repoburg 1.3.148 → 1.3.149

package/.playwright-mcp/page-2026-05-15T00-15-00-945Z.yml

@@ -1,3 +0,0 @@
-- generic [ref=e4]:
-  - img [ref=e5]
-  - paragraph [ref=e7]: Connecting to Repoburg backend...

package/.run/backend _ build.run.xml

@@ -1,12 +0,0 @@
-<component name="ProjectRunConfigurationManager">
-  <configuration default="false" name="backend &gt; build" type="js.build_tools.npm" nameIsGenerated="true">
-    <package-json value="$PROJECT_DIR$/backend/package.json" />
-    <command value="run" />
-    <scripts>
-      <script value="build" />
-    </scripts>
-    <node-interpreter value="project" />
-    <envs />
-    <method v="2" />
-  </configuration>
-</component>

package/.run/backend _ lint.run.xml

@@ -1,12 +0,0 @@
-<component name="ProjectRunConfigurationManager">
-  <configuration default="false" name="backend &gt; lint" type="js.build_tools.npm" nameIsGenerated="true">
-    <package-json value="$PROJECT_DIR$/backend/package.json" />
-    <command value="run" />
-    <scripts>
-      <script value="lint" />
-    </scripts>
-    <node-interpreter value="project" />
-    <envs />
-    <method v="2" />
-  </configuration>
-</component>

package/.run/backend _ test.run.xml

@@ -1,12 +0,0 @@
-<component name="ProjectRunConfigurationManager">
-  <configuration default="false" name="backend &gt; test" type="js.build_tools.npm" nameIsGenerated="true">
-    <package-json value="$PROJECT_DIR$/backend/package.json" />
-    <command value="run" />
-    <scripts>
-      <script value="test" />
-    </scripts>
-    <node-interpreter value="project" />
-    <envs />
-    <method v="2" />
-  </configuration>
-</component>

package/.run/frontend _ build.run.xml

@@ -1,12 +0,0 @@
-<component name="ProjectRunConfigurationManager">
-  <configuration default="false" name="frontend &gt; build" type="js.build_tools.npm" nameIsGenerated="true">
-    <package-json value="$PROJECT_DIR$/frontend/package.json" />
-    <command value="run" />
-    <scripts>
-      <script value="build" />
-    </scripts>
-    <node-interpreter value="project" />
-    <envs />
-    <method v="2" />
-  </configuration>
-</component>

package/.run/frontend _ lint_fix.run.xml

@@ -1,12 +0,0 @@
-<component name="ProjectRunConfigurationManager">
-  <configuration default="false" name="frontend &gt; lint:fix" type="js.build_tools.npm" nameIsGenerated="true">
-    <package-json value="$PROJECT_DIR$/frontend/package.json" />
-    <command value="run" />
-    <scripts>
-      <script value="lint:fix" />
-    </scripts>
-    <node-interpreter value="project" />
-    <envs />
-    <method v="2" />
-  </configuration>
-</component>

