Antigravity Runtime: Autonomous Local Execution Infrastructure (v6.0)
Antigravity Runtime is a production-grade, local-first Autonomous Execution Platform. It is engineered for teams that require absolute operational trust, deterministic orchestration, and governed self-healing.
Unlike generic AI assistants, Antigravity Runtime is an Execution Operating System. It doesn't just generate code—it governs its entire lifecycle, validates its integrity with cryptographic attestation, repairs its failures autonomously, and audits every micro-step via the Replay Engine.
🏛️ Core Architectural Pillars
1. Deterministic Orchestration (The Kernel)
The Antigravity Kernel maintains absolute authority over all execution cycles. Every task is orchestrated through a 7-step autonomous loop (Scaffold → Bootstrap → Provision → Pre-Flight → Judge → Self-Heal → Delivery), ensuring a predictable path from intent to verified artifact.
2. Time-Travel Replay (The Replay Engine)
Experience absolute observability. Every orchestration task generates a "Flight Log" with sub-millisecond telemetry fidelity. The Replay Engine allows you to scrub through execution timelines, audit autonomous decisions, and inspect confidence evolution in real-time.
3. Infrastructure Governance (Watchdog Sentinel)
Execution is never "unbounded." Our centralized ProcessWatchdog enforces strict, policy-driven timeouts and manages tiered escalation (SIGTERM → SIGKILL), ensuring system stability and preventing resource leakage.
4. Autonomous Recovery (Failure Taxonomy)
When orchestration fails, the system doesn't guess. It classifies errors using a formal Failure Taxonomy and routes them to specialized repairers. Our AST-Aware Patching engine applies deterministic fixes to dependencies, syntax, and configurations.
5. Operational Confidence Scoring
Trust is earned, not assumed. The Confidence Engine quantifies execution reliability based on validation grounding, repair frequency, and environment attestation. Only high-confidence builds (Score > 85) are permitted for delivery.
🔄 The 7-Step Autonomous Loop
Antigravity Runtime orchestrates work through a bulletproof lifecycle:
- SCAFFOLD: Template-driven initialization of the execution context.
- BOOTSTRAP: Sandbox creation with SHA-256 Cryptographic Attestation.
- PROVISION: Atomic dependency injection and AST-aware feature patching.
- PRE-FLIGHT: Multi-layer static analysis (Ruff, Syntax) prior to execution.
- JUDGE: Multi-vector validation of outputs, artifacts, and exit codes.
- SELF-HEAL: Taxonomy-driven autonomous repair cycles.
- DELIVERY: High-confidence artifact synchronization to remote infra.
🛠️ Technology Stack
- Kernel Core: Python 3.10+ (FastAPI, Pydantic V2, Asyncio)
- Governance: Psutil-backed ProcessWatchdog
- Static Analysis: Ruff, LibCST (AST-Aware Patching)
- Desktop HUD: Tauri, React, Vanilla CSS (High-Performance Replay UX)
- Local AI: Ollama-integrated Intent Extraction (Qwen2.5-Coder)
🚀 Getting Started
Antigravity Runtime is distributed as globally installable infrastructure software.
1. Installation
Install the runtime engine and CLI tools via pip:
pip install agrt
2. Verify Infrastructure
Ensure your local environment and required dependencies (like Ollama) are configured:
agrt doctor
3. CLI Ecosystem
The agrt CLI is the primary entry point for orchestrating autonomous workflows:
# Execute an autonomous template build
agrt run --template fastapi_basic --features uvicorn fastapi
# Validate the technical integrity of artifacts
agrt validate .
# List all autonomous execution sessions
agrt list
# Inspect a specific execution flight log
agrt inspect graph_1778195994
# View granular structured telemetry
agrt logs graph_1778195994
# Run operational performance benchmarks
agrt benchmark
4. SDK Integration
Integrate Antigravity Runtime directly into your infrastructure using the agrt Python SDK:
from agrt import Runtime
# Initialize the Runtime Kernel
runtime = Runtime()
# Dispatch an autonomous task
report = runtime.run(
task="Build FastAPI backend with JWT auth",
template="fastapi_basic",
features=["fastapi", "uvicorn"]
)
print(f"Task ID: {report.task_id} | Success: {report.success}")
5. Desktop Observability (Optional)
The optional React/Tauri Desktop HUD provides a graphical interface for Time-Travel Replay:
cd apps/desktop
npm install && npm run dev
📈 Operational Benchmarks
We prioritize empirical data over marketing hype. Our latest benchmarks show:
- Recovery Success Rate: 100% for missing dependencies and linter violations.
- Orchestration Latency: ~1.1s average overhead for the entire 7-step loop.
- Governance Accuracy: 100% timeout enforcement with zero process leaks.
View full benchmarks in docs/benchmarks.md.
📜 License
Antigravity Runtime is licensed under the MIT License. Built for the era of Autonomous Infrastructure. 🚀
Release files for agrt 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| agrt-0.1.0.tar.gz | 42.6 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| agrt-0.1.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 116.1 kB
Release files / agrt-0.1.0.tar.gz
| Download URL | agrt-0.1.0.tar.gz |
|---|---|
| Size | 42.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
ae287ee5285286f6cce03c26534d593ed9cd39dec2000c7e99144848d174d45f
|
|
BLAKE2b-256 checksum How to use checksums |
7e6f2e50b1d9b4c4b06b136a9ce6ba11cef86411aa11972aaeacdbce81868ddd
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on May 13, 2026.
Transparency logRelease files / agrt-0.1.0-py3-none-any.whl
| Download URL | agrt-0.1.0-py3-none-any.whl |
|---|---|
| Size | 73.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
b993839e10641ccaa220eabe7d4aada178d2ad348e3594c0fa3f5b779b265515
|
|
BLAKE2b-256 checksum How to use checksums |
de870d39ddf0b974baed1cdd06f8630cf7f2e95b0d4bcba77f9979d060ff91ad
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on May 13, 2026.
Transparency log