autobot-swarm 0.1.4

Hierarchical multi-cluster coding swarm CLI

Autobots is a Python CLI for running a structured, approval-gated coding swarm against target repositories. It can check target-owned context files, generate phased plans, route implementation work through hierarchical model clusters, and execute autonomous work with validation and repair loops.

Table of Contents

• Overview
• Features
• Architecture
• Installation
• Configuration
• CLI Commands
• NVIDIA Models Registry
• Execution Modes
• Context Architecture
• Cross-Platform Usage
• Environment Variables
• Troubleshooting
• Development
• License

---

Overview

Autobots transforms your development workflow by orchestrating multiple AI models as a hierarchical swarm:

1. Optimus (Command Cluster) - Plans and orchestrates the mission
2. UltraMagnus (Backend Cluster) - Implements backend logic and APIs
3. Jazz (Frontend Cluster) - Creates UI components and visual elements
4. RedAlert (Security Cluster) - Reviews code for safety and correctness
5. Ratchet (Repair Cluster) - Fixes validation failures and bugs
6. Perceptor (Retrieval Cluster) - Handles document parsing and RAG
7. Bumblebee (Media Cluster) - Processes speech, audio, and video
8. Ironhide (Simulation Cluster) - Runs physics and optimization tasks
9. Wheeljack (Science Cluster) - Handles molecular and research tasks

---

Features

• Context Checks - Auto-detect project type (Python, Node, etc.) and report missing context filenames
• Phase Planning - Generate implementation roadmaps with dependencies and acceptance criteria
• Model Routing - Intelligent cluster selection based on task keywords
• Multi-Root File Writing - Write to src/, app/, lib/, tests/, docs/, scripts/
• Validation Commands - Run tests, linters, and build commands automatically
• Automatic Repair - Self-healing execution with validation-driven repair loops
• Session Management - Durable checkpoints, resumable runs, and audit trails
• Configurable Modes - Supervised, milestone, or fully autonomous execution

---

Architecture

Autobots follows a modular architecture:

autobots/
├── cli.py              # CLI entry point and command handlers
├── bootstrap.py        # Project profiling and context filename contract
├── planning.py         # Roadmap and progress tracker generation
├── router/
│   ├── core.py        # Main routing orchestration
│   ├── models.py      # Data models (ClusterPlan, PhaseRecord, etc.)
│   ├── planning.py    # Cluster assignment and model selection
│   ├── stages.py      # Stage execution (command, specialist, review, repair)
│   └── phases.py      # Phase reading and status management
├── executor/
│   ├── autonomy.py    # Autonomous execution engine
│   ├── modes.py       # Execution modes (supervised, milestone, autonomous)
│   ├── state.py       # Session state and checkpoint management
│   ├── commands.py    # Command validation and execution
│   └── validation.py  # Validation result handling
├── catalog.py         # Cluster definitions and model registry
├── config.py         # Configuration management
└── workspace.py       # Target workspace safety and locking

---

Installation

Prerequisites

• Python 3.11 or higher
• NVIDIA API Key (for model execution)

Quick Install

# Clone the repository
git clone https://github.com/DanielDeshmukh/autobots.git
cd autobots

# Install in development mode
python -m pip install -e . --no-build-isolation

# Verify installation
autobots --help

pip install (when published)

pip install autobot-swarm

---

Configuration

API Key Setup

Create a .env file in the project root:

NVIDIA_API_KEY=your_nvidia_api_key_here

Or set the environment variable:

$env:NVIDIA_API_KEY = "your_key_here"

Configuration File

Create a .autobots.toml or autobots.toml in your project or home directory:

[autobots]
# Model selection: balanced, speed, or quality
model_selection_profile = "balanced"

# Enable parallel workstream planning
parallel_planning = false

# Use bundled models only (no live catalog)
disable_live_catalog = false

# Safety branch name
safety_branch = "autobots-safety"

# Default execution mode: supervised, milestone, or autonomous
default_mode = "supervised"

# Phases before approval in milestone mode
milestone_threshold = 3

# Max verification attempts per phase
max_verification_attempts = 3

# Optional: custom model registry
# model_registry_path = "./my-registry.json"

# Optional: extra custom clusters
# [autobots.extra_clusters]
# MyCluster = ["nvidia/custom-model-1"]

---

CLI Commands

All commands automatically detect the target project from the current working directory.

Check Project Context

autobots init

Checks for the required context filenames and detects the project profile. Autobots does not create these files for target projects.

Generate/Refresh Plan

autobots plan [options]

Generates roadmap and progress tracker from project context.

Flag	Description
--goal "text"	Set the planning goal
--append	Append new phases instead of replacing
--insert-after "phase"	Insert after specific phase
--dry-run	Preview without writing files

Execute Phases

autobots run [options]

Execute phases with specified autonomy mode.

Flag	Description
--supervised	Manual approval per phase (default)
--milestone	Approval every N phases
--autonomous	Fully autonomous execution
--dry-run	Preview without executing

Resume Execution

autobots resume

Resume from the last checkpoint after interruption.

