v1.3.0

What's Changed

• ci: enable Claude review workflow for fork PRs by @alexx-ftw in #53
• Reduce aggressive response truncation to preserve more content by @RichardAtCT in #54
• Replace query() with ClaudeSDKClient for native session management by @RichardAtCT in #56
• Remove CLI subprocess backend, pass disallowed_tools to SDK by @RichardAtCT in #59
• fix: handle unknown message types (e.g. rate_limit_event) gracefully by @lyra63237 in #72
• Improve error reporting with descriptive, type-aware error messages by @RichardAtCT in #74
• Update claude-agent-sdk from 0.1.38 to 0.1.39 by @RichardAtCT in #75
• Document available Claude Code tools with descriptions and configuration by @RichardAtCT in #76
• chore: bump poetry.lock dependencies by @RichardAtCT in #77
• chore: migrate pyproject.toml to PEP 621 format for Poetry 2.x by @RichardAtCT in #78
• Fix: prevent MessageParseError from killing SDK message stream by @RichardAtCT in #81
• feat: add version management and release workflow by @RichardAtCT in #82

New Contributors

• @lyra63237 made their first contribution in #72

Full Changelog: v1.2.0...v1.3.0

v1.2.0

What's New

• Remote Mac support: New make run-remote target for running the bot on a Mac accessed via SSH — handles macOS keychain unlock and tmux session management
• Python 3.11 minimum: Bumped minimum Python version from 3.10 to 3.11
• Setup docs: New "Running on a Remote Mac (SSH)" section covering keychain issues, workarounds, and tmux workflow

v1.1.0

What's New in v1.1.0

Added

• Tunable Verbose Output (#29):

  • VERBOSE_LEVEL setting (0-2) controls how much of Claude's background activity is shown during execution
  • Level 0: Quiet — only final response, typing indicator still active
  • Level 1: Normal — tool names + reasoning snippets shown during execution
  • Level 2: Detailed — tool names with input summaries + longer reasoning text
  • /verbose 0|1|2 command for per-session override
  • Persistent typing indicator refreshed every ~2 seconds at all levels

• HTML Parse Mode (#26):

  • Migrated from Markdown v1 to HTML parse mode for Telegram message formatting
  • More reliable rendering across Telegram clients

Changed

• Updated documentation to reflect agentic mode as default interaction model (#25)

Full Changelog: v1.0.0...v1.1.0

v1.0.0

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.

[Unreleased]

[1.0.0] - 2026-02-13

Added

• Agentic Mode (default interaction model):
  • MessageOrchestrator routes messages to agentic (3 commands) or classic (13 commands) handlers based on AGENTIC_MODE setting
  • Natural language conversation with Claude -- no terminal commands needed
  • Automatic session persistence per user/project directory
• Event-Driven Platform:
  • EventBus -- async pub/sub system with typed event subscriptions (UserMessage, Webhook, Scheduled, AgentResponse)
  • AgentHandler -- bridges events to ClaudeIntegration.run_command() for webhook and scheduled event processing
  • EventSecurityMiddleware -- validates events before handler processing
• Webhook API Server (FastAPI):
  • POST /webhooks/{provider} endpoint for GitHub, Notion, and generic providers
  • GitHub HMAC-SHA256 signature verification
  • Generic Bearer token authentication
  • Atomic deduplication via webhook_events table
  • Health check at GET /health
• Job Scheduler (APScheduler):
  • Cron-based job scheduling with persistent storage in scheduled_jobs table
  • Jobs publish ScheduledEvent to event bus on trigger
  • Add, remove, and list jobs programmatically
• Notification Service:
  • Subscribes to AgentResponseEvent for Telegram delivery
  • Per-chat rate limiting (1 msg/sec) to respect Telegram limits
  • Message splitting at 4096 char boundary
  • Broadcast to configurable default chat IDs
• Database Migration 3: scheduled_jobs and webhook_events tables, WAL mode enabled
• Automatic Session Resumption: Sessions are now automatically resumed per user+directory
  • SDK integration passes resume parameter to Claude Code for real session continuity
  • Session IDs extracted from Claude's ResultMessage instead of generated locally
  • /cd looks up and resumes existing sessions for the target directory
  • Auto-resume from SQLite database survives bot restarts
  • Graceful fallback to fresh session when resume fails
  • /new and /end are the only ways to explicitly clear session context

Recently Completed

Storage Layer Implementation (TODO-6) - 2025-06-06

• SQLite Database with Complete Schema:
  • 7 core tables: users, sessions, messages, tool_usage, audit_log, user_tokens, cost_tracking
  • Foreign key relationships and proper indexing for performance
  • Migration system with schema versioning and automatic upgrades
  • Connection pooling for efficient database resource management
• Repository Pattern Data Access Layer:
  • UserRepository, SessionRepository, MessageRepository, ToolUsageRepository
  • AuditLogRepository, CostTrackingRepository, AnalyticsRepository
• Persistent Session Management:
  • SQLiteSessionStorage replacing in-memory storage
  • Session persistence across bot restarts and deployments
• Analytics and Reporting System:
  • User dashboards with usage statistics and cost tracking
  • Admin dashboards with system-wide analytics

Telegram Bot Core (TODO-4) - 2025-06-06

• Complete Telegram bot with command routing, message parsing, inline keyboards
• Navigation commands: /cd, /ls, /pwd for directory management
• Session commands: /new, /continue, /status for Claude sessions
• File upload support, progress indicators, response formatting

Claude Code Integration (TODO-5) - 2025-06-06

• Async process execution with timeout handling
• Session state management and cross-conversation continuity
• Streaming JSON output parsing, tool call extraction
• Cost tracking and usage monitoring

Authentication & Security Framework (TODO-3) - 2025-06-05

• Multi-provider authentication (whitelist + token)
• Rate limiting with token bucket algorithm
• Input validation, path traversal prevention
• Security audit logging with risk assessment
• Bot middleware framework (auth, rate limit, security, burst protection)

[0.1.0] - 2025-06-05

Added

Project Foundation (TODO-1)

• Complete project structure with Poetry dependency management
• Exception hierarchy, structured logging, testing framework
• Code quality tools: Black, isort, flake8, mypy with strict settings

Configuration System (TODO-2)

• Pydantic Settings v2 with environment variable loading
• Environment-specific overrides (development, testing, production)
• Feature flags system for dynamic functionality control
• Comprehensive validation with cross-field dependencies

Development Status

• TODO-1: Project Structure & Core Setup -- Complete
• TODO-2: Configuration Management -- Complete
• TODO-3: Authentication & Security Framework -- Complete
• TODO-4: Telegram Bot Core -- Complete
• TODO-5: Claude Code Integration -- Complete
• TODO-6: Storage & Persistence -- Complete
• TODO-7: Advanced Features -- Complete (agentic platform, webhooks, scheduler, notifications)
• TODO-8: Complete Testing Suite -- In progress
• TODO-9: Deployment & Documentation -- In progress