claude-ai-switcher 1.1.4

package/README.md

@@ -0,0 +1,833 @@
+# Claude AI Switcher
+
+Switch between AI providers (Anthropic, GLM, Alibaba Qwen, OpenRouter, Ollama, Gemini) for **Claude Code** with ease. Also provides helper commands to manage providers for **OpenCode**.
+
+## Features
+
+- **Quick Switching**: Switch between Anthropic, GLM/Z.AI, Alibaba Coding Plan, OpenRouter, Ollama (local), and Gemini (Google) with a single command
+- **Model Aliases**: Automatically sets `ANTHROPIC_DEFAULT_OPUS_MODEL`, `ANTHROPIC_DEFAULT_SONNET_MODEL`, and `ANTHROPIC_DEFAULT_HAIKU_MODEL` in Claude Code's settings so you always know what model is active
+- **Custom Tier Overrides**: Use `--opus`, `--sonnet`, `--haiku` to pin specific models per tier
+- **API Key Verification**: Run `claude-switch status` to check current config and verify API keys are valid
+- **Local Models**: Use Ollama for fully local, private AI with DeepSeek R1, Qwen, Llama, and more
+- **Google Gemini**: Access Gemini 2.5 Pro/Flash models with 1M context via LiteLLM proxy
+- **Auto Proxy Management**: Automatically starts LiteLLM translation proxy for Ollama and Gemini
+- **Model Information**: See model capabilities, context windows, and descriptions when switching
+- **Token Tracking**: Visual context bar with percentage usage display
+- **Visual Enhancements**: Model card with provider info, active model display, and context usage
+- **Secure Storage**: API keys stored locally in `~/.claude-ai-switcher/config.json`
+- **Safe Configuration**: Backs up existing settings before any modifications
+- **Auto Onboarding**: Automatically sets `hasCompletedOnboarding: true` to prevent connection errors
+- **OpenCode Helper**: Add/remove provider configurations for OpenCode with simple commands
+
+## Installation
+
+```bash
+cd claude-ai-switcher
+npm ci  # reproducible install from package-lock.json (does NOT modify it)
+npm run build
+npm link  # Install globally
+```
+
+> **Dependency flow:** Use `npm ci` for routine installs and after every pull — it installs exactly from `package-lock.json` and never modifies it, so it can't drift the lockfile or cause pull conflicts. Use `npm install <package>` or `npm update` only when intentionally adding or upgrading a dependency, then commit the updated `package-lock.json`.
+
+## Platform Support
+
+| Platform | Status | Notes |
+|----------|--------|-------|
+| macOS | ✅ Full Support | Default development platform |
+| Linux | ✅ Full Support | All features supported |
+| Windows 11 | ✅ Full Support | Use Git Bash, WSL, or PowerShell for build commands |
+
+### Windows Notes
+
+- Build commands (`npm ci`, `npm run build`) should be run in Git Bash, WSL, or PowerShell
+- The CLI works in any terminal after installation
+- GLM/Z.AI provider requires `coding-helper` to be in your PATH
+- The `clean` script uses `rimraf` for cross-platform compatibility
+
+## Quick Start
+
+```bash
+# Run setup wizard (recommended for first-time use)
+claude-switch setup
+
+# Switch Claude Code to Alibaba Coding Plan
+claude-switch alibaba
+
+# Switch Claude Code with specific model
+claude-switch alibaba qwen3.6-plus
+
+# Switch Claude Code to OpenRouter (free models)
+claude-switch openrouter
+
+# Switch to local Ollama models (requires LiteLLM proxy)
+claude-switch ollama
+
+# Switch to Google Gemini (requires LiteLLM proxy)
+claude-switch gemini
+
+# Switch back to default Anthropic
+claude-switch anthropic
+
+# Add Alibaba provider to OpenCode
+claude-switch opencode add alibaba
+
+# Add OpenRouter provider to OpenCode
+claude-switch opencode add openrouter
+
+# Add Ollama, Gemini, or GLM provider to OpenCode
+claude-switch opencode add ollama
+claude-switch opencode add gemini
+claude-switch opencode add glm
+
+# Remove providers
