v1.5.62

v1.5.62 - Enable Dynamic Model Fetching for Deepseek/GLM Providers

🎉 What's New

This release enables dynamic model fetching for Deepseek and GLM providers, removing the previous bypass behavior for a more consistent experience across all providers.

🔧 Key Changes

• Dynamic Model Fetching: Deepseek and GLM providers now fetch their actual model lists dynamically via their OpenAI-compatible endpoints
• Better Error Handling: Improved error messages for unauthorized API keys (401/403 responses)
• Enhanced Testing: New comprehensive test suites for endpoint resolution and model list normalization
• Updated Documentation: README has been refreshed with more concise, modern information

🧪 Testing

All existing tests pass with the new behavior. New tests verify:

• Endpoint resolution for correct model API paths
• Model list normalization for different provider formats
• Dynamic fetching for Anthropic-native providers (Deepseek/GLM)

🚀 Installation

Install the VSIX from the release assets or update via VS Code extension marketplace.

📋 Technical Details

• Extension Version: 1.5.62
• Files: 828 files packaged (1.6MB)
• Breaking Changes: None - this is a feature enhancement
• Migration: Automatic - no user action required

v1.5.61

VS Code extension v1.5.61 test build.

• Settings UI restructured into logical sections
• Packaging + VSIX attached for testing
• Quick Combos: updated OpenRouter Power Combo models

Install: code --install-extension thronekeeper-1.5.61.vsix

Thronekeeper v1.5.55

Highlights

• Restore the Thronekeeper panel after development guardrail refactors
• Add schema coverage for custom endpoint overrides (updateEndpointKind)
• Bundle redaction helper inside the extension for reliable activation

v1.5.52 - Realtime Model List Improvements & CI Fixes

🎯 Key Features

• Realtime Model List Updates: Dynamic model fetching with provider-aware caching
• Race Condition Protection: Sequence tokens prevent stale model list rendering
• Provider Isolation: Clean separation between OpenRouter, GLM, Deepseek, and custom providers
• Improved Timeout Handling: 50-second model fetch budget with explicit classifications

🔧 Technical Improvements

• Schema Alignment: Root and extension schemas fully synchronized
• TypeScript Compilation: Fixed all CI TypeScript errors with proper module resolution
• Proxy Stability: Enhanced health check endpoints and startup reliability
• Zod Schema Validation: Proper discriminated union structure (v3.23.0)
• Secret Redaction: Improved utility for sanitizing API keys in logs
• Anthropic Provider Detection: Added anthropic.ai host recognition
• Branding Consistency: Unified to "Thronekeeper" across all UI elements

🐛 Bug Fixes

• Fixed proxy startup timeout issues in CI
• Resolved TypeScript discriminated union errors
• Corrected Anthropic provider detection for api.anthropic.ai
• Enhanced health check to support both /health and /healthz endpoints
• Fixed missing dependencies (undici, zod)

📦 Dependencies

• Added TypeScript ^5.6.3 as devDependency
• Zod v3.23.0 (downgraded from v4 for compatibility)
• undici ^6.22.0 for HTTP requests

🔐 Security

• Enhanced API key redaction patterns
• Improved credential storage with VS Code Secrets API

📝 Documentation

• Added CLAUDE.md with Constitution-guided workflow
• Updated CONSTITUTION.md with guarded areas
• Comprehensive test coverage documentation
• Area labels for PR tracking

---

Full Changelog: v1.5.45...v1.5.52

v1.5.45 - Schema Alignment & Timeout Improvements

Changes

• Schema Alignment: Root-level schemas now match extension schemas (structured error payloads, completion key requirement)
• Sequence Token Consistency: All cached model responses include sequence tokens for proper race condition protection
• Timeout Cap: Model fetching now has a 50-second overall budget with explicit timeout classification

Files Changed

• src/schemas/messages.ts - Schema alignment
• extensions/claude-throne/src/schemas/messages.ts - Extension schema updates
• extensions/claude-throne/src/services/Models.ts - Timeout budget implementation
• extensions/claude-throne/src/views/PanelViewProvider.ts - Sequence token consistency
• tests/contract.test.js - Updated test expectations

v1.5.17 - Native GLM Coding Plan and Deepseek Support

Removed Groq as a default provider and replaced it with GLM Coding Plan and Deepseek providers. These both use native Anthriopic API Endpoints so with these providers, or any custom Anthropic providers you supply, the Proxy is only handling Authentication and secret storage (via VSCode Secrets) so that you don't have to pollute your terminal ENV with additional API keys which in turn force Claude-Code to complain about having both oAuth and API credentials present simultaneously.

v0.1.16 – Panel-first config, xAI (Grok), detection + bypass

Highlights

• Panel-first configuration: provider, custom URL, and models in the Claude Throne panel
• xAI (Grok): correct API base https://api.x.ai, keys via XAI_API_KEY/GROK_API_KEY (legacy GROQ_API_KEY)
• Custom endpoint detection: OpenAI-style vs Anthropic-style, with one-click Bypass and Apply
• Inline provider key detection with Add/Update button under dropdown
• VS Code settings trimmed to port, debug, auto-apply, customEndpointKind

Changes

• ProxyManager/daemon updated for xAI base and env vars
• PanelViewProvider: detection & bypass UI, provider key status
• README updated to reflect current state; provider endpoints table

VSIX

• Packaged claude-throne-0.1.16.vsix