event.type === "content" && event.part?.content) {
+              const text = event.part.content;
+              fullResponse += text;
+              yield text;
             } else if (event.type === "step_finish") {
               if (event.part?.snapshot) {}
               if (event.part?.tokens) {
@@ -16888,11 +16897,20 @@ class SpecGenerator {
 7. \u67B6\u6784\u51B3\u7B56\u8BB0\u5F55 (ADR)
 
 \u786E\u4FDD\u6587\u6863\u6E05\u6670\u3001\u53EF\u6267\u884C\uFF0C\u4FBF\u4E8E\u540E\u7EED\u4EFB\u52A1\u62C6\u89E3\u3002`;
-    const response = await this.llm.chat([{ role: "user", content: prompt }], {
-      temperature: 0.7,
-      maxTokens: 4000
-    });
-    return response.content;
+    try {
+      const response = await this.llm.chat([{ role: "user", content: prompt }], {
+        temperature: 0.7,
+        maxTokens: 4000
+      });
+      if (response.content && response.content.trim().length > 100) {
+        return response.content;
+      }
+      console.warn("\u26A0\uFE0F AI \u751F\u6210\u5185\u5BB9\u8FC7\u77ED\u6216\u4E3A\u7A7A\uFF0C\u5C06\u4F7F\u7528\u6A21\u677F\u751F\u6210\u4F5C\u4E3A\u56DE\u9000\u3002");
+      return this.generateFromTemplate(collected);
+    } catch (error) {
+      console.warn(`\u26A0\uFE0F AI \u751F\u6210\u5931\u8D25: ${error.message}\u3002\u5C06\u4F7F\u7528\u6A21\u677F\u751F\u6210\u4F5C\u4E3A\u56DE\u9000\u3002`);
+      return this.generateFromTemplate(collected);
+    }
   }
   generateFromTemplate(collected) {
     const techStackNames = {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@arrislink/axon",
-  "version": "1.0.2",
+  "version": "1.0.4",
   "description": "AI-Powered Development Operating System with unified LLM provider support",
   "type": "module",
   "main": "dist/index.js",
@@ -12,7 +12,7 @@
     "registry": "https://registry.npmjs.org/"
   },
   "files": [
-    "dist",
+    "dist/index.js",
     "templates",
     "README.md",
     "README.zh-CN.md",
@@ -27,7 +27,7 @@
     "format": "bunx @biomejs/biome format --write .",
     "type-check": "tsc --noEmit",
     "clean": "rm -rf dist",
-    "prepublishOnly": "bun run build:js"
+    "prepublishOnly": "npm run clean && npm run build:js"
   },
   "keywords": [
     "ai",