Skip to content

CHANGELOG.md

Latest commit

 

History

History
209 lines (172 loc) · 10.2 KB

CHANGELOG.md

File metadata and controls

209 lines (172 loc) · 10.2 KB

Changelog

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

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

2.3.0 - 2026-03-03

Added

  • Channel Monitoring (#17): Watch a Discord channel for bot/webhook messages and auto-trigger a Claude investigation in a thread. Messages are batched over a 30-second debounce window. Requires MONITOR_CHANNEL_ID and MONITOR_BOT_IDS env vars and the Message Content privileged intent.
  • Thread-per-Session (#19): New /claude-thread command starts each conversation in a dedicated Discord thread with a custom name.
  • Per-Channel Session Tracking: Regular /claude commands in the same channel automatically reuse the last session — no need to pass session_id manually.
  • Custom Thread Names: /claude-thread name:"Fix auth bug" prompt:"Review the auth module" creates a thread titled "Fix auth bug".
  • Thread-Aware Channel Detection: Commands work correctly inside session threads (bot recognizes the parent channel).

Changed

  • discord/session-threads.ts added — SessionThreadManager handles thread creation, lookup, and lifecycle
  • discord/types.ts — InteractionContext now exposes getChannelId() for per-channel session mapping
  • core/handler-registry.ts — channelSessionMap maintains channel → session mapping
  • .env.example updated with Channel Monitoring section

2.2.0 - 2025-07-18

Added

  • Version in /status: Shows bot version (e.g. v2.2.0) and update status in the /status embed
  • Periodic Update Checks: Automatically checks for updates every 12 hours and notifies in Discord
  • Semver in Startup Embed: Startup message now shows version number
  • BOT_VERSION Export: deno.json version read at startup and available throughout codebase

Changed

  • Dockerfile Optimized: Removed unnecessary Node.js and @anthropic-ai/claude-code CLI install — bot uses SDK directly via Deno imports, reducing image size

Fixed

  • deno.json version bumped from 1.0.0 to 2.2.0 (was not updated for v2.1.0)
  • CHANGELOG legacy reference to deprecated CLI corrected to @anthropic-ai/claude-agent-sdk

[2.1.0] - 2025-07-17

Added

  • Interactive Permission Requests: Allow/Deny buttons in Discord when Claude wants to use an unapproved tool (replaces CLI TUI prompt, makes default and acceptEdits modes fully usable)
  • Settings Autocomplete: /settings action and value fields now show dropdown suggestions based on selected category
  • Fast Mode (/fast): Toggle Opus 4.6 speed-optimized API config (2.5x faster, same quality)
  • AskUserQuestion Handler: Claude can now ask clarifying questions mid-session via Discord buttons
  • Startup Buttons: Quick-action buttons on the startup embed (Status, Sessions, Help, Shutdown)
  • MCP Mid-Session Management (/mcp toggle, /mcp reconnect): Enable/disable or reconnect MCP servers without restarting
  • Granular Sandbox Config: Full SDK SandboxSettings support (network rules, filesystem ACLs, excluded commands, violation ignoring)
  • Additional Directories: Multi-repo access via additionalDirectories setting
  • Fork Session: Branch conversations into new sessions via forkSession option
  • Hooks System: Passive SDK callbacks for tool use, notification, and task completion observability
  • stopTask(): Stop background tasks mid-session via /claude-control action:stop-task
  • Agent Teams: CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 env var support with delegate permission mode
  • stop_reason Display: Shows why Claude stopped (end_turn, max_tokens, stop_sequence) in completion embeds
  • Tasks Env Var: CLAUDE_CODE_ENABLE_TASKS=1 automatically set for background task support
  • Known Issues Documentation: docs/known-issues.md documenting 22 accepted risks

Changed

  • /continue renamed to /resume for clarity
  • /claude-settings and /output-settings removed — consolidated into /settings (use category:claude or category:output)
  • Button UX overhaul: consistent styling, proper disabled states, contextual labels
  • Settings display now shows SDK features (hooks, agent teams, sandbox config, additional dirs)
  • Help system updated with all new command documentation

Fixed

  • Getter pattern for claudeController prevents stale abort controller references
  • Abort state checks prevent sending to cancelled sessions
  • Continue button properly resumes the last session
  • Pagination titles show correct page info
  • Unicode-safe message splitting prevents mid-codepoint truncation

2.0.0 - 2026-02-18

Breaking Changes

  • Migrated from deprecated @anthropic-ai/claude-code to @anthropic-ai/claude-agent-sdk v0.2.45
  • Default model no longer hardcoded; SDK auto-selects the best available model

