░████             ░██        ░██░██       ░███                                        ░██
    ░██  ░██            ░██           ░██      ░██░██                                       ░██
   ░██   ░██  ░███████  ░████████  ░██░██     ░██  ░██   ░████████  ░███████  ░████████  ░████████
  ░██    ░██ ░██    ░██ ░██    ░██ ░██░██    ░█████████ ░██    ░██ ░██    ░██ ░██    ░██    ░██
  ░██    ░██ ░█████████ ░██    ░██ ░██░██    ░██    ░██ ░██    ░██ ░█████████ ░██    ░██    ░██
  ░██    ░██ ░██        ░███   ░██ ░██░██    ░██    ░██ ░██   ░███ ░██        ░██    ░██    ░██
  ░█████████  ░███████  ░██░█████  ░██░██    ░██    ░██  ░█████░██  ░███████  ░██    ░██     ░████
░██        ░██                                                 ░██
Агент который только говорит ДА                          ░███████

Multiple workstations, dozens of projects, across hundereds of instances of Claude Code/Gemini/Codex. No problem. Tired of CLI agents asking for permissions in Approve All mode, but afraid to go YOLO. No problem.

An unfinished, evolving collection of everything agentic — prompt engineering patterns, autonomous agent instructions, multi-agent orchestration experiments, and tooling. This is a working dumping ground, not a polished framework.

Table of Contents

• In development
  • DA (ДА)
  • Gadya (Гадя)
• No longer developed
  • DAPY CLI
• What's where

In development

• DA (ДА): Personal multi-agent CLI and TUI built directly on the Anthropic SDK. Multi-session support, tool execution, Claude session browsing, and remote host management.
• Gadya (Гадя): A voice-first iOS/Android mobile assistant built with React Native / Expo and TypeScript. Designed for hands-free, one-handed interaction with LLMs — speak a question, get an AI answer read aloud, and optionally save or search personal notes via voice.
• DAPY CLI: A more ambitious attempt to wrap the whole workflow into a LangChain-based CLI tool

---

DA (ДА) - Агент который только говорит ДА

Superagent to manage cli coding agents, projects, tools. Makes Claude Code compliant (not asking too many questions).

Features

Session Management

Browse sessions across multiple machines.

Delete, rename, move and merge.

The tree expands to show full session history with timestamps. Selected sessions show recent conversation messages and available actions in the detail panel.

Obsidian Integration

Obsidian vault integrated. Used to manage Projects - an abstraction above sessions.

Editor

A tree-based browser for tools, agents, and YAML config files. Da can be fully configured from its own ui or using its own chat interface.

REPL

CLI Commands

Command	Description
da	Launch full TUI (default)
da rich	Rich TUI — 5-tab layout with Projects, Chat, Sessions, Obsidian, Config
da tui	Sidebar TUI — session list + conversation layout
da repl	Interactive REPL with slash commands and tab completion
da manage	Session manager — browse, rename, copy, move, delete
da ask "<query>"	One-shot query to the orchestrator agent
da sessions	List recent sessions
da resume <id>	Resume a previous session
da analyze	Analyze Claude Code conversation history
da diag	Show diagnostic info (model, hosts, tools)
da next	Read 2Do.md, ROADMAP.md, git status — recommend what to do
da close	Close session — update 2Do.md, document progress
da document	Generate CHANGELOG.md from git diff
da push [msg]	Commit and push with changelog verification

Flags: --model / -m (model override), --debug, --config / -c.