package/CODEMAP.md

@@ -1,53 +0,0 @@
-# Repoburg Project Codemap
-
-Repoburg is a local-first, AI-powered code generation and modification system. It bridges state-of-the-art LLMs with your local development environment through a multi-layered architecture.
-
-## 🚀 System Architecture
-
-Repoburg is composed of several specialized projects that interact to form a cohesive workflow:
-
-### 1. Core Logic & Interface
-- **[`backend/`](backend/CODEMAP.md)**: The "brain" of the system. A NestJS server that manages LLM providers (OpenRouter, Z.AI, Alibaba, Ollama, OpenAI-compatible), Eta-based context templates, and the deterministic `llm-orchestration` engine.
-- **[`frontend/`](frontend/CODEMAP.md)**: The primary dashboard. A Next.js 15 application providing real-time chat, plan review interfaces, and a built-in code editor for workspace navigation.
-
-### 2. Browser Integrations (PowerTools)
-These userscripts bridge browser-based LLM interfaces with the local backend for "manual mode" or state-sync workflows:
-- **[`userscript-chatgpt/`](userscript-chatgpt/CODEMAP.md)**: SSE-interception for **ChatGPT** to capture and sync assistant responses.
-- **[`userscript-claude/`](userscript-claude/CODEMAP.md)**: Linear stream capture for **Claude.ai**.
-
-### 3. Local Management & Auth
-- **[`daemon/`](daemon/CODEMAP.md)**: A headless Express process manager using `pm2`. It handles device-level JWT authentication and the lifecycle of multiple backend instances.
-- **[`daemon-desktop/`](daemon-desktop/CODEMAP.md)**: An Electron GUI for the daemon, allowing users to manage services and updates without the CLI.
-- **[`website/`](website/CODEMAP.md)**: The public portal (repoburg.com). Manages user accounts and the OAuth-based device authorization flow.
-
-### 4. Advanced Tooling
-- **[`visual-editor-proxy/`](visual-editor-proxy/CODEMAP.md)**: A local proxy that enables "Live View" selection. It injects a script into target apps to resolve React components to source files visually.
-
----
-
-## 🔄 Primary Workflows
-
-### The AI Orchestration Loop
-1.  **Input**: User submits a prompt via the Frontend.
-2.  **Context**: Backend renders Eta templates (fetching files, git diffs, or using MCP tools).
-3.  **LLM**: The configured LLM provider generates a plan, or the prompt is sent to a Userscript via WebSocket (Manual Mode).
-4.  **Parsing**: The `llm-orchestration` module parses tool calls from the LLM response.
-5.  **Execution**: Actions are sorted (e.g., `create_file` before `patch`) and executed.
-6.  **Self-Correction**: If a tool fails validation (e.g., non-unique search block), an automatic feedback turn is triggered.
-
-### Visual Selection (Live View)
-1.  `visual-editor-proxy` serves a local target application.
-2.  Developer clicks the floating "Repo" button and selects a UI element.
-3.  Client script traverses the React Fiber tree to identify the component and its source path.
-4.  Component context + prompt are sent to the Backend to initiate a generated plan.
-
----
-
-## 🛠 Global Tech Stack
--   **Languages**: TypeScript (Full-stack)
--   **Backend**: NestJS, Express, TypeORM (SQLite), Eta.js, `pm2`
--   **Frontend**: Next.js 15, React, Tailwind CSS 4, Zustand, Monaco Editor
--   **Desktop**: Electron, Vite, Electron Forge
--   **Inter-Process**: WebSockets (Strongly-typed events), HTTP/JSON
--   **Parsing**: Tiktoken (BPE), Tree-sitter (WASM)
--   **Cloud**: Supabase (Auth/DB)

package/backend/.repoburg/hooks/README.md

@@ -1,53 +0,0 @@
-# Repoburg Tool Hooks
-
-This directory contains TypeScript scripts that can be attached to AI tool execution events.
-You can configure these hooks in the "Tool Hooks" section of the Repoburg UI.
-
-## How it works
-
-1.  **Create a script** in this folder (e.g., `audit-patch.ts`).
-2.  **Go to UI**: Open Repoburg -> Tool Hooks -> Add Hook.
-3.  **Configure**: Select the tool (e.g., `patch`), timing (`before` or `after`), and your script.
-
-## Script Interface
-
-Your script must **export a default function** that receives the context.
-
-### Input Context
-
-```typescript
-interface HookContext {
-  hook_type: 'before' | 'after';
-  
-  action: {
-    tool_name: string; // e.g. "patch"
-    args: any;         // e.g. { file_path: "src/app.ts", patch_code: "..." }
-    
-    // Only present for 'after' hooks
-    result?: {
-      status: 'SUCCESS' | 'FAILURE';
-      output?: string;
-      error?: string;
-    };
-  };
-
-  plan_context: any; // Shared state of the current AI plan
-}
-```
-
-### Output Format
-
-Your function should return an object (or void if passive):
-
-```typescript
-interface HookOutput {
-  // If true, the AI plan stops immediately.
-  should_halt_plan?: boolean;
-  
-  // Optional message to log in the UI
-  message?: string;
-  
-  // Optional data to merge back into the plan context
-  update_context?: Record<string, any>;
-}
-```