+claude-switch opencode remove alibaba
+claude-switch opencode remove openrouter
+claude-switch opencode remove ollama
+claude-switch opencode remove gemini
+claude-switch opencode remove glm
+```
+
+## Commands
+
+### Switch Claude Code (top-level commands)
+
+```bash
+claude-switch anthropic
+claude-switch alibaba
+claude-switch alibaba qwen3.6-plus       # specific model
+claude-switch glm
+claude-switch openrouter                   # default: qwen/qwen3.6-plus:free
+claude-switch openrouter openrouter/free   # specific model
+claude-switch ollama                       # default: deepseek-r1:latest (starts proxy)
+claude-switch ollama qwen2.5-coder:latest  # specific model
+claude-switch gemini                       # default: gemini-2.5-pro (starts proxy)
+claude-switch gemini gemini-2.5-flash      # specific model
+```
+
+### Switch Claude Code (explicit targeting)
+
+```bash
+claude-switch claude anthropic
+claude-switch claude alibaba
+claude-switch claude alibaba qwen3.6-plus
+claude-switch claude glm
+claude-switch claude openrouter
+claude-switch claude openrouter qwen/qwen3.6-plus:free
+claude-switch claude ollama
+claude-switch claude ollama deepseek-r1:latest
+claude-switch claude gemini
+claude-switch claude gemini gemini-2.5-flash
+```
+
+### OpenCode Helper Commands
+
+Add or remove providers from OpenCode configuration:
+
+```bash
+# Add providers to OpenCode
+claude-switch opencode add alibaba
+claude-switch opencode add openrouter
+claude-switch opencode add ollama
+claude-switch opencode add gemini
+claude-switch opencode add glm
+
+# Remove providers from OpenCode
+claude-switch opencode remove alibaba
+claude-switch opencode remove openrouter
+claude-switch opencode remove ollama
+claude-switch opencode remove gemini
+claude-switch opencode remove glm
+```
+
+**Note**: OpenCode configuration is stored at `~/.config/opencode/opencode.json`. The `add` command adds the provider with all available models. The `remove` command removes only the specified provider, preserving any other providers you have configured.
+
+### Model Aliases (Claude Code)
+
+When switching Claude Code to a non-Anthropic provider, the tool writes model alias env vars into `~/.claude/settings.json` so Claude Code knows which model to route each tier to:
+
+| Env Var | Default (Alibaba) | Default (GLM) | Default (OpenRouter) | Default (Ollama) | Default (Gemini) |
+|---------|-------------------|---------------|---------------------|-----------------|-----------------|
+| `ANTHROPIC_DEFAULT_OPUS_MODEL` | `qwen3.7-plus` | `glm-5.2[1m]` | `qwen/qwen3.6-plus:free` | `deepseek-r1:latest` | `gemini-2.5-pro` |
+| `ANTHROPIC_DEFAULT_SONNET_MODEL` | `qwen3.6-plus` | `glm-5-turbo` | `openrouter/free` | `qwen2.5-coder:latest` | `gemini-2.5-flash` |
+| `ANTHROPIC_DEFAULT_HAIKU_MODEL` | `kimi-k2.5` | `glm-5v-turbo` | `openrouter/free` | `llama3.1:latest` | `gemini-2.5-flash-lite` |
+
+Override any tier at switch time:
+
+```bash
+# Set all three tiers (default for Alibaba)
+claude-switch alibaba qwen3.7-plus
+
+# Override individual tiers
+claude-switch alibaba --opus qwen3-max-2026-01-23 --sonnet qwen3.6-plus --haiku kimi-k2.5
+
+# Custom configuration with specific model assignments
+claude-switch alibaba --opus qwen3.7-plus --sonnet qwen3.6-plus --haiku kimi-k2.5
+
+# GLM with custom haiku tier
+claude-switch glm --haiku glm-4.7
+
+# OpenRouter with custom tiers
+claude-switch openrouter --sonnet qwen/qwen3.6-plus:free --haiku openrouter/free
+
+# Ollama with custom tiers
+claude-switch ollama --opus deepseek-r1:latest --sonnet qwen2.5-coder:latest
+
+# Gemini with custom tiers
+claude-switch gemini --opus gemini-2.5-pro --haiku gemini-2.5-flash-lite
+```
+
+Switching back to Anthropic clears these env vars so native Claude models are used again.
