gencode-ai 0.1.0 → 0.1.2

package/docs/proposals/README.md

@@ -1,6 +1,6 @@
-# mycode Enhancement Proposals
+# gencode Enhancement Proposals
 
-This directory contains enhancement proposals for the mycode project. Each proposal describes a feature inspired by Claude Code that we plan to implement.
+This directory contains enhancement proposals for the gencode project. Each proposal describes a feature inspired by Claude Code that we plan to implement.
 
 ## Proposal Process
 
@@ -16,18 +16,20 @@ This directory contains enhancement proposals for the mycode project. Each propo
 
 | ID | Title | Status |
 |----|-------|--------|
-| [0001](./0001-web-fetch-tool.md) | WebFetch Tool | Draft |
-| [0002](./0002-web-search-tool.md) | WebSearch Tool | Draft |
+| [0001](./0001-web-fetch-tool.md) | WebFetch Tool | Implemented |
+| [0002](./0002-web-search-tool.md) | WebSearch Tool | Implemented |
 | [0003](./0003-task-subagents.md) | Task Tool & Subagents | Draft |
 | [0004](./0004-plan-mode.md) | Plan Mode | Draft |
-| [0005](./0005-todo-system.md) | Todo System | Draft |
-| [0006](./0006-memory-system.md) | Memory System (MYCODE.md) | Draft |
+| [0005](./0005-todo-system.md) | Todo System | Implemented |
+| [0006](./0006-memory-system.md) | Memory System (MYCODE.md) | Implemented |
 | [0007](./0007-context-management.md) | Context Management | Draft |
 | [0008](./0008-checkpointing.md) | Checkpointing | Draft |
 | [0009](./0009-hooks-system.md) | Hooks System | Draft |
 | [0010](./0010-mcp-integration.md) | MCP Integration | Draft |
 | [0011](./0011-custom-commands.md) | Custom Commands | Draft |
 | [0012](./0012-ask-user-question.md) | AskUserQuestion Tool | Draft |
+| [0041](./0041-configuration-system.md) | Configuration System | Implemented |
+| [0042](./0042-prompt-optimization.md) | Prompt System Optimization | Implemented |
 
