0.133.0

What's Changed

0.133.0 - 2026-06-20

Highlights

Bug Fixes

• Reduce mid-run stops and eliminate wasted duplicate tool calls (884b39b)
• Reduce mid-run stops and eliminate wasted duplicate tool calls (a925cb8)
• Silently ignore floating overlay Selection events when no palette is active (e67db4e)
• Stop constant writes of background_subagents.json (5cfc6c1)

Features

• Add Z.ai GLM Coding Plan with subscription details to README (593ba1b) (@vinhnx)
• Enhance verification handling and refactor tool execution methods (d4f9164)
• Implement duplicate tool call detection and enhance recovery handling (89e92b2)
• Enhance agent selection and permissions handling for planning workflow (5c372a6)
• Enhance planning workflow with real-time header updates and tool output handling (16c55cf)
• Improve planning confirmation flow and enhance tool output handling (49d6411)
• Enhance planning workflow with trust checks and tool behavior adjustments (be0b38e)
• Implement tool policy overrides and enhance denial diagnostics in planning workflow (0cf67cf)
• Merge session and subagent lifecycle events in primary hook configuration (ba26433)
• Prevent TUI interrupt from exiting the program and improve signal handling (8cac5ee)
• Enhance planning workflow with agent type handling and patch file extraction (25227d3)
• Deactivate planning workflow on agent switch and refine exit triggers (4e95879)
• Add mode command to switch active agent mode and implement mode palette (ef1b52b)
• Enhance agent mode handling and improve tool output formatting (02c5a15)
• Add kimi-k2.7-code, glm-5.2, minimax-m3 models (ce6c722)
• Improve signal handler responsiveness and refine tool action filtering (36da47e)
• Enhance Ctrl+C handling and add agent context to tool logging (474fcdf)
• Implement tool-free recovery mode switch and refactor related logic (a379f55)
• Enhance workspace detection and logging for safe-mode prompts (ed91aee)
• Add support for native inline compaction in LLMProvider and related providers (f8146fc)

Other Changes

Other

• Remove references to Nemotron 3 Super model from Ollama provider and related tests (e8a7077) (@vinhnx)
• Refactor documentation links and update references across multiple files (8838ff0) (@vinhnx)
• Update documentation for VT Code version 0.132.0, including new modules, features, and provider integrations. (1a2fe62) (@vinhnx)
• Revise VT Code description for clarity and conciseness (7a3b9ab) (@vinhnx)
• Rename sections for OpenAI and Anthropic in guide (f801db7) (@vinhnx)
• Add global read-only budget limit and update harness limits prompt (7f96113)
• Implement subagent mode with tighter read-only budgets and navigation limits (e7c2f74)
• Update TODO (69b98ae)
• Update notes (8c0d97a)
• Implement adaptive budget recovery mechanism and related tool definitions (1f518af)
• Refactor manual compaction support and improve provider interface (29030ca)

Refactors

• Update tool references from grep_file and read_file to unified_search and unified_file across documentation (8739109) (@vinhnx)
• Clean up code formatting and improve readability in multiple files (f5b06c9)
• Streamline fallback handling and improve code readability in tool execution (a6b542f)
• Consolidate HTTP client and optimized client code, enhance command argument handling, and improve read-only command detection (6d3915d)
• Consolidate duplicated tool dispatch patterns and add fallback validation (0b37a84)

Full Changelog: 0.132.0...0.133.0