+
+### View Information
+
+```bash
+# Show current config + verify API keys (recommended)
+claude-switch status
+
+# Show current configuration for both clients
+claude-switch current
+
+# List all providers and models
+claude-switch list
+
+# Show models for specific provider
+claude-switch models alibaba
+claude-switch models glm
+claude-switch models anthropic
+claude-switch models openrouter
+claude-switch models ollama
+claude-switch models gemini
+```
+
+### API Key Management
+
+```bash
+# Set API key
+claude-switch key alibaba <your-api-key>
+claude-switch key openrouter <your-api-key>
+claude-switch key gemini <your-api-key>
+
+# Check if API key is set
+claude-switch key alibaba
+claude-switch key openrouter
+claude-switch key gemini
+```
+
+> **Note**: Ollama does not require an API key (runs locally). Anthropic uses the `ANTHROPIC_API_KEY` environment variable.
+
+### Hooks - Token Tracking & Visual Enhancements
+
+Install visual enhancements and token tracking for Claude Code:
+
+```bash
+# Install all hooks (token tracker + visual enhancements)
+claude-switch hooks install
+
+✓ Hooks installed successfully!
+
+  Installed:
+    • Token Tracker (~/.claude/token-tracker.js)
+    • Visual Enhancements (~/.claude/visual-enhancements.js)
+
+  Usage:
+    • Token usage is tracked automatically
+    • Run 'claude-switch hooks status' to see current usage
+    • Run 'claude-switch hooks reset' to reset counters
+```
+
+```bash
+# Show current token usage and visual status
+claude-switch hooks status
+
+=== Hooks Status ===
+
+  Token Tracker: ✓ Installed
+  Visual Enhancements: ✓ Installed
+
+╔══════════════════════════════════════════════════════════════╗
+║  🤖 Active Model: Qwen3 6 Plus                                ║
+╠══════════════════════════════════════════════════════════════╣
+║  📊 Token Usage:                                              ║
+║    Input:  12,450      tokens                                   ║
+║    Output: 8,320       tokens                                   ║
+║    Total:  20,770      tokens                                   ║
+╠══════════════════════════════════════════════════════════════╣
+║  📈 Context Window:                                           ║
+║    Used:   20,770      tokens                                   ║
+║    Total:  1,000,000   tokens                                   ║
+║    ████░░░░░░░░░░░░░░░░   2.1%                                  ║
+╚══════════════════════════════════════════════════════════════╝
+```
+
+```bash
+# Reset token usage counters
+claude-switch hooks reset
+
+Token usage reset complete.
+```
+
+```bash
+# Remove all hooks
+claude-switch hooks remove
+
+✓ All hooks removed
+```
+
+**What you get:**
+
+| Feature | Description |
+|---------|-------------|
+| **Token Tracker** | Tracks input/output tokens across your session with visual percentage bar |
+| **Model Card** | Shows active model, provider, context window, and capabilities |
+| **Context Bar** | Visual progress bar showing context usage percentage (color-coded) |
+| **Auto Display** | Shows model info and token usage when you start a Claude Code session |
+
+**Context Bar Colors:**
+
+| Color | Usage Range | Meaning |
+|-------|-------------|---------|
+| 🟢 Green | 0-50% | Safe usage |
+| 🟡 Yellow | 50-75% | Moderate usage |
+| 🔴 Red | 75-90% | High usage |
+| 🟣 Magenta | 90-100% | Critical - approaching limit |
+
+**Example Output:**
+
+```bash
+# Token tracker standalone display
+node ~/.claude/token-tracker.js
+
+╔══════════════════════════════════════════════════════════════╗
+║  🤖 Active Model: GLM 5 1                                     ║
+╠══════════════════════════════════════════════════════════════╣
+║  📊 Token Usage:                                              ║
+║    Input:  45,200      tokens                                   ║
