Skip to content

CHANGELOG.md

Latest commit

 

History

History
373 lines (321 loc) · 16.1 KB

CHANGELOG.md

File metadata and controls

373 lines (321 loc) · 16.1 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.

[Unreleased]

Enhanced

  • TUI Professional Layout - Complete redesign of interactive mode

    • Header bar with agent status, model, tool count, LLM calls, session timer
    • Current tool indicator shows which tool is being executed
    • Activity panel shows detailed tool information (name, arguments, duration)
    • Token counts displayed in activity log for LLM responses
    • Timestamped chat messages with role indicators (YOU/AI/SYS/ERR)
    • Color-coded token gauge (green/yellow/red based on usage)
    • Professional footer with context-aware keyboard shortcuts

  • Full Input Editing - Claude Code-like input experience

    • Cursor movement with ←/→ arrow keys
    • Word-by-word navigation with Ctrl+←/→
    • Home/End keys to jump to start/end of input
    • Ctrl+A/E for bash-style start/end navigation
    • Backspace/Delete work at cursor position
    • Ctrl+W to delete word before cursor
    • Ctrl+U to clear entire input
    • Multi-line input with Alt+Enter, Ctrl+J (cross-terminal compatible)
    • Animated cursor shows position in text

  • Double-ESC to Exit - Vim-style exit

    • Press ESC twice within 500ms to quit (when not busy)
    • Single ESC still cancels running agent

  • Header Tool Count Fix

    • Now shows "Tools: X (Y used)" where X = available, Y = executed

  • Real-time Tool Activity Events

    • Activity panel now shows tool executions in real-time
    • Tool name, arguments (truncated), and execution duration displayed
    • Streaming events from runtime for accurate tool tracking
    • Current tool indicator in header during execution

[0.4.0-beta] - 2026-01-23

Added

  • Interactive TUI Mode - Full-featured terminal user interface for agent conversations

    • Launch with aofctl run agent <config.yaml> (no --input flag)
    • Chat panel with syntax-highlighted conversation history
    • Activity log showing real-time agent events (thinking, analyzing, tool use, LLM calls)
    • Context gauge displaying token usage and execution time
    • Help overlay with keyboard shortcuts (press ?)
    • LazyGit-inspired styling with clear visual hierarchy

  • Agent Cancellation - Stop running agents with ESC key

    • Graceful cancellation using tokio CancellationToken
    • Clean abort of LLM calls and tool executions
    • Status indicator shows "Cancelling..." during abort

  • Session Persistence - Conversation history saved automatically

    • Sessions stored in ~/.aof/sessions/<agent-name>/
    • Includes complete message history, token usage, activity logs
    • JSON format for easy inspection and backup

  • Session Resume - Continue previous conversations

    • --resume flag to continue latest session: aofctl run agent config.yaml --resume
    • --session <id> flag to resume specific session
    • Restored sessions show previous context to the agent

  • Session Management Commands

    • aofctl get sessions - List all saved sessions across agents
    • aofctl get sessions <agent> - List sessions for specific agent
    • Output shows session ID, agent, model, message count, tokens, age
    • Supports -o json and -o yaml output formats

  • Activity Event System - Real-time agent activity tracking

    • New ActivityEvent enum in aof-core with event types:
      • Thinking, Analyzing, LlmCall, ToolUse, ToolComplete, Warning, Error
    • ActivitySender for emitting events from runtime
    • ActivityReceiver for consuming events in TUI

Changed

  • TUI keyboard shortcuts updated:
    • ESC now cancels running agent (was: do nothing)
    • Ctrl+S saves session manually
    • Ctrl+L clears chat and starts new session
    • Shift+↑/↓ scrolls chat history
    • PageUp/Down scrolls 5 lines

Documentation

  • Updated getting-started guide with interactive mode examples
  • Added TUI keyboard shortcuts to CLI reference
  • Added session management documentation
  • Updated aofctl reference with --resume and --session flags