 ### P2 Priority - Enhanced Features
 

package/docs/providers.md

@@ -0,0 +1,220 @@
+# Providers
+
+GenCode supports multiple LLM providers and Search providers. Use the `/provider` command to manage all provider connections.
+
+## LLM Providers
+
+### Anthropic (Claude)
+
+Claude models from Anthropic with multiple connection options:
+
+| Connection Method | Environment Variables | Description |
+|-------------------|----------------------|-------------|
+| API Key | `ANTHROPIC_API_KEY` | Direct API access |
+| Google Vertex AI | `ANTHROPIC_VERTEX_PROJECT_ID` or `GOOGLE_CLOUD_PROJECT` | Claude via GCP |
+| Amazon Bedrock | `AWS_ACCESS_KEY_ID` or `AWS_PROFILE` | Claude via AWS (coming soon) |
+
+### OpenAI
+
+GPT models from OpenAI:
+
+| Connection Method | Environment Variables | Description |
+|-------------------|----------------------|-------------|
+| API Key | `OPENAI_API_KEY` | Direct API access |
+
+### Google Gemini
+
+Gemini models from Google:
+
+| Connection Method | Environment Variables | Description |
+|-------------------|----------------------|-------------|
+| API Key | `GOOGLE_API_KEY` or `GEMINI_API_KEY` | Direct API access |
+
+## Search Providers
+
+GenCode includes a WebSearch tool that can search the web for current information. Multiple search providers are supported:
+
+### Exa AI (Default)
+
+AI-native search engine using Exa's public MCP endpoint. Works out of the box without any configuration.
+
+| Connection Method | Environment Variables | Description |
+|-------------------|----------------------|-------------|
+| Public MCP | (none required) | Uses public `mcp.exa.ai` endpoint |
+
+**Features:**
+- No configuration required - works immediately
+- AI-optimized search results
+- Live crawl support for fresh content
+
+**Note:** Uses Exa's public MCP endpoint. For heavy usage, consider using Serper or Brave with your own API key.
+
+### Serper.dev
+
+Google Search results via Serper API.
+
+| Connection Method | Environment Variables | Description |
+|-------------------|----------------------|-------------|
+| API Key | `SERPER_API_KEY` | Google Search via Serper |
+
+**Features:**
+- 2,500 free queries (no credit card required)
+- Google Search quality results
+- Fast response times
+
+**Get API Key:** https://serper.dev
+
+### Brave Search
+
+Privacy-focused search from Brave.
+
+| Connection Method | Environment Variables | Description |
+|-------------------|----------------------|-------------|
+| API Key | `BRAVE_API_KEY` | Privacy-focused search |
+
+**Features:**
+- 2,000 free queries per month
+- Privacy-first approach
+- Independent search index
+
+**Get API Key:** https://brave.com/search/api
+
+### Search Provider Priority
+
+When multiple search providers are configured, the priority is:
+
+1. **Configured in `/provider`** - Explicitly selected provider
+2. **Environment variables** - `SERPER_API_KEY` > `BRAVE_API_KEY`
+3. **Default** - Exa AI (no configuration needed)
+
+## Commands
+
+### `/provider` - Provider Management
+
+Opens the provider management interface with two tabs:
+
+**[L] LLM Providers Tab:**
+- View connected and available LLM providers
+- Connect to new providers
+- Refresh model lists for connected providers
+- Remove provider connections
+
+**[S] Search Providers Tab:**
+- View available search providers
+- Select which search provider to use for WebSearch
+- See configuration status (configured/not configured)
+
+**Keyboard shortcuts:**
+- `Tab` or `L`/`S` - Switch between tabs
+- `↑↓` - Navigate between providers
+- `Enter` - Connect/Select provider
+- `r` - Remove a connected LLM provider
+- `Esc` - Exit
+
+### `/model` - Model Selection
+
+Opens the model selector to switch between models from all connected providers:
+
+- Models are grouped by provider
+- Shows cached models (fast, no API call)
+- Supports fuzzy search filtering
+
+**Keyboard shortcuts:**
+- `↑↓` - Navigate between models
+- `Enter` - Select model
+- `Esc` - Cancel
+
+## Google Vertex AI Setup
+
+To use Claude models via Google Vertex AI:
+
+### 1. Set Environment Variables
+
+```bash
+# Required: Your GCP project ID
+export ANTHROPIC_VERTEX_PROJECT_ID="your-project-id"
+
+# Optional: Region (defaults to us-east5)
+export ANTHROPIC_VERTEX_REGION="us-east5"
+```
+
+Alternative variables also supported:
+- `GOOGLE_CLOUD_PROJECT` - GCP project ID
+- `CLOUD_ML_REGION` - GCP region
+
+### 2. Authenticate with Google Cloud
+
+```bash
+# Login and set up Application Default Credentials
+gcloud auth application-default login
+
+# Verify authentication
+gcloud auth application-default print-access-token
+```
+
+### 3. Enable Vertex AI API
+
+Ensure the Vertex AI API is enabled in your GCP project:
+
+```bash
+gcloud services enable aiplatform.googleapis.com
+```
+
+### 4. Connect in GenCode
+
+```
+/provider
+# Select "Anthropic"
+# Choose "Google Vertex AI" connection method
+# Press Enter to connect
+```
+
+## Configuration Storage
+
+Provider connections and cached models are stored in:
+
+```
+~/.gencode/
+├── settings.json      # Current model and provider settings
+└── providers.json     # Provider connections and model cache
+```
+
+### providers.json Structure
+
+```json
+{
+  "connections": {
+    "anthropic": {
+      "method": "vertex",
+      "connectedAt": "2025-01-15T10:00:00Z"
+    }
+  },
+  "models": {
+    "anthropic": {
+      "cachedAt": "2025-01-15T10:00:00Z",
+      "list": [
+        { "id": "claude-sonnet-4-5@20250929", "name": "Claude Sonnet 4.5" }
+      ]
+    }
+  },
+  "searchProvider": "exa"
+}
+```
+
+The `searchProvider` field stores the selected search provider. Valid values: `exa`, `serper`, `brave`. If not set, defaults to `exa`.
+
+## Troubleshooting
+
+### "Not configured" status
+
+If a provider shows "(not configured)", set the required environment variables and restart GenCode.
+
+### Model list is empty
+
+Use `/provider` and press Enter on a connected provider to refresh the model cache.
+
+### Vertex AI authentication errors
+
+1. Verify you're logged in: `gcloud auth application-default print-access-token`
+2. Check project ID: `echo $ANTHROPIC_VERTEX_PROJECT_ID`
+3. Ensure Vertex AI API is enabled in your project

package/jest.config.js

@@ -0,0 +1,26 @@
+/** @type {import('ts-jest').JestConfigWithTsJest} */
+export default {
+  preset: 'ts-jest/presets/default-esm',
+  testEnvironment: 'node',
+  extensionsToTreatAsEsm: ['.ts'],
+  moduleNameMapper: {
+    '^(\\.{1,2}/.*)\\.js$': '$1',
+  },
+  transform: {
+    '^.+\\.tsx?$': [
+      'ts-jest',
+      {
+        useESM: true,
+        isolatedModules: true,
+      },
+    ],
+  },
+  testMatch: ['**/*.test.ts'],
+  collectCoverageFrom: [
+    'src/**/*.ts',
+    '!src/**/*.d.ts',
+    '!src/**/*.test.ts',
+  ],
+  coverageDirectory: 'coverage',
+  moduleFileExtensions: ['ts', 'tsx', 'js', 'jsx', 'json', 'node'],
+};