+║    Output: 32,100      tokens                                   ║
+║    Total:  77,300      tokens                                   ║
+╠══════════════════════════════════════════════════════════════╣
+║  📈 Context Window:                                           ║
+║    Used:   77,300      tokens                                   ║
+║    Total:  200,000     tokens                                   ║
+║    ████████████░░░░░░░░  38.7%                                  ║
+╚══════════════════════════════════════════════════════════════╝
+```
+
+```bash
+# Visual enhancements standalone display
+node ~/.claude/visual-enhancements.js
+
+┌─────────────────────────────────────────────────────────────┐
+│ 🤖 Alibaba Model Studio                                          │
+├─────────────────────────────────────────────────────────────┤
+│ Model: qwen3.6-plus                                             │
+│ Context: 1M tokens                                              │
+│ Capabilities:                                                    │
+│   Text Generation • Deep Thinking • Visual Understanding        │
+─────────────────────────────────────────────────────────────┘
+```
+
+### Setup
+
+```bash
+# Interactive setup wizard
+claude-switch setup
+```
+
+## Available Models
+
+### Alibaba Coding Plan
+
+| Model | Context | Capabilities |
+|-------|---------|--------------|
+| qwen3.7-plus | 1M tokens | Text Generation, Deep Thinking, Visual Understanding |
+| qwen3.6-plus | 1M tokens | Text Generation, Deep Thinking, Visual Understanding |
+| qwen3-max-2026-01-23 | 262K tokens | Text Generation, Deep Thinking |
+| qwen3-coder-next | 262K tokens | Text Generation (Advanced Coding Agent) |
+| qwen3-coder-plus | 1M tokens | Text Generation (Code, Tool Calling, Autonomous) |
+| glm-5 | 200K tokens | Text Generation, Deep Thinking |
+| glm-4.7 | 256K tokens | Text Generation, Deep Thinking |
+| glm-4.7-flash | 256K tokens | Text Generation, Fast Inference |
+| kimi-k2.5 | 200K tokens | Text Generation, Deep Thinking, Visual |
+| MiniMax-M2.5 | 200K tokens | Text Generation, Deep Thinking |
+
+### GLM/Z.AI (via coding-helper)
+
+| Model | Context | Capabilities |
+|-------|---------|--------------|
+| glm-5.2[1m] | 1M tokens | Text Generation, Deep Thinking |
+| glm-5v-turbo | 200K tokens | Text Generation, Deep Thinking, Visual Understanding, Visual Programming |
+| glm-5-turbo | 200K tokens | Text Generation, Deep Thinking, Fast Responses |
+| glm-5.1 | 200K tokens | Text Generation, Deep Thinking |
+| glm-4.7 | 256K tokens | Text Generation, Deep Thinking |
+
+### Anthropic (Default)
+
+| Model | Context | Capabilities |
+|-------|---------|--------------|
+| claude-opus-4-6-20250205 | 200K tokens | Text, Code, Vision, Complex Reasoning |
+| claude-opus-4-5-20251101 | 200K tokens | Text, Code, Vision, Complex Reasoning |
+| claude-sonnet-4-6-20250219 | 200K tokens | Text, Code, Vision |
+| claude-sonnet-4-5-20250814 | 200K tokens | Text, Code, Vision |
+| claude-haiku-4-5-20251015 | 200K tokens | Text, Fast Responses |
+
+### OpenRouter
+
+| Model | Context | Capabilities |
+|-------|---------|--------------|
+| `qwen/qwen3.6-plus:free` | 131K tokens | Text Generation, Deep Thinking |
+| `openrouter/free` | 131K tokens | Text Generation |
+
+### Ollama (Local, via LiteLLM proxy)
+
+| Model | Context | Capabilities |
+|-------|---------|--------------|
+| `deepseek-r1:latest` | 128K tokens | Text Generation, Deep Thinking, Reasoning |
+| `qwen2.5-coder:latest` | 128K tokens | Text Generation, Coding, Tool Calling |
+| `llama3.1:latest` | 128K tokens | Text Generation, Code, Vision |
+| `codellama:latest` | 100K tokens | Text Generation, Coding |
+
+> **Prerequisites**: [Ollama](https://ollama.com) must be installed and running on port 11434. [LiteLLM](https://github.com/BerriAI/litellm) proxy is auto-started on port 4000.