[0.3.2-beta] - 2026-01-02

Added

  • Built-in command handler support via agent: builtin in trigger command bindings
    • Use agent: builtin for /help, /agent, /fleet to get interactive menus
    • Interactive menus include fleet/agent selection buttons (Telegram/Slack)
    • Keeps built-in UI handlers separate from LLM-routed commands
  • Stale message filtering for webhook handlers
    • Messages older than 60 seconds are silently dropped
    • Prevents processing of queued messages when daemon restarts
    • Configurable via max_message_age_secs in handler config
  • cargo install aofctl support via crates.io publishing
    • All AOF crates now published to crates.io
    • Automated publishing on tagged releases
  • New documentation: Built-in Commands Guide (docs/guides/builtin-commands.md)

Fixed

  • aofctl serve now produces visible startup output
    • Changed from tracing (default level: error) to println for critical startup messages
    • Users can now see server bind address, registered platforms, loaded agents/flows
    • Error messages use stderr for proper output separation
  • GitHub/GitLab/Bitbucket PR reviews now post a single response instead of multiple comments
    • Intermediate acknowledgment messages ("Thinking...", "Processing...") are skipped for Git platforms
    • Only the final response is posted, keeping PR threads clean
    • Slack/Telegram/Discord still show real-time progress indicators
  • Improved library:// URI path resolution for agent library

[0.3.1-beta] - 2025-12-26

Added

  • Token usage tracking for AgentFlow execution
    • Agent nodes now report input/output tokens
    • Flow completion summary shows total token usage
    • Script nodes correctly show 0 tokens (no LLM usage)
  • --library flag for aofctl get agents to list built-in agents
    • Shows all 30 production-ready agents from the library
    • Displays domain (category), agent name, status, and model
    • Supports filtering by agent name: aofctl get agents pod-doctor --library
    • Supports JSON/YAML output formats
  • library:// URI syntax for running agents from the built-in library
    • Format: library://domain/agent-name
    • Example: aofctl run agent library://kubernetes/pod-doctor --prompt "debug CrashLoopBackOff"
    • Helpful error messages showing available agents when agent not found
  • --prompt as an alias for --input in the run command
    • More intuitive for LLM-style interactions

Fixed

  • Script node YAML field naming (scriptConfig camelCase)
  • Flow completion display formatting for token line

[0.3.0-beta] - 2025-12-24

Added

Agent Library (30 Production-Ready Agents)

  • Kubernetes Domain (5 agents)
    • deploy-guardian: Validates deployments before production rollout
    • node-doctor: Diagnoses and auto-heals node problems
    • resource-optimizer: Right-sizes containers based on actual usage
    • pod-debugger: Troubleshoots pod failures with context
    • rollout-manager: Manages progressive deployments
  • Observability Domain (5 agents)
    • alert-manager: Manages Prometheus alerts with runbook automation
    • slo-guardian: Monitors SLI/SLO compliance and error budgets
    • log-analyzer: Analyzes logs for patterns and anomalies
    • metrics-explorer: Queries and visualizes Prometheus metrics
    • trace-investigator: Analyzes distributed traces
  • Incident Domain (5 agents)
    • rca-agent: Performs automated root cause analysis
    • incident-commander: Orchestrates incident response
    • escalation-manager: Routes incidents to appropriate teams
    • postmortem-writer: Generates blameless postmortems
    • runbook-executor: Executes runbooks with safety checks
  • CI/CD Domain (5 agents)
    • pipeline-fixer: Diagnoses and fixes failing pipelines
    • build-optimizer: Optimizes build performance
    • release-manager: Coordinates release workflows
    • test-analyzer: Analyzes test failures and flakiness
    • artifact-manager: Manages build artifacts and images
  • Security Domain (5 agents)
    • vuln-scanner: Scans for vulnerabilities with Trivy
    • secret-auditor: Audits secrets management with Vault
    • compliance-checker: Validates compliance with OPA policies
    • access-reviewer: Reviews RBAC permissions
    • security-responder: Responds to security incidents
  • Cloud Domain (5 agents)
    • cost-optimizer: Analyzes cloud costs and recommends savings
    • drift-detector: Detects infrastructure drift with Terraform
    • capacity-planner: Forecasts capacity needs
    • backup-validator: Validates backup integrity
    • multi-cloud-coordinator: Coordinates across AWS/Azure/GCP