Show Status

autobots status

Display current execution status, session details, and checkpoint information.

Interactive Mode

autobots engage

Launch the interactive approval-gated workflow with full cluster orchestration.

Validate Models

autobots validate-models

Test model contracts and JSON responses against the live NVIDIA API.

---

NVIDIA Models Registry

Autobots includes a comprehensive model registry with 9 clusters:

Cluster Overview

Cluster	Role	Models	Keywords
Optimus	Command & Routing	9	plan, roadmap, phase, orchestrate
UltraMagnus	Backend & Architecture	12	backend, api, database, service
RedAlert	Security & Safety	10	security, auth, guardrail, validation
Jazz	Frontend & Creative	11	ui, ux, css, image, visual
Ratchet	Debug & Repair	10	debug, fix, refactor, patch, repair
Perceptor	Retrieval & Parsing	11	ocr, rag, embedding, document
Bumblebee	Communication & Media	9	speech, voice, translation, audio
Ironhide	Physical & Simulation	8	simulation, physics, autonomous
Wheeljack	Scientific Specialist	7	science, molecule, protein, quantum

Model List by Cluster

Optimus (Command & Planning)

• nvidia/nemotron-3-super-120b-a12b
• nvidia/llama-3.3-nemotron-super-49b-v1.5
• nvidia/mistral-large-3-675b-instruct-2512
• nvidia/kimi-k2-thinking
• nvidia/step-3.5-flash
• nvidia/gpt-oss-120b
• nvidia/glm-5.1
• nvidia/llama-4-maverick-17b-128e-instruct
• nvidia/stockmark-2-100b-instruct

UltraMagnus (Backend & Architecture)

• nvidia/kimi-k2.6
• nvidia/deepseek-v4-pro
• nvidia/qwen3.5-397b-a17b
• nvidia/mistral-medium-3.5-128b
• nvidia/gemma-4-31b-it
• nvidia/qwen3-next-80b-a3b-thinking
• nvidia/dracarys-llama-3.1-70b-instruct
• nvidia/mixtral-8x22b-instruct-v0.1
• nvidia/evo2-40b
• nvidia/boltz-2
• nvidia/alphafold2-multimer
• nvidia/msa-search

RedAlert (Security & Safety)

• nvidia/llama-3.1-nemotron-70b-instruct
• nvidia/nemotron-4-340b-instruct
• deepseek-ai/deepseek-v4-pro
• nvidia/llama-3.1-nemotron-51b-instruct
• meta/llama-3.1-405b-instruct
• deepseek-ai/deepseek-v4-flash
• nvidia/llama-3.3-nemotron-super-49b-v1.5
• nvidia/mistral-large-3-675b-instruct-2512
• nvidia/qwen3.5-397b-a17b
• nvidia/mistral-medium-3.5-128b

Jazz (Frontend & Creative)

• nvidia/qwen-image-edit
• nvidia/qwen-image
• nvidia/flux.2-klein-4b
• nvidia/flux.1-dev
• nvidia/flux.1-schnell
• nvidia/stable-diffusion-3.5-large
• nvidia/FLUX.1-Kontext-dev
• nvidia/phi-4-multimodal-instruct
• nvidia/NVIDIA AI for Media Relighting
• nvidia/TRELLIS
• nvidia/vista-3d

Ratchet (Debug & Repair)

• nvidia/deepseek-v4-flash
• nvidia/qwen3.5-coder-480b-a35b-instruct
• nvidia/qwen2.5-coder-32b-instruct
• nvidia/mistral-small-4-119b-2603
• nvidia/devstral-2-123b-instruct-2512
• nvidia/magistral-small-2506
• nvidia/phi-4-mini-instruct
• nvidia/llama-3.2-3b-instruct
• nvidia/llama-3.2-1b-instruct
• nvidia/nemotron-mini-4b-instruct

Perceptor (Retrieval & Parsing)

• nvidia/nemotron-ocr-v1
• nvidia/nemotron-parse
• nvidia/paddleocr
• nvidia/nemotron-table-structure-v1
• nvidia/nemotron-page-elements-v3
• nvidia/nemotron-graphic-elements-v1
• nvidia/llama-3.2-nemoretriever-300m-embed-v2
• nvidia/llama-3.2-nv-embedqa-1b-v2
• nvidia/llama-3.2-nv-rerankqa-1b-v2
• nvidia/nv-embedcode-7b-v1
• nvidia/bge-m3

Bumblebee (Communication & Media)

• nvidia/whisper-large-v3
• nvidia/canary-1b-asr
• nvidia/riva-translate-4b-instruct-v1_1
• nvidia/magpie-tts-zeroshot
• nvidia/nemotron-voicechat
• nvidia/LipSync
• nvidia/Background Noise Removal
• nvidia/Active Speaker Detection
• nvidia/parakeet-1.1b-rnnt-multilingual-asr