+
+### Gemini (Google, via LiteLLM proxy)
+
+| Model | Context | Capabilities |
+|-------|---------|--------------|
+| `gemini-2.5-pro` | 1M tokens | Text Generation, Deep Thinking, Code, Vision |
+| `gemini-2.5-flash` | 1M tokens | Text Generation, Fast Responses, Code |
+| `gemini-2.5-flash-lite` | 1M tokens | Text Generation, Cost-optimized |
+
+> **Prerequisites**: Google API key from [AI Studio](https://aistudio.google.com/apikey). [LiteLLM](https://github.com/BerriAI/litellm) proxy is auto-started on port 4001.
+
+## Configuration Files
+
+| Client | Config File | Purpose |
+|--------|-------------|---------|
+| Claude Code | `~/.claude/settings.json` | Environment variables for provider config + model alias env vars |
+| Claude Code | `~/.claude.json` | `hasCompletedOnboarding` flag |
+| OpenCode | `~/.config/opencode/opencode.json` | Provider configuration (bailian-coding-plan) |
+| API Keys | `~/.claude-ai-switcher/config.json` | Secure local API key storage |
+
+## How It Works
+
+### Provider Architecture
+
+```
+┌─────────────┐     ┌──────────────┐     ┌───────────────────┐
+│ Claude Code  │────►│  ANTHROPIC_  │────►│ Target Provider   │
+│ / OpenCode   │     │ BASE_URL     │     │                   │
+└─────────────┘     └──────────────┘     ├───────────────────┤
+                                          │ Anthropic (direct)│
+                                          │ Alibaba  (direct)│
+                                          │ OpenRouter(direct)│
+                                          │ GLM (via helper) │
+                                          ├───────────────────┤
+                                          │ LiteLLM Proxy     │
+                                          │ ├─ Ollama :4000   │
+                                          │ └─ Gemini :4001   │
+                                          └───────────────────┘
+```
+
+Anthropic, Alibaba, OpenRouter, and GLM speak the Anthropic API format natively. Ollama and Gemini only speak OpenAI format, so a **LiteLLM proxy** translates between protocols. The proxy is auto-started when you switch to Ollama or Gemini.
+
+### Claude Code Configuration
+
+When you switch Claude Code to Alibaba, the tool writes these environment variables to `~/.claude/settings.json`:
+
+```json
+{
+  "env": {
+    "ANTHROPIC_AUTH_TOKEN": "YOUR_API_KEY",
+    "ANTHROPIC_BASE_URL": "https://coding-intl.dashscope.aliyuncs.com/apps/anthropic",
+    "ANTHROPIC_MODEL": "qwen3.7-plus",
+    "ANTHROPIC_DEFAULT_OPUS_MODEL": "qwen3.7-plus",
+    "ANTHROPIC_DEFAULT_SONNET_MODEL": "qwen3.6-plus",
+    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "kimi-k2.5"
+  }
+}
+```
+
+- `ANTHROPIC_AUTH_TOKEN`, `ANTHROPIC_BASE_URL`, `ANTHROPIC_MODEL` — Route Claude Code to Alibaba's endpoint
+- `ANTHROPIC_DEFAULT_*_MODEL` — Map Claude's model tiers (opus/sonnet/haiku) to specific Alibaba models
+
+Switching back to Anthropic clears all these env vars.