MCP Server Catalog (10 Documented Servers)

  • Core: filesystem, fetch, puppeteer
  • Development: github, gitlab
  • Databases: postgres (read-only), sqlite (read/write)
  • Communication: slack
  • Search: brave-search

Each catalog entry includes:

  • Configuration examples for AOF agents
  • Full tool reference with parameters
  • Use case examples with agent specs
  • Troubleshooting guides

Comprehensive Documentation

  • Agent Library user guide with domain overviews
  • MCP integration guide with tested servers
  • Wired both sections into Docusaurus sidebar

Changed

  • Improved getting-started with zero-setup examples
  • Cleaned up architecture documentation
  • Updated all agent specs to use correct tool naming (underscore-separated)

[0.2.0-beta] - 2025-12-20

Added

Composable Architecture (Major Refactor)

  • Simplified to 4 Core Concepts: Agent, Fleet, Flow, Trigger
  • Composable Design: Mix and match agents, tools, and triggers
  • Removed complex FlowBinding in favor of direct trigger→agent mapping

New Trigger Platforms (5 New)

  • Microsoft Teams - Bot Framework integration with Adaptive Cards
    • JWT Bearer token authentication
    • Tenant and channel restrictions
    • Action.Submit handling for button clicks
  • WhatsApp Business - Cloud API integration
    • HMAC-SHA256 signature verification
    • Interactive buttons and lists
    • Template message support
  • GitHub - Webhook integration for PR/Issue automation
    • PR opened/updated/merged events
    • Issue created/updated events
    • Comment triggers with @mention detection
  • GitLab - Webhook integration for MR automation
    • Merge request events
    • Pipeline status triggers
    • Note (comment) events
  • Bitbucket - Webhook integration for PR automation
    • Pull request events
    • Repository push events
  • Jira - Issue tracking platform abstraction
    • Issue created/updated/transitioned events
    • JQL query support
    • Comment and attachment handling

Comprehensive Documentation

  • Concepts: Teams, Discord, WhatsApp, Jira integration overviews
  • Reference: Full API reference for each platform
  • Tutorials: Step-by-step ops bot tutorials
  • Quickstart Guides: 10-15 minute setup guides

Platform Capabilities System

  • Thread support detection
  • Interactive element support
  • File attachment support
  • Reaction support
  • Rich text support
  • Approval workflow support

Changed

  • Simplified Agent Switching - Easy agent switching via /agent and /help commands
  • Platform-Based Safety Layer - Read-only mode for mobile platforms
  • Simplified Output - Text-only responses for better mobile display
  • Daemon configuration simplified with direct platform webhook paths

Fixed

  • Telegram inline keyboard callback handling
  • Agent loading when flows directory doesn't exist
  • System prompts correctly loaded from agent YAML
  • Model configuration from agent YAML (was hardcoded)

Technical Details

  • 9 trigger platforms: Slack, Discord, Telegram, WhatsApp, Teams, GitHub, GitLab, Bitbucket, Jira
  • 7 built-in tools: kubectl, docker, aws, terraform, git, shell, http
  • Platform registry with factory pattern for extensibility
  • Ed25519 (Discord) and HMAC-SHA256 (others) signature verification
  • ~60,000 lines of Rust code