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "gencode-ai",
-  "version": "0.1.0",
-  "description": "GenCode - An open-source, provider-agnostic AI coding assistant",
+  "version": "0.1.2",
+  "description": "An open-source AI assistant for your terminal",
   "type": "module",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -11,7 +11,8 @@
   "scripts": {
     "build": "tsc",
     "dev": "tsc --watch",
-    "start": "npx tsx src/cli/index.tsx",
+    "start": "node dist/cli/index.js",
+    "start:dev": "npx tsx src/cli/index.tsx",
     "test": "node --experimental-vm-modules node_modules/jest/bin/jest.js",
     "example": "npx tsx examples/basic.ts"
   },
@@ -28,22 +29,32 @@
   "dependencies": {
     "@anthropic-ai/sdk": "^0.71.2",
     "@google/generative-ai": "^0.24.1",
+    "@types/turndown": "^5.0.6",
     "chalk": "^5.6.2",
     "dotenv": "^17.2.3",
     "fast-glob": "^3.3.3",
     "glob": "^13.0.0",
+    "google-auth-library": "^10.5.0",
+    "gray-matter": "^4.0.3",
     "ink": "^6.6.0",
     "ink-spinner": "^5.0.0",
     "ink-text-input": "^6.0.0",
+    "marked": "^17.0.1",
+    "minimatch": "^10.1.1",
     "openai": "^6.16.0",
     "ora": "^9.0.0",
     "react": "^19.2.3",
+    "turndown": "^7.2.2",
     "undici": "^7.18.2",
     "zod": "^4.3.5"
   },
   "devDependencies": {
+    "@types/jest": "^30.0.0",
+    "@types/minimatch": "^5.1.2",
     "@types/node": "^25.0.8",
     "@types/react": "^19.2.8",
+    "jest": "^30.2.0",
+    "ts-jest": "^29.4.6",
     "tsx": "^4.21.0",
     "typescript": "^5.9.3"
   }

package/src/agent/agent.ts

@@ -3,31 +3,30 @@
  */
 
 import type { LLMProvider, Message, ToolResultContent } from '../providers/types.js';
-import { createProvider } from '../providers/index.js';
+import { createProvider, inferProvider } from '../providers/index.js';
 import { ToolRegistry, createDefaultRegistry } from '../tools/index.js';
-import { PermissionManager } from '../permissions/index.js';
+import {
+  PermissionManager,
+  type ApprovalAction,
+  type ApprovalSuggestion,
+  type PromptPermission,
+  type PermissionSettings,
+} from '../permissions/index.js';
 import { SessionManager } from '../session/index.js';
+import { MemoryManager, type LoadedMemory } from '../memory/index.js';
 import type { AgentConfig, AgentEvent } from './types.js';
-
-const DEFAULT_SYSTEM_PROMPT = `You are a helpful AI assistant with access to tools for reading, writing, and executing code.
-
-When using tools:
-- Use Read to view file contents before editing
-- Use Glob/Grep to find files
-- Use Edit for precise changes (old_string must be unique)
-- Use Write for new files or full rewrites
-- Use Bash for commands, git operations, etc.
-
-Be concise and focus on completing the user's task.`;
+import { buildSystemPromptForModel, debugPromptLoading } from '../prompts/index.js';
 
 export class Agent {
   private provider: LLMProvider;
   private registry: ToolRegistry;
   private permissions: PermissionManager;
   private sessionManager: SessionManager;
+  private memoryManager: MemoryManager;
   private config: AgentConfig;
   private messages: Message[] = [];
   private sessionId: string | null = null;
+  private loadedMemory: LoadedMemory | null = null;
 
   constructor(config: AgentConfig) {
     this.config = {
@@ -38,17 +37,93 @@ export class Agent {
 
     this.provider = createProvider({ provider: config.provider });
     this.registry = createDefaultRegistry();
-    this.permissions = new PermissionManager(config.permissions);
+    this.permissions = new PermissionManager({
+      config: config.permissions,
+      projectPath: config.cwd,
+    });
     this.sessionManager = new SessionManager();
+    this.memoryManager = new MemoryManager();
+  }
+
+  /**
+   * Initialize permission system (load persisted rules)
+   */
+  async initializePermissions(settings?: PermissionSettings): Promise<void> {
+    await this.permissions.initialize(settings);
   }
 
   /**
-   * Set permission confirmation callback
+   * Set simple permission confirmation callback (backward compatible)
    */
   setConfirmCallback(callback: (tool: string, input: unknown) => Promise<boolean>): void {
+    this.permissions.setSimpleConfirmCallback(callback);
+  }
+
+  /**
+   * Set enhanced permission confirmation callback with approval options
+   */
+  setEnhancedConfirmCallback(
+    callback: (
+      tool: string,
+      input: unknown,
+      suggestions: ApprovalSuggestion[]
+    ) => Promise<ApprovalAction>
+  ): void {
     this.permissions.setConfirmCallback(callback);
   }
 
