autobot-swarm 0.1.9

Hierarchical multi-cluster coding swarm CLI

Autobots

Hierarchical multi-cluster coding swarm CLI powered by NVIDIA NIM

Autobots is a Python CLI tool that orchestrates multiple AI models as a coding swarm to autonomously plan, implement, review, repair, and validate code in your target repositories. It routes tasks to specialized clusters, injects project context into model prompts, and runs validation loops with automatic repair and rollback support.

---

Quick Start

pip install autobot-swarm

# Check context files in your project
autobots init /path/to/your/project

# Generate an implementation roadmap
autobots plan /path/to/your/project

# Execute a task through the swarm
autobots run P1-T1

# Launch the interactive mode
autobots engage

---

What Autobots Does

1. Plans — Parses a roadmap into phases and tasks with dependencies
2. Routes — Intelligently assigns tasks to the best-fit AI cluster
3. Implements — Specialist models generate code with your project context
4. Reviews — A reviewer cluster validates correctness and safety
5. Repairs — A repair cluster fixes issues found during review
6. Validates — Runs your tests, linters, and build commands
7. Rolls Back — Snapshots before writes, autobots undo reverts changes

---

Key Features

Skill Injection

Autobots reads your context/ directory (architecture, conventions, testing strategy, security docs) and injects them into model prompts. Models see your project's actual documentation before writing code.

NVIDIA Skills

17 NVIDIA-specific skills are automatically injected into cluster system prompts, providing domain expertise for:

• RAG — NVIDIA RAG Blueprint architecture, Milvus vector DB, RAGAS evaluation
• Deployment — Dynamo Kubernetes recipes, model serving, GPU scaling
• Optimization — cuOpt routing (VRP, TSP), numerical optimization (LP/MILP/QP)
• Data — cuDF GPU DataFrames (100x faster than pandas on large datasets)
• Training — Nemotron model customization, NeMo distributed training
• Infrastructure — Kubernetes setup, OSMO workflows, Holoscan video analytics
• Quantum — CUDA-Q quantum circuit simulation

Skills are loaded automatically based on cluster role and roadmap keywords.

Nine Specialized Clusters

Cluster	Role	Use Case
Optimus	Planner	Mission briefs and orchestration
UltraMagnus	Backend	APIs, databases, server logic
Jazz	Frontend	UI components, React, CSS
RedAlert	Reviewer	Code review, security checks
Ratchet	Repair	Bug fixes, validation failures
Perceptor	Retrieval	Document parsing, RAG
Bumblebee	Media	Speech, audio, video
Ironhide	Simulation	Physics, optimization
Wheeljack	Science	Molecular, research

Execution Modes

• Supervised — Approval required before each phase
• Milestone — Approval every N phases (configurable)
• Autonomous — No approval, runs to completion

Configurable Models

# .autobots.toml
[autobots]
model_selection_profile = "balanced"  # fast, balanced, quality
temperature = 0.2
max_tokens = 4096

Additional Features

• Rollback/Undo — Snapshot system with autobots undo and autobots snapshots
• Response Streaming — Live character counter during model calls
• Doctor Preflight — Health checks for API, git, config, and dependencies
• Structured Errors — Contextual messages with actionable suggestions
• Test Gate — Run tests before commit
• Git Integration — Auto-commit after phase completion
• Config Validation — autobots config validate checks TOML settings
• Shell Completions — bash, zsh, and fish tab completion
• Context Budget — Warns and truncates when prompts approach model limits
• Plugin System — before/after hooks for custom extensions
• Skill Marketplace — Built-in packs for FastAPI, Django, React, Next.js
• Web Dashboard — Real-time status on port 8080
• Rich Status Output — Progress bars, estimated time, branch info
• Explain Command — autobots explain P2-T3 shows audit trail details
• Usage Stats — autobots stats shows totals, averages, costs
• Verbose Mode — --verbose flag logs full prompts sent to models

---

Requirements

• Python 3.11+
• NVIDIA NIM API key (free tier available at build.nvidia.com)

---

Configuration

# .autobots.toml (place in project root or $HOME)
[autobots]
model_selection_profile = "balanced"
safety_branch = "autobots-safety"
default_mode = "supervised"
milestone_threshold = 3
max_verification_attempts = 3
temperature = 0.2
max_tokens = 4096

All settings can also be set via environment variables:

export AUTOBOTS_MODEL_SELECTION_PROFILE=quality
export AUTOBOTS_SAFETY_BRANCH=main

---

CLI Commands

Command	Description
autobots init	Check context files in target project
autobots init --interactive	Interactive setup wizard
autobots plan	Generate implementation roadmap
autobots run <task>	Execute a specific task
autobots resume	Resume from last checkpoint
autobots engage	Interactive mode with startup screen
autobots status	Rich status with progress bars
autobots explain <id>	Show audit trail for a phase/task
autobots stats	Usage statistics and costs
autobots undo	Rollback to previous snapshot
autobots snapshots	List available snapshots
autobots diff	Compare workspace to snapshot
autobots logs	View audit trail
autobots doctor	Preflight health checks
autobots catalog	Browse NVIDIA model registry
autobots config validate	Validate TOML configuration
autobots completions	Generate shell completions
autobots marketplace	Browse skill packs
autobots dashboard	Launch web dashboard
autobots validate-models	Test NVIDIA API connectivity
autobots publish	Build and publish to PyPI

---

Context Architecture

Create these files in your project's context/ directory:

context/
  architecture.md       # System design, tech stack, patterns
  conventions.md        # Code style, naming, formatting rules
  testing-strategy.md   # Test framework, coverage, patterns
  security-auth.md      # Auth patterns, OWASP checklist
  roadmap.md            # Implementation plan (autobots reads this)
  progress-tracker.md   # Current status (autobots updates this)

These files are injected into model prompts so generated code matches your project's patterns.

---

How It Works

[Your Task] → [Optimus Planner] → [Specialist Cluster] → [RedAlert Reviewer]
                                          ↓                      ↓
                                    [Code Output]         [Pass / Revise]
                                                              ↓
                                                      [Ratchet Repair]
                                                              ↓
                                                      [Validation Loop]
                                                              ↓
                                                      [Snapshot + Write]

---

Safety Features

• Command whitelist blocks dangerous operations (rm -rf, sudo, etc.)
• File locking prevents concurrent writes to critical context files
• Atomic writes ensure no partial file corruption
• Path traversal protection keeps writes within allowed directories
• Audit trail logs all actions for debugging
• Test gate runs tests before commit
• Pre-commit snapshots enable rollback

---

Development Status

• Version: 0.1.9
• Status: Production-ready
• Tests: 465 passing
• License: MIT

---

Links

• PyPI: https://pypi.org/project/autobot-swarm/
• GitHub: https://github.com/DanielDeshmukh/autobots
• Issues: https://github.com/DanielDeshmukh/autobots/issues
• NVIDIA NIM: https://build.nvidia.com

---

License

MIT License — see LICENSE file for details.