Tools (23)

Category	Tools
Shell	shell_exec — run arbitrary commands
Git	git_status, git_diff, git_log, git_commit_push
Docker	docker_ps, docker_logs, docker_exec, docker_compose
SSH	ssh_exec, ssh_batch (parallel multi-host)
Files	read_file, write_file, list_dir, find_files, delete_file, copy_file, move_file, edit_file
Search	grep_search, glob_find
Sessions	list_sessions, get_session, delete_session

REPL Shortcuts

Slash commands in the REPL and TUI input:

Command	Action
/model [opus|sonnet|haiku]	Switch model or show current
/tools	List all available tools
/hosts	Show configured remote hosts
/git [status|diff|log|commit|push]	Git operations
/docker [ps|logs|exec|up|down]	Docker operations
/ssh <host> [cmd]	Execute on remote host
/grep <pattern>, /find <glob>	Search files
/cat <path>, /ls [path]	Read file / list directory
/sessions, /switch <id>, /delete [id]	Session management
/compact	Summarize and trim conversation context
!<command>	Direct shell execution

Tab completion for commands, model names, hosts, git/docker subcommands, and file paths.

Keyboard Shortcuts

Key	Action
F1–F5	Switch views: Projects, Chat, Sessions, Obsidian, Config
Ctrl+N	New session
Ctrl+Q	Quit
Esc	Back / close editor

---

See Gadya/README.md for full documentation.

---

What's where

.
├── AGENTS.md                  → Master instructions for autonomous agents (Claude, Gemini, etc.)
├── AGENTS_mistakes.md         → Log of agent mistakes and lessons learned
├── CLAUDE.md / GEMINI.md      → Per-LLM instruction pointers
│
├── DA/                       → Personal multi-agent CLI/TUI (Anthropic SDK)
│   ├── da/                   → Python package (cli, tui, agents, tools)
│   ├── scripts/              → Install script
│   └── config.yaml           → Model, hosts, session settings
│
├── DAPY/                → LangChain/LangGraph CLI for agentic workflows
│   ├── dapy/            → Python package (orchestrator, tools, middleware)
│   ├── deployment/            → Docker, GCP, LangChain Cloud configs
│   └── tools/                 → LLM log ingestion and querying
│
├── Gadya/                     → Voice-first mobile assistant (React Native / Expo)
│   ├── app/                   → Expo app screens and navigation
│   ├── server/                → Express + tRPC backend with LangChain chains
│   ├── hooks/                 → Voice system, continuous listening
│   └── services/              → LLM integration, notes, storage
│
├── agents/                    → Specialized subagent definitions (legacy)
│   ├── knowledge-management-agent.md
│   ├── changelog-agent.md
│   ├── docs-agent.md
│   ├── archive-agent.md
│   ├── validation-agent.md
│   ├── push-agent.md
│   └── mistake-review-agent.md
│
├── agent-swarm/               → Multi-agent orchestration experiments (legacy)
│   ├── ORCHESTRATOR.md
│   ├── claude-code-orchestration-plan.md
│   └── agents/                → Alternative agent definitions
│
└── archive/                   → Previous iterations, old plans, templates, guides
    ├── INSTRUCTIONS_CORE.md, INSTRUCTIONS_EXPERIMENTAL.md
    ├── ROADMAP.md, 2Do.md
    ├── templates/             → README, AGENTS, archive templates
    ├── plans/                 → Subagent project plans
    └── guides/                → Best practices (archiving, etc.)

No longer developed

• AGENTS.md as a pattern: A single markdown file that tells any LLM agent how to behave in a repository — session management, git workflow, mistake tracking, changelog conventions
• Multi-agent orchestration: Experiments with specialized subagents (changelog, docs, archive, validation, push) coordinated by a parent knowledge-management agent
• Mistake-driven improvement: Agents log their own mistakes; those patterns feed back into better instructions

DAPY CLI

See DAPY/README.md for full documentation.

---

      ░████                     ░███    
    ░██  ░██                   ░██░██                      
   ░██   ░██                  ░██  ░██  
  ░██    ░██       _    _ _  ░█████████               _   
  ░██    ░██   ___| |__(_) | ░██    ░██  __ _ ___ _ _| |_ 
  ░██    ░██  / -_) '_ \ | | ░██    ░██ / _` / -_) ' \  _|
  ░█████████  \___|_.__/_|_| ░██    ░██ \__, \___|_||_\__|
░██        ░██                           |___/             
Агент который только говорит ДА