package/backend/.repoburg/hooks/eslint-hook.ts

@@ -1,46 +0,0 @@
-/**
- * ESLint Validation Hook
- * 
- * Usage: Attach as 'after' hook to 'create_file', 'edit_file', 'patch'.
- */
-import { exec } from 'child_process';
-import { promisify } from 'util';
-
-const execAsync = promisify(exec);
-
-// Redefining types locally since this runs in a standalone VM context
-export interface HookContext {
-  action: {
-    tool_name: string;
-    args: { file_path?: string; [key: string]: any };
-  };
-  plan_context: {
-    feedback: {
-      validationErrors: Array<{ tool_name: string; error: string }>;
-    };
-  };
-}
-
-export default async function(context: HookContext) {
-  const filePath = context.action.args.file_path;
-  if (!filePath) return;
-
-  console.log(`[Hook] Running ESLint on ${filePath}`);
-
-  try {
-    // Adjust command as needed for your project structure
-    // We use --no-color to keep the error message clean for the AI
-    await execAsync(`npx eslint "${filePath}" --no-color`);
-  } catch (e: any) {
-    const errorOutput = e.stdout || e.stderr || e.message;
-    
-    console.log(`[Hook] ESLint failed for ${filePath}`);
-
-    // Inject error into the shared plan context
-    // The orchestration engine will pick this up and present it to the AI in the next turn (or immediate feedback)
-    context.plan_context.feedback.validationErrors.push({
-      tool_name: context.action.tool_name,
-      error: `ESLint check failed:\n${errorOutput}`
-    });
-  }
-}

package/backend/.repoburg/hooks/example-hook.ts

@@ -1,62 +0,0 @@
-/**
- * Example Hook Script
- * 
- * Usage: 
- * 1. Attach to 'patch' or 'edit_file' tool as a 'before' hook.
- * 2. It will block edits to any file with 'lock' in the name.
- */
-
-// --- Type Definitions ---
-
-export interface HookAction {
-  tool_name: string;
-  args: Record<string, any>;
-  // Only present for 'after' hooks
-  result?: {
-    status: 'SUCCESS' | 'FAILURE';
-    output?: string;
-    error?: string;
-  };
-}
-
-export interface HookContext {
-  hook_type: 'before' | 'after';
-  action: HookAction;
-  plan_context: any; // Shared state of the current AI plan
-}
-
-export interface HookOutput {
-  // If true, the AI plan stops immediately.
-  should_halt_plan?: boolean;
-  
-  // Optional message to log in the UI
-  message?: string;
-  
-  // Optional data to merge back into the plan context
-  update_context?: Record<string, any>;
-}
-
-// --- Hook Logic ---
-
-export default async function(context: HookContext): Promise<HookOutput | void> {
-  // 1. Extract Data
-  const filePath = context.action?.args?.file_path || '';
-  
-  // Log to console (appears in backend logs)
-  console.log(`[Hook] analyzing action on: ${filePath}`);
-
-  // 2. Logic
-  if (filePath.includes('lock')) {
-    // Return Halt Signal
-    return {
-      should_halt_plan: true,
-      message: `BLOCKED: Editing files with "lock" in the name is forbidden by policy. File: ${filePath}`
-    };
-  } else {
-    // Return Pass Signal
-    return {
-      should_halt_plan: false,
-      message: `Hook allowed action on ${filePath}`
-    };
-  }
-}