[0.1.15] - 2025-12-18

Added

  • Human-in-the-Loop Approval Workflow - Reaction-based command approval for Slack
    • ✅/❌ reactions for approve/deny destructive operations
    • Automatic ✅ ❌ reaction buttons added to approval messages
    • Support for multiple approval reactions (thumbs up, checkmark variants)
    • Configurable approval_allowed_users for role-based approval
  • Conversation Memory System - Context persistence across Slack messages
    • Per-channel and per-thread memory isolation
    • Automatic context injection for follow-up messages
    • Supports "delete it", "scale that down" style contextual commands
    • Configurable memory limits (20 messages default, 10 in context)
  • AgentFlow Multi-Tenant Routing - Route messages to different agents based on patterns
    • Pattern-based agent selection from incoming messages
    • Support for multiple agents in a single flow
  • Bot Self-Approval Prevention - Auto-detects bot_user_id at startup
    • Uses Slack's auth.test API to get bot's own user ID
    • Filters out bot's own reactions to prevent self-approval
  • New documentation guides:
    • docs/guides/approval-workflow.md - Complete approval workflow guide
    • docs/guides/conversation-memory.md - Conversation memory guide

Changed

  • kubectl-style verb-noun CLI commands for Fleet and Flow resources
  • Flow resource type with short name fw (alias for workflow execution)
  • All documentation updated to use aofctl <verb> <resource> syntax
  • Updated slack-k8s-bot agent with explicit kubectl syntax guidance

Fixed

  • Telegram inline keyboard callback handling - Fixed "Invalid selection" error when tapping agent/flow buttons
    • Improved callback format parsing to handle both callback:agent:name and agent:name formats
    • Added better debug logging for callback troubleshooting
    • More descriptive error messages when callback parsing fails
  • System prompts from agent YAML files now correctly loaded and used
  • Agent execution error handling with proper response to platforms
  • Bracket structure in handle_natural_language function
  • Agent loading flag logic - Fixed issue where agents weren't loaded when flows directory didn't exist
    • Introduced proper agents_loaded boolean tracking
    • Agents now correctly pre-load regardless of flows configuration
  • /run agent using wrong model - Fixed hardcoded Anthropic model to use pre-loaded agent's configuration
    • Now uses google:gemini-2.5-flash from agent YAML
    • Tools and system prompts correctly applied

Notes

  • Config changes to approval_allowed_users require server restart (hot-reload coming in Issue #22)

[0.1.14] - 2025-12-17

Added

  • Initial release with core aofctl functionality
  • Agent execution with multiple LLM providers (OpenAI, Anthropic, Google, Ollama, Groq)
  • MCP (Model Context Protocol) server support
  • Memory backends (InMemory, File, SQLite)
  • Built-in tools (Shell, HTTP, FileSystem)
  • AgentFleet multi-agent coordination
  • AgentFlow workflow orchestration
  • Trigger system (Webhook, Schedule, FileWatch, Slack, GitHub, PagerDuty)

Commands

  • aofctl run agent <file> - Execute an agent
  • aofctl run workflow <file> - Execute a workflow
  • aofctl run fleet <file> - Execute a fleet
  • aofctl get <resource> - List resources
  • aofctl describe <resource> <name> - Show resource details
  • aofctl apply -f <file> - Apply configuration
  • aofctl delete <resource> <name> - Delete resources
  • aofctl logs <resource> <name> - View logs
  • aofctl api-resources - List available resource types
  • aofctl version - Show version information

Resource Types

Resource Short Name Description
agents ag AI agents
workflows wf Multi-step workflows
fleets fl Multi-agent coordination
flows fw Workflow aliases
tools tl MCP tools

Release Notes Format

Version Number Meaning

  • MAJOR: Breaking changes to CLI or API
  • MINOR: New features, backwards compatible
  • PATCH: Bug fixes, documentation updates

Categories

  • Added: New features
  • Changed: Changes to existing functionality
  • Deprecated: Features to be removed in future
  • Removed: Features removed in this release
  • Fixed: Bug fixes
  • Security: Security-related changes