+  /**
+   * Add prompt-based permissions (Claude Code ExitPlanMode style)
+   */
+  addAllowedPrompts(prompts: PromptPermission[]): void {
+    this.permissions.addAllowedPrompts(prompts);
+  }
+
+  /**
+   * Clear prompt-based permissions
+   */
+  clearAllowedPrompts(): void {
+    this.permissions.clearAllowedPrompts();
+  }
+
+  /**
+   * Set callback to save permission rules to settings
+   * This enables saving rules to settings.local.json instead of permissions.json
+   */
+  setSaveRuleCallback(callback: (tool: string, pattern?: string) => Promise<void>): void {
+    this.permissions.setSaveRuleCallback(callback);
+  }
+
+  /**
+   * Get permission manager for direct access
+   */
+  getPermissionManager(): PermissionManager {
+    return this.permissions;
+  }
+
+  /**
+   * Get memory manager for direct access
+   */
+  getMemoryManager(): MemoryManager {
+    return this.memoryManager;
+  }
+
+  /**
+   * Get loaded memory (null if not loaded yet)
+   */
+  getLoadedMemory(): LoadedMemory | null {
+    return this.loadedMemory;
+  }
+
+  /**
+   * Load memory for the current working directory
+   */
+  async loadMemory(): Promise<LoadedMemory> {
+    const cwd = this.config.cwd ?? process.cwd();
+    this.loadedMemory = await this.memoryManager.load({ cwd });
+    return this.loadedMemory;
+  }
+
   /**
    * Get current session ID
    */
@@ -64,10 +139,17 @@ export class Agent {
   }
 
   /**
-   * Set the model to use
+   * Set the model to use (auto-switches provider if needed)
    */
   setModel(model: string): void {
     this.config.model = model;
+
+    // Auto-switch provider based on model name
+    const newProvider = inferProvider(model);
+    if (newProvider !== this.config.provider) {
+      this.config.provider = newProvider;
+      this.provider = createProvider({ provider: newProvider });
+    }
   }
 
   /**
@@ -179,6 +261,11 @@ export class Agent {
       await this.startSession();
     }
 
+    // Load memory if not already loaded
+    if (!this.loadedMemory) {
+      await this.loadMemory();
+    }
+
     // Add user message
     const userMessage: Message = { role: 'user', content: prompt };
     this.messages.push(userMessage);
@@ -196,11 +283,26 @@ export class Agent {
       // Call LLM
       let response;
       try {
+        // Debug prompt loading (enabled with GENCODE_DEBUG_PROMPTS=1)
+        debugPromptLoading(this.config.model, this.config.provider);
+
+        // Build system prompt based on model → provider → prompt flow
+        // Looks up provider from ~/.gencode/providers.json, falls back to config.provider
+        const systemPrompt =
+          this.config.systemPrompt ??
+          buildSystemPromptForModel(
+            this.config.model,
+            this.config.cwd ?? process.cwd(),
+            true, // Assume git repo for now
+            this.loadedMemory?.context,
+            this.config.provider // Fallback provider if model lookup fails
+          );
+
         response = await this.provider.complete({
           model: this.config.model,
           messages: this.messages,
           tools: toolDefs,
-          systemPrompt: this.config.systemPrompt ?? DEFAULT_SYSTEM_PROMPT,
+          systemPrompt,
           maxTokens: 4096,
         });
       } catch (error) {
@@ -237,7 +339,7 @@ export class Agent {
       for (const call of toolCalls) {
         yield { type: 'tool_start', id: call.id, name: call.name, input: call.input };
 
-        const allowed = await this.permissions.checkPermission(call.name, call.input);
+        const allowed = await this.permissions.requestPermission(call.name, call.input);
         const result = allowed
           ? await this.registry.execute(call.name, call.input, { cwd })
           : { success: false, output: '', error: 'Permission denied by user' };

package/src/agent/types.ts

@@ -5,7 +5,7 @@
 import type { PermissionConfig } from '../permissions/types.js';
 
 export interface AgentConfig {
-  provider: 'openai' | 'anthropic' | 'gemini';
+  provider: 'openai' | 'anthropic' | 'gemini' | 'vertex-ai';
   model: string;
   systemPrompt?: string;
   tools?: string[];
@@ -35,6 +35,14 @@ export interface AgentEventToolResult {
     success: boolean;
     output: string;
     error?: string;
+    metadata?: {
+      title?: string;
+      subtitle?: string;
+      size?: number;
+      statusCode?: number;
+      contentType?: string;
+      duration?: number;
+    };
   };
 }