Z.ai sdk integration (GLM-4.7, GLM-4.6v, GLM-4.6, GLM-4.5-air)

CHANGELOG.md

@@ -0,0 +1,292 @@
+# Changelog
+
+All notable changes to AutoMaker will be documented in this file.
+
+## [Unreleased] - 2025-01
+
+### Added - Multi-Provider UX Enhancements
+
+This release enhances the multi-provider system with improved user experience for provider management and model equivalence routing.
+
+#### Backend Changes
+
+**Model Resolver System**
+
+- Added `resolveModelWithProviderAvailability()` function for provider-aware model resolution
+- `MODEL_EQUIVALENCE` map for cross-provider model fallback:
+  - `claude-opus-4.5-...` ↔ `glm-4.7` (Premium tier)
+  - `claude-sonnet-4.5-...` ↔ `glm-4.6` (Balanced tier)
+  - `claude-haiku-4.5-...` ↔ `glm-4.5-air` (Speed tier)
+- `getProviderForModel()` helper function for provider detection
+- `EnabledProviders` type for provider availability configuration
+
+**Provider Query Layer**
+
+- `executeProviderQuery()` accepts `enabledProviders` option
+- Automatic model substitution when provider is disabled
+- Logged fallback behavior for debugging
+
+**Type System**
+
+- `GlobalSettings` includes `enabledProviders: { claude: boolean; zai: boolean }`
+- `DEFAULT_GLOBAL_SETTINGS` has both providers enabled by default
+- Exported `MODEL_EQUIVALENCE` and `getProviderForModel` from types package
+
+#### Frontend Changes
+
+**Settings UI - Provider Cards**
+
+- Redesigned API Keys section with provider cards
+- Each provider card includes:
+  - Provider name, icon, and description
+  - Enable/disable toggle switch
+  - API key input with visibility toggle
+  - Test connection button
+  - Verification status indicator
+- Auto-enable provider when valid API key is saved
+- Manual toggle to disable providers without removing API keys
+
+**Setup Wizard - Provider Selection**
+
+- New `ProviderSelectionStep` component for choosing primary provider
+- Provider selection cards with feature comparison:
+  - Claude: CLI support + API key, multiple models (Haiku, Sonnet, Opus)
+  - Zai: API key only, GLM models (GLM-4.7, GLM-4.6v, GLM-4.6, GLM-4.5-Air)
+- New `ZaiSetupStep` component for Zai API key configuration
+- Updated wizard flow: `welcome → theme → provider_selection → provider_setup → github → complete`
+- Dynamic routing based on selected provider
+
+**State Management**
+
+- `app-store` includes `enabledProviders` state with actions:
+  - `setEnabledProviders()` - Set provider enable states
+  - `toggleProvider()` - Toggle single provider
+- `setup-store` includes `selectedProvider` for wizard flow
+- Both providers persisted in settings
+
+**New Components**
+
+- `provider-card.tsx` - Reusable provider configuration card
+- `provider-selection-step.tsx` - Provider choice in setup wizard
+- `zai-setup-step.tsx` - Zai-specific setup step
+
+#### Architecture Updates
+
+- Updated `architecture.md` with:
+  - Provider enable/disable UX documentation
+  - Model equivalence routing explanation
+  - Onboarding wizard flow documentation
+  - Provider configuration in state management
+
+### Added - Zai (GLM) Provider Integration
+
+This release adds support for Z.ai's GLM models as an alternative AI provider alongside Claude.
+
+#### Backend Changes
+
+**New Provider Implementation**
+
+- `ZaiProvider` class implementing the `BaseProvider` interface
+  - Full support for GLM-4.7, GLM-4.6v, GLM-4.6, and GLM-4.5-Air models
+  - Tool calling support (Read, Write, Edit, Glob, Grep, Bash)
+  - Vision support via GLM-4.6v with automatic fallback for non-vision models
+  - Extended thinking mode support with `reasoning_content` streaming
+  - Structured output via JSON response format
+
+**Provider-Agnostic Query Layer**
+
+- `executeProviderQuery()` function for unified AI provider routing
+- Automatic model-based provider selection
+- Provider-specific API key management
+- Structured output handling for non-Claude providers
+
+**Provider Factory Updates**
+
+- GLM model prefix routing (`glm-*` → ZaiProvider)
+- Alias support (`glm` → `glm-4.5-air`)
+- Provider discovery and status checking
+
+**Service Layer Updates**
+
+- `AgentService`: Provider-agnostic model selection and query execution
+- `AutoModeService`: Zai model support for autonomous feature execution
+- `SettingsService`: Zai API key storage and migration
+
+**Route Updates**
+
+- All AI routes converted to use `executeProviderQuery()`:
+  - `/api/agent` - Agent conversations
+  - `/api/app-spec/generate-spec` - Specification generation
+  - `/api/app-spec/generate-features` - Feature generation
+  - `/api/suggestions` - AI suggestions
+  - `/api/enhance-prompt` - Prompt enhancement
+  - `/api/context/describe-file` - File description
+
+**New Endpoints**
+
+- `GET /api/models/providers` - Provider status and availability
+- `GET /api/models/available` - Lists all available models from all providers
+
+#### Frontend Changes
+