package/backend/CODEMAP.md

@@ -1,48 +0,0 @@
-# Repoburg Backend Codemap
-
-The local server orchestrating the Repoburg AI workflow. Built with NestJS and TypeORM.
-
-## 🚀 Core AI Loop
-1.  **Input**: `SessionInputsService` takes a prompt, generates context via Eta templates (`SessionInputContextService`), and calls the LLM.
-2.  **Response**: `LlmResponsesService` receives the raw text (from the LLM provider or external sync) and routes it to orchestration.
-3.  **Orchestration**: `LlmTurnProcessorService` parses the response into atomic `Action`s, sorts them (e.g., create before edit), and executes them.
-4.  **Correction**: If a tool fails (validation or syntax), hooks trigger a self-correction turn automatically.
-5.  **Hooks**: `PostExecutionHook`s handle follow-up AI calls, frontend notifications, and "YOLO mode" loops.
-
-## ✨  KKey Feature Modules (`src/`)
--   **`llm-orchestration/`**: [Detailed Codemap](src/llm-orchestration/CODEMAP.md). The engine. Contains the parser, deterministic action sorter, and the registry of `ActionHandler`s.
--   **`ai-actions/`**: [Detailed Codemap](src/ai-actions/CODEMAP.md). Tracks the state of all proposed changes. Handles user approval, manual execution, and the complex `revert` logic.
--   **`application-state/`**: [Detailed Codemap](src/application-state/CODEMAP.md). Global settings and feature flags (e.g., YOLO mode, token limits, execution strategies).
--   **`core-entities/`**: [Detailed Codemap](src/core-entities/CODEMAP.md). The TypeORM data model for projects, sessions, and actions.
--   **`context-generation/`**: [Detailed Codemap](src/context-generation/CODEMAP.md). The rendering engine. Uses Eta.js to build prompts with secure filesystem access (rg, tree, it).
--   **`context-templates/`**: [Detailed Codemap](src/context-templates/CODEMAP.md). Library of reusable prompt structures and partials.
--   **`events/`**: [Detailed Codemap](src/events/CODEMAP.md). Real-time communication layer using WebSockets for UI updates and browser integration.
--   **`execution-logs/`**: [Detailed Codemap](src/execution-logs/CODEMAP.md). captures and stores real-time output from all AI tool executions.
--   * **`interactive-chat/`**: [Detailed Codemap](src/interactive-chat/CODEMAP.md). Manages the in-memory conversation history and system instructions for active sessions.
--   **`llm-provider/`**: [Detailed Codemap](src/llm-provider/CODEMAP.md). Abstraction layer for different LLM engines (OpenRouter, Z.AI, Alibaba, Ollama, OpenAI).
--   **`llm-responses/`**: [Detailed Codemap](src/llm-responses/CODEMAP.md). Ingests and normalizes LLM output from integrated or external sources, triggering the orchestration engine.
--   * **`mcp/`**: [Detailed Codemap](src/mcp/CODEMAP.md). Client implementation for the Model Context Protocol, connecting the AI to external tools and servers.
--   **`message-bus/`**: [Detailed Codemap](src/message-bus/CODEMAP.md). Polling-based communication fallback for browser extensions.
--   **`projects/`**: [Detailed Codemap](src/projects/CODEMAP.md). Maps workspace paths to AI sessions and enforces root directoryoundaries.
--   **`seeding/`**: [Detailed Codemap](src/seeding/CODEMAP.md). Populates the database with default prompts, templates, and snippets on first run.
--   **`sessions/`**: [Detailed Codemap](src/sessions/CODEMAP.md). Manages conversation metadata, projects, and session-specific default templates.
--   **`session-followup/`**: [Detailed Codemap](src/session-followup/CODEMAP.md). logic for spawning new sessions while inheriting context from previous ones.
--   **`session-inputs/`**: [Detailed Codemap](src/session-inputs/CODEMAP.md). Manages the user prompts, context generation via templates, and interaction with the LLM provider.
--   **`session-transfer/`**: [Detailed Codemap](src/session-transfer/CODEMAP.md). Deep import/export functionality for migrating session history and associated metadata.
--   **`system-prompts/`**: [Detailed Codemap](src/system-prompts/CODEMAP.md). Manages the AI's master instructions and handles tool-definition injection.
--   **`tool-hooks/`**: [Detailed Codemap](src/tool-hooks/CODEMAP.md). Extensibility layer. Executes user-provided TypeScript scripts (compiled via VM) before/after tool runs.
--   **`utils/`**: [Detailed Codemap](src/utils/CODEMAP.md). Shared helpers for fuzzy search, tracing, and tool-call generation.
--   **`visual-editor/`**: [Detailed Codemap](src/visual-editor/CODEMAP.md). Backend handler for component selection prompts from the Visual Editor Proxy.
--   **`workspace/`**: [Detailed Codemap](src/workspace/CODEMAP.md). Secure filesystem access. Provides the file tree and `Cmd+P` fuzzy search using RipG   **`mcp/`**: Client implementation for the Model Context Protocol, connecting the AI to external tools and servers.
--   **`interactive-chat/`**: Maintains the conversation history and manages the `LlmProvider` lifecycle.
--   **`events/` & `message-bus/`**: Communication layer. Real-time updates via WebSockets and a polling-based fallback.
-
-## 📦 � Internal Packages (`packages/`)
--   **`tokenpatch/`**: [Detailed Codemap](packages/tokenpatch/CODEMAP.md). A custom, intelligent patching library using token-based matching.
-
-## 🛠 Tech Stack
--   **Framework**: NestJS
--   **Database**: SQLite via TypeORM
--   **Templating**: Eta.js (for context generation)
--   **Observability**: OpenTelemetry (custom `@Trace()` decorator)
--   **Validation**: class-validator & class-transformer

package/backend/README.md

@@ -1,3 +0,0 @@
-# Backend
-
-This is the backend for the Repoburg application.