feat(statusline): add multi-provider support with z.ai GLM auto-detection

Detect model provider from JSON input (glm-* prefix) and fetch rate
limits from the appropriate API. Add token counts and progress bar to
context display, per-tool breakdown for z.ai (Search/Web/ZRead), and
provider-aware caching to prevent stale cross-provider data.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: claude

statusline/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "statusline",
-  "version": "1.6.0",
-  "description": "Configurable status bar showing context usage, API rate limits (5h/7d for Anthropic, detailed metrics for z.ai GLM), git branch, and active tools. Supports 1/2/3 line layouts with smart hiding of empty values.",
+  "version": "1.7.0",
+  "description": "Multi-provider status bar with context health, API rate limits (Anthropic 5h/7d + z.ai GLM tokens/tools), git branch, and active tools. Auto-detects provider from model name. Supports 1/2/3 line layouts with smart hiding.",
   "author": {
     "name": "duyet"
   }

statusline/README.md

@@ -1,20 +1,21 @@
 # Statusline
 
-A Claude Code plugin that provides real-time visibility into your coding session. Track context usage, active tools, running agents, and task progress at a glance.
+A Claude Code plugin that provides real-time visibility into your coding session. Track context usage, API rate limits, active tools, running agents, and task progress at a glance.
 
-**Inspired by [claude-hud](https://github.com/jarrodwatts/claude-hud) — enhanced for the claude-plugins ecosystem.**
+**Multi-provider support:** Automatically detects Anthropic vs z.ai GLM models and shows provider-specific metrics.
 
 ## Overview
 
-Statusline keeps you informed about what's happening during your Claude Code session without interrupting your workflow. Get immediate feedback on:
+Statusline keeps you informed about what's happening during your Claude Code session without interrupting your workflow.
 
-| Metric | Purpose |
-|--------|---------|
-| **Context Health** | Know exactly how much of your token window is in use |
-| **Active Tools** | See which tools Claude is using right now |
-| **Agent Tracking** | Monitor running agents and their execution time |
-| **Task Progress** | Track todo completion in real-time |
-| **API Rate Limits** | Monitor usage across Anthropic (5h/7d) or z.ai GLM (5h) |
+| Metric | Anthropic | z.ai GLM |
+|--------|-----------|----------|
+| **Context Health** | Token count, percentage, progress bar | Token count, percentage, progress bar |
+| **Rate Limits** | 5h utilization, 7d utilization | Token quota %, monthly tools % |
+| **Tool Breakdown** | — | Search, Web, ZRead usage counts |
+| **Active Tools** | MCP server detection | MCP server detection |
+| **Agent Tracking** | Running agents and count | Running agents and count |
+| **Task Progress** | Todo completion tracking | Todo completion tracking |
 