+**Model Selection**
+
+- Zai models integrated into model selector with badges (Premium, Vision, Balanced, Speed)
+- Model constants with provider metadata
+- Display names and model aliases
+
+**Settings & Configuration**
+
+- Zai provider configuration in API providers config
+- API key management with live validation
+- Authentication status display for Zai
+- Feature defaults support for Zai models
+
+**Profiles**
+
+- Provider auto-detection from model selection
+- Thinking level support for Zai models
+- Model support utilities (`getProviderFromModel`, `modelSupportsThinking`)
+
+**Type System**
+
+- `ModelProvider` type: `'claude' | 'zai'`
+- `ZAI_MODEL_MAP` for model alias resolution
+- Credentials interface includes `zai` property
+- Model display metadata for UI rendering
+
+**Agent View**
+
+- Reasoning content display with collapsible sections
+- Brain icon for extended thinking
+- User setting `showReasoningByDefault` for default expanded state
+
+#### Library Changes
+
+**Model Resolver (`libs/model-resolver`)**
+
+- Provider-aware model resolution with `providerHint` parameter
+- Zai model aliases and display names
+- Unified model string resolution
+
+**Types (`libs/types`)**
+
+- `ModelProvider` type definition
+- `ZAI_MODEL_MAP` constant
+- Model display metadata for GLM models
+- Updated settings types for Zai credentials
+
+#### Tests
+
+**New Test Files**
+
+- `zai-provider.test.ts` (560 lines) - Comprehensive unit tests covering:
+  - Basic query execution
+  - Model selection and configuration
+  - Tool calling (all 6 tools)
+  - System prompt handling
+  - Reasoning content support
+  - Vision support and fallback
+  - Structured output requests
+  - Error handling
+  - Installation detection
+  - Configuration validation
+
+**Updated Tests**
+
+- `provider-factory.test.ts` - Zai model routing tests
+
+#### Documentation
+
+**Architecture Documentation**
+
+- Comprehensive Zai integration section in `architecture.md`:
+  - Provider architecture overview
+  - Model information and capabilities table
+  - Tool mapping documentation
+  - Vision handling explanation
+  - Extended thinking mode documentation
+  - Security features
+  - Configuration guide
+  - Extension points for adding new providers
+
+#### Configuration
+
+**Environment Variables**
+
+- `ZAI_API_KEY` - Z.ai API authentication
+- `AUTOMAKER_MODEL_*` - Use-specific model selection (supports Zai models)
+
+**Model Selection**
+| Variable | Purpose |
+|----------|---------|
+| `AUTOMAKER_MODEL_SPEC` | Model for spec generation |
+| `AUTOMAKER_MODEL_FEATURES` | Model for feature generation |
+| `AUTOMAKER_MODEL_SUGGESTIONS` | Model for suggestions |
+| `AUTOMAKER_MODEL_CHAT` | Model for chat sessions |
+| `AUTOMAKER_MODEL_AUTO` | Model for auto-mode execution |
+| `AUTOMAKER_MODEL_DEFAULT` | Default fallback model |
+
+### Breaking Changes
+
+None. The Zai integration is fully additive and maintains backward compatibility with existing Claude-only workflows.
+
+### Migration Guide
+
+**Adding Zai API Key:**
+
+1. Open Settings → API Keys
+2. Enter your Z.ai API key (get from https://open.bigmodel.cn/)
+3. Click "Test" to validate
+4. Save
+
+**Using Zai Models:**
+
+- Select a GLM model from the model selector in any AI feature
+- Or set `AUTOMAKER_MODEL_DEFAULT=glm-4.7` environment variable
+
+**Provider-Aware Defaults:**
+
+```typescript
+import { resolveModelString } from '@automaker/model-resolver';
+
+// Default to Zai
+const model = resolveModelString(undefined, undefined, 'zai');
+// Returns: 'glm-4.7'
+
+// Default to Claude
+const model = resolveModelString(undefined, undefined, 'claude');
+// Returns: 'claude-opus-4.5-20251101'
+```
+
+**Provider Name Change:**
+
+If your code reads `ModelDefinition.provider`, it will return `'claude'` instead of `'anthropic'`.
+Use `BaseProvider.normalizeProviderName()` for backward compatibility if needed:
+
+```typescript
+import { BaseProvider } from '@/providers/base-provider';
+
+// Legacy code expecting 'anthropic'
+const provider = model.provider; // Returns 'claude'
+
+// Normalize for backward compatibility
+const normalized = BaseProvider.normalizeProviderName(provider);
+// Returns 'anthropic' if provider is 'claude'
+```
+
+### Known Issues
+
+- ~~Streaming not yet implemented for Zai~~ (streaming support is now implemented)
+- Tool execution uses shell commands (requires security hardening for production use)
+- Structured output for non-Claude models uses regex parsing (planned: capability flag)
+- Model display naming inconsistent between views ("Opus" vs "GLM-4.7")
+
+### Future Enhancements
+
+- ~~Add streaming support for Zai models~~ (implemented)
+- Implement native structured output capability flag
+- Security hardening for tool execution
+- E2E tests for Zai integration
+- Provider registration map instead of if/else chains
+- Visual provider indicators in UI