Added

  • Mid-Session Controls (/claude-control): interrupt, change model, change permissions without restarting
  • File Rewind (/rewind): undo file changes to any conversation turn with dry-run preview
  • Structured Output: force JSON responses matching a configurable schema
  • Info Commands (/claude-info): view account info, available models, MCP server status
  • Dynamic Model Discovery: auto-fetches models from Anthropic API and CLI at startup
  • 1M Token Context Beta: opt-in extended context window
  • File Checkpointing: enables rewind support for file changes
  • Sandbox Mode: run Claude in a sandboxed environment
  • Startup Version Check: compares local commit vs GitHub, sends Discord notification if behind
  • GHCR Docker Publishing: GitHub Actions workflow builds and pushes images on every push to main
  • Watchtower Support: auto-update Docker containers when new images are published
  • DRY Documentation: README slimmed to hub, 7 focused doc files in /docs
  • OS-Specific Install Guides: separate instructions for Linux/macOS, Windows, Docker

Changed

  • SDK integration uses AsyncGenerator<SDKMessage> streaming instead of CLI subprocess
  • Agents converted to native SDK AgentDefinition format
  • Normal permission mode uses acceptEdits instead of default
  • All settings (thinking, effort, system prompt, permissions, git context) now wired to SDK
  • Dockerfile includes Node.js 20 and Claude CLI
  • ANTHROPIC_API_KEY passthrough added to Docker Compose

Fixed

  • Duplicate "Claude Code Complete" embeds per query
  • Model default causing rate limit fallback to haiku
  • Settings not being passed through to SDK queries

1.0.0 - 2026-02-05

🎉 First Public Release

This is the first stable release of Claude Code Discord Bot - a Discord bot that brings Claude AI capabilities to your Discord server.

Features

Core Commands

  • /claude - Chat with Claude AI with thinking modes
  • /continue - Continue previous conversations
  • /claude-cancel - Cancel ongoing requests

Enhanced Claude

  • /claude-enhanced - Advanced Claude with model selection
  • /claude-models - List available Claude models
  • /claude-sessions - Manage conversation sessions
  • /claude-context - View and manage context

Development Tools

  • /claude-explain - Get code explanations
  • /claude-debug - Debug code issues
  • /claude-optimize - Optimize code performance
  • /claude-review - Code review assistant
  • /claude-generate - Generate code snippets
  • /claude-refactor - Refactor code
  • /claude-learn - Learn programming concepts

Task Management

  • /todos - Task management with priorities and persistence
  • /mcp - Model Context Protocol server management (reads from .claude/mcp.json)
  • /agent - 7 specialized AI agents

Settings

  • /settings - Unified settings management
  • /claude-settings - Claude-specific settings
  • /output-settings - Output formatting options
  • /quick-model - Quick model switching

Git Operations

  • /git - Execute git commands
  • /worktree - Create git worktrees
  • /worktree-list - List all worktrees
  • /worktree-remove - Remove worktrees
  • /worktree-bots - Manage worktree bots
  • /worktree-kill - Kill worktree processes

Shell Management

  • /shell - Execute shell commands
  • /shell-input - Send input to running processes
  • /shell-list - List running processes
  • /shell-kill - Kill processes

System Monitoring

  • /system-info - System information
  • /processes - Process listing
  • /system-resources - Resource usage
  • /network-info - Network information
  • /disk-usage - Disk space info
  • /env-vars - Environment variables
  • /system-logs - System logs
  • /port-scan - Port scanning
  • /service-status - Service status
  • /uptime - System uptime
  • /screenshot - Capture screen (Windows/macOS/Linux GUI)

Utilities

  • /status - Bot status
  • /pwd - Current directory
  • /shutdown - Graceful shutdown
  • /help - Command help

AI Agents

7 specialized AI agents for different development tasks:

  • Code Reviewer - Quality analysis and security review
  • Software Architect - System design and architecture
  • Debug Specialist - Bug analysis and troubleshooting
  • Security Analyst - Vulnerability assessment
  • Performance Engineer - Optimization and profiling
  • DevOps Engineer - Deployment and infrastructure
  • General Assistant - Multi-purpose development help

Infrastructure

  • Docker Support - Production-ready Dockerfile and docker-compose.yml
  • Cross-Platform - Windows, macOS, and Linux support
  • Persistence - Todos, sessions, and settings persist across restarts
  • Branch-Aware - Automatic channel organization by git branch
  • Secure - Runs as non-root user in Docker, resource limits

MCP (Model Context Protocol) Support

  • Reads server configurations from standard .claude/mcp.json file
  • Add/remove/test/list MCP servers via Discord commands
  • Cross-platform command testing (Windows where / Unix which)

Technical Details

  • Built with Deno 2.x and TypeScript
  • Uses Discord.js 14.14.1
  • Claude API via @anthropic-ai/claude-agent-sdk