 ## Installation
 
@@ -24,211 +25,105 @@ Add the plugin to your Claude Code instance:
 /plugin install statusline
 ```
 
-Then enable monitoring:
-
-```bash
-/statusline:enable
-```
-
-## Quick Start
-
-### View Current Status
-Check your session metrics anytime:
-```bash
-/statusline:status
-```
-
-### Continuous Monitoring
-Enable automatic status updates (5-second interval by default):
-```bash
-/statusline:enable
-```
-
-### Custom Update Interval
-Set a custom monitoring interval (in seconds):
-```bash
-/statusline:enable 10
-```
-
-### Stop Monitoring
-Pause background updates while keeping the plugin active:
-```bash
-/statusline:disable
-```
-
 ## Commands
 
 | Command | Purpose |
 |---------|---------|
-| `/statusline:status` | Display current session metrics immediately |
-| `/statusline:enable [interval]` | Enable real-time monitoring with optional custom interval |
+| `/statusline:status` | Display current session metrics |
+| `/statusline:config` | Configure display format (1/2/3 lines) |
+| `/statusline:setup` | Interactive setup wizard |
 | `/statusline:disable` | Pause automatic status updates |
 
-## What Each Metric Shows
+## Provider Auto-Detection
 
-### Context Window
-A visual progress bar showing your token usage with color coding:
-- 🟢 Green: 0-60% — Comfortable usage
-- 🟡 Yellow: 60-85% — Getting full
-- 🔴 Red: 85%+ — Approaching limit
+The plugin automatically detects your model provider:
 
-Example:
-```
-Context: ████░░░░░░ 45% (90,000 / 200,000 tokens)
-```
-
-### Compact One-Line Format
-All metrics displayed in a single line with empty values hidden:
-```
-📊 🟡 45% | Model: Opus 4.5 | 12m 34s | Tools: Glob×4 Bash×8 Read×12 | Agents: Explore(12s) | Tasks: 🔄 1 ⏳ 2 ✓ 5 | Context: 3 prompts
-```
-
-**Smart hiding:**
-- ✓ No "None" values displayed
-- ✓ Empty collections omitted (no agents/tools if zero)
-- ✓ Claude Code version hidden
-- ✓ Context shows matching system prompts and tools
+- **Anthropic** — Any Claude model (claude-opus, claude-sonnet, claude-haiku, etc.)
+- **z.ai GLM** — Any model starting with `glm-` (glm-5.1, glm-4, glm-flash, etc.)
 
-### API Rate Limits
+Detection happens from the model ID injected by Claude Code into the statusline JSON input.
 
-The plugin automatically detects your model provider and shows appropriate usage metrics:
+## Display Formats
 
-**Anthropic Models** (default, Claude, Opus, Sonnet, etc.):
-```
-5h: 42% | 7d: 28%
-```
-- Shows 5-hour and 7-day utilization percentages
-- Requires Claude Code OAuth credentials
+### 1-line (Compact)
 
-**z.ai GLM Models** (glm-4, glm-4.7, glm-flash, glm-plus, etc.):
+Everything on a single line:
 ```
-z.ai: 15%
+claude-plugins (master) │ GLM 5.1 │ ctx: 21% 43k/200k │ tokens: 15% │ tools: 11% (883 left)
 ```
-- Shows token quota percentage (5h equivalent)
-- Requires z.ai API key from opencode auth file
-
-#### z.ai Setup
-
-For GLM models, the plugin tries multiple credential sources in order:
-
-**1. macOS Keychain** (macOS only - highest priority)
-- Looks for keychain entries named: `z.ai`, `zai`, `opencode`, or `zai-coding-plan`
-- Parses JSON or uses raw value as API key
 
-**2. Environment Variables** (cross-platform)
-- `ZAI_API_KEY` - Generic z.ai API key
-- `ZAI_CODING_PLAN_KEY` - Coding plan specific key
+### 2-line (Default)
 
-**3. Configuration Files** (cross-platform fallback)
-Tries these paths in order:
+Location + model on line 1, all metrics on line 2:
 ```
-~/.local/share/opencode/auth.json    # Linux/Unix XDG standard
-~/.config/opencode/auth.json         # Alternative XDG location
-~/.opencode/auth.json                # Legacy location
-~/.zai/auth.json                     # z.ai specific
+claude-plugins (master) │ GLM 5.1 │ v2.0.76
+ctx: 21% 43k/200k │ tokens: 15% │ tools: 11% (883 left) Web:35 ZRead:82
 ```
 
-Example auth file structure:
-```json
-{
-  "zai-coding-plan": {
-    "key": "your-zai-api-key-here"
-  }
-}
-```
+### 3-line (Detailed)
 
-Alternative JSON structures supported:
-```json
-{"zai": {"key": "..."}}
-{"apiKey": "..."}
-{"key": "..."}
+Full layout with progress bar and token details:
+```
+claude-plugins (master) │ GLM 5.1 │ v2.0.76 │ explanatory
+ctx: ██░░░░░░░░ 21% (43k/200k tokens) │ tokens: 15% │ tools: 11% (883 left) Web:35 ZRead:82
+Tools: Seq Ctx7 │ Agents: 2 active │ Tasks: 3/5 (60%)
 ```
 