README.md

@@ -322,6 +322,32 @@ DATA_DIR=./data
 
 The application can store your API key securely in the settings UI. The key is persisted in the `DATA_DIR` directory.
 
+#### Option 3: Z.ai (GLM) API Key
+
+If using Z.ai GLM models, you can provide your API key using one of these methods:
+
+##### 3a. Shell Configuration
+
+Add to your `~/.bashrc` or `~/.zshrc`:
+
+```bash
+export ZAI_API_KEY="your-zai-key"
+```
+
+Then restart your terminal or run `source ~/.bashrc` (or `source ~/.zshrc`).
+
+##### 3b. .env File
+
+Add to your `.env` file in the project root:
+
+```bash
+ZAI_API_KEY=your-zai-key
+```
+
+##### 3c. In-App Storage
+
+The Z.ai API key can also be stored securely in the settings UI under API Keys.
+
 ## Features
 
 ### Core Workflow
@@ -334,7 +360,7 @@ The application can store your API key securely in the settings UI. The key is p
 
 ### AI & Planning
 
-- 🧠 **Multi-Model Support** - Choose from Claude Opus, Sonnet, and Haiku per feature
+- 🧠 **Multi-Model Support** - Choose from Claude (Opus, Sonnet, Haiku) or Z.ai (GLM-4.7, GLM-4.6v, GLM-4.6, GLM-4.5-Air) models per feature
 - 💭 **Extended Thinking** - Enable thinking modes (none, medium, deep, ultra) for complex problem-solving
 - 📝 **Planning Modes** - Four planning levels: skip (direct implementation), lite (quick plan), spec (task breakdown), full (phased execution)
 - ✅ **Plan Approval** - Review and approve AI-generated plans before implementation begins
@@ -480,7 +506,7 @@ automaker/
 ### Key Architectural Patterns
 
 - **Event-Driven Architecture** - All server operations emit events that stream to the frontend
-- **Provider Pattern** - Extensible AI provider system (currently Claude, designed for future providers)
+- **Provider Pattern** - Extensible AI provider system supporting Claude (Anthropic) and Z.ai (GLM) models
 - **Service-Oriented Backend** - Modular services for agent management, features, terminals, settings
 - **State Management** - Zustand with persistence for frontend state across restarts
 - **File-Based Storage** - No database; features stored as JSON files in `.automaker/` directory

apps/server/package.json

@@ -31,6 +31,7 @@
     "cors": "^2.8.5",
     "dotenv": "^17.2.3",
     "express": "^5.2.1",
+    "express-rate-limit": "^8.2.1",
     "morgan": "^1.10.1",
     "node-pty": "1.1.0-beta41",
     "ws": "^8.18.3"