Skip to content

CHANGELOG.md

Latest commit

 

History

History
108 lines (83 loc) · 4.97 KB

CHANGELOG.md

File metadata and controls

108 lines (83 loc) · 4.97 KB
Raw

Changelog

All notable changes to Andor will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

[0.2.1] - 2026-03-10

Changed

  • Version bump — incremented to 0.2.1 to resolve marketplace version conflict

[0.2.0] - 2026-03-10

Added — Agent Execution Improvements

  • Structured sub-agent workflow — internal agents now respond with explicit Analysis, Plan, Execution, Verification, Blockers, and Completion sections
  • Completion state handling — tasks can now terminate more cleanly as COMPLETE, INCOMPLETE, or BLOCKED
  • Progress summaries — sub-agents report more meaningful progress updates during execution

Added — Orchestration Improvements

  • Stronger complexity detection — better routing of complex requests into orchestrated multi-agent execution
  • Higher-quality planning prompt — planner now prefers clearer decomposition, better role selection, stronger validation, and less redundant work
  • Structured orchestrator summary — multi-agent runs now produce organized plan execution, verification, and blocker summaries

Changed

  • Sub-agent prompting now emphasizes senior-engineer behavior, root-cause analysis, explicit blockers, and verification-first outputs
  • Failure handling for sub-agents now returns clearer errors after retry exhaustion
  • README updated to document the new agent execution model and release packaging workflow

[0.1.0] - 2025-03-07

Added — Multi-Provider Architecture

  • Provider system — New src/providers/ module with base interfaces, provider registry, and smart fallback logic
  • 6 AI providers — Puter.js (default, free), NVIDIA NIM, Groq, Google Gemini, Mistral AI, OpenRouter
  • 70+ models available across all providers, with tier badges (fast/balanced/powerful) and FREE indicators
  • Secure API key storage using VS Code SecretStorage API (andor.apikey.<providerId>)
  • Provider registry with automatic model resolution, connection testing, and model caching
  • Smart fallback — automatic retry with alternative providers on rate limit (429) or server errors (5xx)
  • Extension-host streaming — non-Puter providers stream directly from the extension host via SSE parsing
  • Dynamic model fetching for OpenRouter (fetches available models from API)

Added — Settings Panel

  • Provider cards showing connection status (configured/untested/unconfigured), model count, and API key management
  • API key management — add, delete, test connection, with links to provider key pages
  • Command allowlist editor for managing auto-approved terminal command patterns

Added — UI Improvements

  • Grouped model selector with search, provider sections, tier icons, FREE/context badges
  • Command approval dialog — Allow Once, Always Allow (with pattern), Deny actions for terminal commands
  • Auto-allowed commands — safe commands (git status, npm run build, tsc, etc.) run without approval
  • Allowlist file at .vscode/andor-allowlist.json with glob pattern matching
  • File operation cards — inline display for modified/created/deleted files with diff preview
  • Terminal output component — collapsible inline terminal output with success/failure styling
  • Settings button in header toolbar

Changed

  • Model selector rewritten from simple dropdown to searchable grouped dropdown with provider context
  • ChatPanel updated to route Puter models through webview and non-Puter models through extension host
  • WebviewBridge extended with provider, command approval, and streaming message handlers
  • System prompt enhanced with coding principles, response format guidelines, and security rules
  • Types expanded with ProviderInfo, ModelInfo, CommandApprovalRequest, AllowlistFile interfaces
  • Version bumped to 0.1.0

Previous (0.0.x)

  • Multiple AI model support (Claude Sonnet 4, Claude Opus 4, GPT-4o, Gemini 2.5 Pro, DeepSeek V3, Llama 4, etc.)
  • Free tier model indicators
  • Checkpoint/revert system for conversation states
  • Prompt editing capability for previous messages
  • Task completion detection with "Continue" button
  • Advanced context assembly with workspace file listing
  • Enhanced error handling for insufficient funds
  • Modern Windsurf-like UI redesign
  • Sticky headers and improved UX
  • File reading capability for AI agent

[0.0.1] - 2025-03-06

Added

  • Initial release of Andor (formerly PuterCoder)
  • Basic AI chat functionality with Puter.js integration
  • File writing and terminal command execution
  • Context file inclusion for AI prompts
  • VS Code diagnostics awareness
  • Authentication with Puter.com
  • Basic UI with model selection
  • Webview-based chat interface

[Previous Versions]

PuterCoder 0.0.x

  • Initial development phase
  • Basic chat functionality
  • File operations support
  • Terminal integration