-**Recommended Setup by Platform:**
-
-| Platform | Recommended Method |
-|----------|-------------------|
-| macOS | Keychain: `security add-generic-password -s "z.ai" -w "your-key"` |
-| Linux | File: `~/.local/share/opencode/auth.json` |
-| Windows | File: `%USERPROFILE%\.local\share\opencode\auth.json` |
-
-The provider is automatically detected from the `CLAUDE_MODEL` or `ANTHROPIC_MODEL` environment variable. Any model starting with `glm-` will use the z.ai API.
+## Context Health
 
-## Features
+Color-coded progress indicator:
+- 🟢 Green: 0-60% — Comfortable
+- 🟡 Yellow: 60-85% — Getting full
+- 🔴 Red: 85%+ — Approaching limit
 
-✅ **Real-time Updates** — Status refreshes automatically at your chosen interval
+The 3-line format includes a visual progress bar: `███░░░░░░░ 30%`
 
-✅ **Non-intrusive** — Displays without interrupting your workflow
+## API Rate Limits
 
-✅ **Customizable** — Set monitoring frequency that works for you
+### Anthropic
 
-✅ **Manual Control** — Check status anytime with `/statusline:status`
+Shows 5-hour and 7-day utilization percentages:
+```
+5h: 42% │ 7d: 28%
+```
 
-✅ **Plugin Integration** — Works seamlessly with other Claude Code plugins
+Requires Claude Code OAuth credentials (auto-detected from keychain or config files).
 
-✅ **Token Awareness** — Shows actual token counts from Claude Code
+### z.ai GLM
 
-## Use Cases
+Shows token quota and monthly tool usage:
+```
+tokens: 15% │ tools: 11% (883 left) Web:35 ZRead:82
+```
 
-- **Monitor Context** — Know when you're approaching context limits before running out
-- **Track Progress** — See how many tasks have been completed in your session
-- **Debug Workflows** — Understand tool and agent activity during complex operations
-- **Optimize Sessions** — Identify bottlenecks and adjust your approach
-- **Stay Informed** — Keep awareness of background operations
+- **tokens** — 5-hour rolling window (equivalent to Anthropic's 5h)
+- **tools** — Monthly tool quota with remaining count
+- **Per-tool breakdown** — Search, Web-reader, ZRead usage
 
-## How It Works
+#### z.ai Credential Sources (tried in order)
 
-The plugin reads session data from Claude Code's internal state and displays aggregated metrics about:
+1. **macOS Keychain** — `security add-generic-password -s "z.ai" -w "your-key"`
+2. **Environment** — `ZAI_API_KEY` or `ZAI_CODING_PLAN_KEY`
+3. **Config files** — `~/.local/share/opencode/auth.json` or `~/.zai/auth.json`
 
-1. **Token Usage** — Real counts from Claude Code's context window
-2. **Tool Activity** — Tracked from tool invocations in your session
-3. **Agent Operations** — Monitored from subagent execution
-4. **Task Management** — Read from your TodoWrite operations
+## Smart Hiding
 
-Updates happen at a regular interval (default 5 seconds) without blocking your work.
+Empty values are never shown:
+- No "None", "0%", or "No tasks" clutter
+- Tool/agent sections hidden when inactive
+- Per-tool breakdown only shown when usage > 0
+- Output style hidden when set to "default"
 
 ## Requirements
 
 - Claude Code v1.0.80+
-- Node.js 18+ or Bun (if running custom build)
-
-## Inspiration
-
-This plugin is inspired by [claude-hud](https://github.com/jarrodwatts/claude-hud) but optimized for the claude-plugins marketplace with enhanced integration and customization options.
-
-## Development
-
-Session Monitor is built with:
-- TypeScript for type safety
-- Claude Code's native APIs
-- Minimal dependencies for quick performance
-
-### Source Structure
-
-```
-session-monitor/
-├── .claude-plugin/
-│   └── plugin.json         # Plugin manifest
-├── commands/
-│   ├── status.md          # Status command documentation
-│   ├── enable.md          # Enable command documentation
-│   └── disable.md         # Disable command documentation
-└── README.md              # This file
-```
+- Node.js 18+ (for TypeScript CLI)
+- `jq` (for Bash statusline script)
 
 ## Contributing
 
-Found an issue or have a feature idea? Contributions are welcome! This plugin is part of the [claude-plugins](https://github.com/duyet/claude-plugins) marketplace.
+Found an issue or have a feature idea? Open an issue on the [claude-plugins repository](https://github.com/duyet/claude-plugins/issues).
 