+
+### Ollama Configuration
+
+When you switch Claude Code to Ollama, the tool:
+
+1. Checks that LiteLLM and Ollama are installed and Ollama is running
+2. Starts a LiteLLM proxy on port 4000 (if not already running)
+3. Writes these environment variables to `~/.claude/settings.json`:
+
+```json
+{
+  "env": {
+    "ANTHROPIC_AUTH_TOKEN": "ollama",
+    "ANTHROPIC_BASE_URL": "http://localhost:4000",
+    "ANTHROPIC_MODEL": "deepseek-r1:latest",
+    "ANTHROPIC_DEFAULT_OPUS_MODEL": "deepseek-r1:latest",
+    "ANTHROPIC_DEFAULT_SONNET_MODEL": "qwen2.5-coder:latest",
+    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "llama3.1:latest"
+  }
+}
+```
+
+### Gemini Configuration
+
+When you switch Claude Code to Gemini, the tool:
+
+1. Checks that LiteLLM is installed
+2. Prompts for a Google API key (if not already saved)
+3. Starts a LiteLLM proxy on port 4001 (if not already running)
+4. Writes these environment variables to `~/.claude/settings.json`:
+
+```json
+{
+  "env": {
+    "ANTHROPIC_AUTH_TOKEN": "YOUR_GEMINI_API_KEY",
+    "ANTHROPIC_BASE_URL": "http://localhost:4001",
+    "ANTHROPIC_MODEL": "gemini-2.5-pro",
+    "ANTHROPIC_DEFAULT_OPUS_MODEL": "gemini-2.5-pro",
+    "ANTHROPIC_DEFAULT_SONNET_MODEL": "gemini-2.5-flash",
+    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "gemini-2.5-flash-lite"
+  }
+}
+```
+
+### OpenRouter Configuration
+
+When you switch Claude Code to OpenRouter, the tool writes these environment variables to `~/.claude/settings.json`:
+
+```json
+{
+  "env": {
+    "ANTHROPIC_AUTH_TOKEN": "YOUR_OPENROUTER_API_KEY",
+    "ANTHROPIC_BASE_URL": "https://openrouter.ai/api/v1",
+    "ANTHROPIC_MODEL": "qwen/qwen3.6-plus:free",
+    "ANTHROPIC_DEFAULT_OPUS_MODEL": "qwen/qwen3.6-plus:free",
+    "ANTHROPIC_DEFAULT_SONNET_MODEL": "openrouter/free",
+    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "openrouter/free"
+  }
+}
+```
+
+### OpenCode Configuration
+
+When you run `claude-switch opencode add alibaba`, the tool writes the `bailian-coding-plan` provider to `~/.config/opencode/opencode.json`:
+
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "provider": {
+    "bailian-coding-plan": {
+      "npm": "@ai-sdk/anthropic",
+      "name": "Model Studio Coding Plan",
+      "options": {
+        "baseURL": "https://coding-intl.dashscope.aliyuncs.com/apps/anthropic/v1",
+        "apiKey": "YOUR_API_KEY"
+      },
+      "models": {
+        "qwen3.6-plus": { ... },
+        "qwen3-max-2026-01-23": { ... },
+        "qwen3-coder-next": { ... },
+        "qwen3-coder-plus": { ... },
+        "MiniMax-M2.5": { ... },
+        "glm-5": { ... },
+        "glm-4.7": { ... },
+        "kimi-k2.5": { ... }
+      }
+    }
+  }
+}
+```
+
+Running `claude-switch opencode remove alibaba` removes only the `bailian-coding-plan` provider, preserving any other providers you have configured.
+
+### GLM/Z.AI Configuration
+
+GLM uses the `@z_ai/coding-helper` package to manage its configuration. For Claude Code, the tool triggers `coding-helper auth reload claude` to apply GLM settings, plus sets the model tier aliases.
+
+For OpenCode, run `claude-switch opencode add glm` to add the GLM provider. This reads the auth credentials from Claude Code's settings (set by `coding-helper auth reload claude`) and writes the GLM provider config with all models to `~/.config/opencode/opencode.json`:
+
+```bash
+# Set up GLM auth in Claude Code first
+claude-switch glm
+
+# Then add to OpenCode
+claude-switch opencode add glm
+
+# Remove when done
+claude-switch opencode remove glm
+```
+
+## Example Output
+
+```bash
+$ claude-switch status
+
+=== Claude AI Switcher Status ===
+
+  Claude Code:
+    Provider: anthropic
+
+  OpenCode:
+    Provider: anthropic
+
+  API Key Verification:
+──────────────────────────────────────────────────
+    ○ alibaba      No key configured
+    ○ openrouter   No key configured
+    ○ anthropic    No key configured
+    ✓ glm          coding-helper installed
+──────────────────────────────────────────────────
+```
+
+```bash
+$ claude-switch alibaba qwen3.7-plus
+
+✓ Switched to: Alibaba Coding Plan
+────────────────────────────────────────────────────────────
+  Model: Qwen3.7-Plus
+  Context: 1M tokens
+  Endpoint: https://coding-intl.dashscope.aliyuncs.com/apps/anthropic
+  Capabilities: Text Generation, Deep Thinking, Visual Understanding
+  Most capable Qwen model with balanced performance, speed, and cost...