Ironhide (Simulation & Optimization)

• nvidia/cosmos-reason2-8b
• nvidia/cosmos-transfer2.5-2b
• nvidia/cosmos-predict1-5b
• nvidia/streampetr
• nvidia/sparsedrive
• nvidia/bevformer
• nvidia/fourcastnet
• nvidia/cuopt

Wheeljack (Scientific Specialist)

• nvidia/ising-calibration-1-35b-a3b
• nvidia/genmol
• nvidia/molmim
• nvidia/rfdiffusion
• nvidia/proteinmpnn
• nvidia/esm2-650m
• nvidia/openfold3

---

Execution Modes

Supervised Mode (Default)

autobots run --supervised

• Requires approval before each phase
• Full human control over execution
• Best for critical or unfamiliar tasks

Milestone Mode

autobots run --milestone

• Requires approval after N phases (default: 3)
• Balance between autonomy and control
• Good for well-understood workflows

Autonomous Mode

autobots run --autonomous

• No approval gates
• Fastest execution
• Best for trusted, well-tested workflows

---

Context Architecture

Autobots expects six target-owned context files under context/:

File	Purpose
architecture.md	Project architecture and structure
roadmap.md	Phase definitions with goals and validation
progress-tracker.md	Phase status tracking (PENDING/IN_PROGRESS/COMPLETE)
ui-components.md	UI component inventory
project-briefing.md	Project overview and metadata
security-auth.md	Security and authentication notes

---

Cross-Platform Usage

All commands automatically use the current working directory as the target project.

Windows

# Install
python -m pip install -e .

# Set API key
$env:NVIDIA_API_KEY = "your_key"

# Run commands (from your project directory)
cd C:\path\to\your\project
autobots init
autobots plan
autobots run

macOS / Linux

# Install
python3 -m pip install -e .

# Set API key
export NVIDIA_API_KEY="your_key"

# Run commands (from your project directory)
cd /path/to/your/project
autobots init
autobots plan
autobots run

---

Environment Variables

Variable	Description	Default
NVIDIA_API_KEY	NVIDIA API key for model access	Required
AUTOBOTS_MODEL_SELECTION_PROFILE	Model selection: balanced/speed/quality	balanced
AUTOBOTS_ENABLE_PARALLEL_PLANNING	Enable parallel workstreams	false
AUTOBOTS_DISABLE_LIVE_CATALOG	Use bundled models only	false
AUTOBOTS_SAFETY_BRANCH	Required git branch name	autobots-safety
AUTOBOTS_DEFAULT_MODE	Default execution mode	supervised
AUTOBOTS_MILESTONE_THRESHOLD	Phases per approval	3
AUTOBOTS_MAX_VERIFICATION_ATTEMPTS	Retry limit	3
AUTOBOTS_MODEL_REGISTRY	Custom model registry path	None

---

Troubleshooting

Missing NVIDIA API Key

RuntimeError: NVIDIA_API_KEY is missing. Cannot execute the swarm.

Solution: Set NVIDIA_API_KEY in .env or environment variable.

Incomplete Context Setup

Command: run
Missing: roadmap.md, progress-tracker.md

Solution: Create the listed files in the target project's context/ folder, then run autobots plan if you need to refresh roadmap/progress content.

Safety Branch Check Fails

Execution blocked. Switch to autobots-safety branch.

Solution: Create and checkout the safety branch:

git checkout -b autobots-safety

Command Policy Violation

Command not in safety whitelist: rm -rf /

Solution: Autobots blocks dangerous commands. Use safe alternatives.

Resume Without Checkpoint

No checkpoint found. Use 'autobots run' to start fresh.

Solution: Start a new run with autobots run.

---

Development

Running Tests

# Run all tests
python -m pytest tests/ -v

# Run specific test file
python -m pytest tests/test_phase_9_config.py -v

# Run with coverage
python -m pytest tests/ --cov=autobots --cov-report=html

Project Structure

autobots/
├── autobots/           # Main package
│   ├── cli.py         # CLI entry point
│   ├── config.py      # Configuration
│   ├── catalog.py     # Model registry
│   ├── bootstrap.py    # Project profiling
│   ├── planning.py     # Roadmap generation
│   ├── router/         # Routing orchestration
│   ├── executor/       # Execution engine
│   └── workspace.py    # Workspace management
├── tests/             # Test suite
├── context/           # Project context (for this repo)
├── setup.cfg          # Package configuration
└── README.md          # This file

---

License

MIT License - See LICENSE for details.

---

Version History

Version	Date	Changes
0.1.4	2026-06-08	Removed Autobots-created starter context files; missing context now reports filenames only
0.1.1	2026-05-15	Phase 9-10 complete: config, error handling, failure tests
0.1.0	2026-05-14	Phase 1-8.5 complete: core functionality

---

Acknowledgments

• Built with NVIDIA NIM for model access
• Powered by Rich for terminal UI
• Uses OpenAI compatible API interface