 ## License
 
-MIT License — see LICENSE file
-
-## Related Plugins
-
-- **[team-agents](../team-agents)** — Coordinated parallel task execution
-- **[commit](../commit)** — Semantic Git commit automation
-
----
-
-**Questions?** Open an issue on the [claude-plugins repository](https://github.com/duyet/claude-plugins/issues)
+MIT License

statusline/commands/config.md

@@ -14,9 +14,9 @@ Ask the user which display format they prefer:
 **Question**: "How many lines should the statusline display?"
 
 **Options**:
-1. **1 line (Compact)** — `dir (branch) │ Model │ Ctx% │ 5h/7d │ Tasks`
+1. **1 line (Compact)** — `dir (branch) │ Model │ Ctx% │ Rate │ Tasks`
 2. **2 lines** — Line 1: location/model, Line 2: all metrics
-3. **3 lines (Default)** — Full layout with separate lines for location, metrics, and activity
+3. **3 lines (Default)** — Full layout with progress bar, token details, and activity
 
 ### Save Configuration
 
@@ -40,28 +40,50 @@ Output:
 
 ## Example Outputs
 
-**1-line format**:
+**Anthropic** (Claude, Opus, Sonnet):
+
+1-line:
+```
+claude-plugins (master) │ Opus 4.5 │ ctx: 21% 43k/200k │ 5h: 42% 7d: 28%
+```
+
+2-line:
+```
+claude-plugins (master) │ Opus 4.5 │ v2.0.76
+ctx: 21% 43k/200k │ 5h: 42% │ 7d: 28% │ Tasks: 2/5 (40%)
+```
+
+3-line:
+```
+claude-plugins (master) │ Opus 4.5 │ v2.0.76 │ explanatory
+ctx: ██░░░░░░░░ 21% (43k/200k tokens) │ 5h: 42% │ 7d: 28%
+Tools: Seq Ctx7 │ Agents: 2 active │ Tasks: 1/5 (20%)
+```
+
+**z.ai GLM** (glm-5.1, glm-4, etc.):
+
+1-line:
 ```
-claude-plugins (master) │ Opus 4 5 │ Ctx: 21% │ 5h: 14% 7d: 46%
+claude-plugins (master) │ GLM 5.1 │ ctx: 21% 43k/200k │ tokens: 15% │ tools: 11% (883 left)
 ```
 
-**2-line format**:
+2-line:
 ```
-claude-plugins (master) │ Opus 4 5 │ v2.0.76
-Ctx: 21% │ 5h: 14% 7d: 46% │ Tasks: 2/5
+claude-plugins (master) │ GLM 5.1 │ v2.0.76
+ctx: 21% 43k/200k │ tokens: 15% │ tools: 11% (883 left) Web:35 ZRead:82
 ```
 
-**3-line format** (default):
+3-line:
 ```
-claude-plugins (master) │ Opus 4 5 │ v2.0.76
-Context: 21% (43,000 tokens) │ 5h: 14% 7d: 46%
-Tools: Sequential │ Tasks: 2/5
+claude-plugins (master) │ GLM 5.1 │ v2.0.76 │ explanatory
+ctx: ██░░░░░░░░ 21% (43k/200k tokens) │ tokens: 15% │ tools: 11% (883 left) Web:35 ZRead:82
+Tools: Seq Ctx7 │ Agents: 1 active
 ```
 
 **Note**: Empty values are automatically hidden (no "None", no "0%", no "No tasks").
+Provider is auto-detected from the model name.
 
 ## Related Commands
 
 - `/statusline:status` — View current metrics
-- `/statusline:enable` — Enable monitoring
 - `/statusline:disable` — Disable monitoring