+
+  Claude model aliases:
+    ANTHROPIC_DEFAULT_OPUS_MODEL   → qwen3.7-plus
+    ANTHROPIC_DEFAULT_SONNET_MODEL → qwen3.6-plus
+    ANTHROPIC_DEFAULT_HAIKU_MODEL  → kimi-k2.5
+```
+
+```bash
+$ claude-switch opencode add alibaba
+
+✓ Added Alibaba Coding Plan provider to OpenCode
+  Config: ~/.config/opencode/opencode.json
+  Provider: bailian-coding-plan
+  Models: qwen3.7-plus, qwen3.6-plus, qwen3-max-2026-01-23, qwen3-coder-next, qwen3-coder-plus, MiniMax-M2.5, glm-5, glm-4.7, kimi-k2.5
+```
+
+```bash
+$ claude-switch glm --opus glm-5.2[1m] --sonnet glm-5-turbo --haiku glm-5v-turbo
+
+✓ Switched to GLM/Z.AI
+
+  Claude model aliases:
+    ANTHROPIC_DEFAULT_OPUS_MODEL   → glm-5.2[1m]
+    ANTHROPIC_DEFAULT_SONNET_MODEL → glm-5-turbo
+    ANTHROPIC_DEFAULT_HAIKU_MODEL  → glm-5v-turbo
+```
+
+```bash
+$ claude-switch openrouter
+
+✓ Switched to: OpenRouter
+────────────────────────────────────────────────────────────
+  Model: Qwen3.6 Plus (Free)
+  Context: 131K tokens
+  Endpoint: https://openrouter.ai/api/v1
+  Capabilities: Text Generation, Deep Thinking
+  Free Qwen3.6 Plus model via OpenRouter with strong reasoning capabilities.
+
+  Claude model aliases:
+    ANTHROPIC_DEFAULT_OPUS_MODEL   → qwen/qwen3.6-plus:free
+    ANTHROPIC_DEFAULT_SONNET_MODEL → openrouter/free
+    ANTHROPIC_DEFAULT_HAIKU_MODEL  → openrouter/free
+```
+
+```bash
+$ claude-switch ollama deepseek-r1:latest
+
+✓ Switched to: Ollama (Local)
+────────────────────────────────────────────────────────────
+  Model: DeepSeek R1
+  Context: 128K tokens
+  Endpoint: http://localhost:4000 (LiteLLM proxy)
+  Capabilities: Text Generation, Deep Thinking, Reasoning
+  DeepSeek's reasoning model with deep thinking capabilities.
+
+  Claude model aliases:
+    ANTHROPIC_DEFAULT_OPUS_MODEL   → deepseek-r1:latest
+    ANTHROPIC_DEFAULT_SONNET_MODEL → qwen2.5-coder:latest
+    ANTHROPIC_DEFAULT_HAIKU_MODEL  → llama3.1:latest
+```
+
+```bash
+$ claude-switch gemini
+
+✓ Switched to: Gemini (Google)
+────────────────────────────────────────────────────────────
+  Model: Gemini 2.5 Pro
+  Context: 1M tokens
+  Endpoint: http://localhost:4001 (LiteLLM proxy)
+  Capabilities: Text Generation, Deep Thinking, Code, Vision
+  Google's most capable Gemini model with deep thinking.
+
+  Claude model aliases:
+    ANTHROPIC_DEFAULT_OPUS_MODEL   → gemini-2.5-pro
+    ANTHROPIC_DEFAULT_SONNET_MODEL → gemini-2.5-flash
+    ANTHROPIC_DEFAULT_HAIKU_MODEL  → gemini-2.5-flash-lite
+```
+
+## Project Documentation (Zread)
+
+This repository includes an AI-generated project wiki in `.zread/` created with [Zread CLI](https://zread.ai/cli). It provides detailed architectural walkthroughs, module breakdowns, and implementation guides.
+
+### Viewing the Wiki
+
+```bash
+# Install Zread CLI
+npm install -g zread_cli
+
+# Browse the wiki in your browser
+zread browse
+
+# Regenerate after code changes
+zread generate
+```
+
+The wiki is versioned alongside the code. Contributors should prune old version snapshots before committing so the repo only contains the latest.
+
+### Contributing to the Wiki
+
+After making significant code changes:
+
+1. Run `zread generate` to create an updated wiki
+2. Delete old dated folders under `.zread/wiki/versions/`, keeping only the latest
+3. Commit the updated `current` pointer and the new version folder
+
+## Requirements
+
+- Node.js >= 18.0.0
+- Claude Code installed (for Claude Code support)
+- OpenCode installed (for OpenCode helper commands)
+- Alibaba API Key (for Alibaba Coding Plan)
+- `coding-helper` package (for GLM/Z.AI support)
+- [LiteLLM](https://github.com/BerriAI/litellm) (for Ollama and Gemini support)
+- [Ollama](https://ollama.com) (for Ollama local models)
+- Google API Key (for Gemini)
+
+## LiteLLM Setup
+
+Ollama and Gemini require [LiteLLM](https://github.com/BerriAI/litellm) to translate between Claude Code's Anthropic API format and the OpenAI-compatible format used by these providers.
+
+```bash
+# Install LiteLLM with proxy support
+pip install 'litellm[proxy]'
+```
+
+The tool automatically starts the proxy when you switch to Ollama (port 4000) or Gemini (port 4001). No manual proxy management is needed.
+
+## Getting API Keys
+
+### Alibaba Coding Plan
+
+1. Visit [Alibaba Cloud Model Studio](https://modelstudio.console.alibabacloud.com/)
+2. Navigate to API Key management
+3. Create a new API key
+4. Run `claude-switch setup` or `claude-switch key alibaba <key>`
+
+### GLM/Z.AI
+
+1. Install coding-helper: `npm install -g @z_ai/coding-helper`
+2. Run: `coding-helper auth`
+3. Follow the interactive setup
+
+### OpenRouter
+
+1. Visit [OpenRouter Keys](https://openrouter.ai/settings/keys)
+2. Create a new API key
+3. Run `claude-switch setup` or `claude-switch key openrouter <key>`
+
+### Gemini (Google)
+
+1. Visit [Google AI Studio](https://aistudio.google.com/apikey)
+2. Create a new API key
+3. Run `claude-switch setup` or `claude-switch key gemini <key>`
+4. Ensure LiteLLM is installed: `pip install 'litellm[proxy]'`
+
+### Ollama (Local)
+
+1. Install [Ollama](https://ollama.com)
+2. Start Ollama: `ollama serve`
+3. Pull a model: `ollama pull deepseek-r1`
+4. Ensure LiteLLM is installed: `pip install 'litellm[proxy]'`
+5. No API key needed — everything runs locally
+
+## Safety Features
+
+- Checks if config files exist before creating
+- Backs up existing settings before modification
+- Never overwrites without confirmation
+- Auto-sets `hasCompletedOnboarding: true` in `~/.claude.json`
+- Local-only storage (no cloud sync)
+- Clears model alias env vars when switching between providers
+- Preserves other OpenCode providers when adding/removing bailian-coding-plan
+
+## Troubleshooting
+
+### "coding-helper not found"
+
+```bash
+npm install -g @z_ai/coding-helper
+```
+
+### "LiteLLM is required for Ollama/Gemini support"
+
+```bash
+pip install 'litellm[proxy]'
+```
+
+### "Ollama is not installed"
+
+Download and install from [ollama.com](https://ollama.com), then start it:
+
+```bash
+ollama serve
+```
+
+### "Ollama is not running"
+
+```bash
+ollama serve
+```
+
+### "Failed to start LiteLLM proxy"
+
+The proxy may take a moment to start. Check that port 4000 (Ollama) or 4001 (Gemini) is not already in use:
+
+```bash
+# Check what's using the port
+lsof -i :4000   # macOS/Linux
+netstat -ano | findstr :4000   # Windows
+```
+
+### "API Key not found"
+
+```bash
+claude-switch setup
+# or
+claude-switch key alibaba <your-api-key>
+claude-switch key gemini <your-api-key>
+```
+
+### "Unable to connect to Anthropic services"
+
+The tool automatically sets `hasCompletedOnboarding: true` in `~/.claude.json`. Run any switch command to trigger this:
+
+```bash
+claude-switch anthropic
+```
+
+### Check current configuration
+
+```bash
+claude-switch status    # Shows config + verifies API keys
+claude-switch current   # Shows config only
+```
+
+## License
+
+MIT