Reinforcement-learning driven alignment engine for OpenClaw workflows
Project description
OpenClaw Alignment System
Reinforcement-learning driven workflow alignment engine (Actor-Critic)
English | 简体中文
Features
- Actor-Critic RL core pipeline
- Four-dimensional reward system (objective/behavior/explicit/pattern)
- Optional Phase3 modules (distributed training, tuning, monitoring, performance)
- Contract drift guards (state/action dimensions + docs consistency)
- Cross-platform support: Windows / macOS / Linux
Support Matrix
- Python: 3.10, 3.11, 3.12, 3.13
- OS: Windows, macOS, Linux
Installation
1) PyPI (Recommended)
pip install openclaw-alignment
Optional Phase3 extras:
pip install "openclaw-alignment[phase3]"
2) Install from source
git clone https://github.com/412984588/openclaw-alignment.git
cd openclaw-alignment
python3 scripts/install.py
Development install:
python3 scripts/install.py --dev --editable
Quick Verification
python3 -m pytest tests/ -v
python3 scripts/check_docs_consistency.py
openclaw-alignment --help
Architecture
Core (Phase 1-2)
lib/reward.py: reward calculation enginelib/environment.py: interaction environmentState: State data class (17 dimensions)Action: Action data class (11 dimensions)
lib/agent.py: Actor-Critic agentlib/learner.py: online learnerlib/trainer.py: training looplib/contracts.py: single source of truth for dimensions
Optional (Phase 3)
lib/distributed_trainer.pylib/hyperparameter_tuner.pylib/monitoring.pylib/performance_optimizer.py
Documentation
- Architecture:
docs/architecture.md - Reward model:
docs/reward-model.md - Configuration:
docs/configuration.md - Optional dependencies:
docs/phase3-optional-deps.md - Contributing:
CONTRIBUTING.md - Security:
SECURITY.md - Support:
SUPPORT.md
Test Coverage
- Total Tests: 80
- Pass Rate: 100%
- Core RL + integration: 54 tests ✅
- Phase 2: 1 test ✅
- Phase 3: 21 tests ✅
- Docs/contract drift guards: 4 tests ✅
Release and Versioning
- Versioning: SemVer (stable branch:
release/1.0.x) - Release runbook:
RELEASING.md/RELEASING.zh-CN.md - Changelog:
CHANGELOG.md
License
MIT
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
openclaw_alignment-1.0.0.tar.gz
(79.4 kB
view details)
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file openclaw_alignment-1.0.0.tar.gz.
File metadata
- Download URL: openclaw_alignment-1.0.0.tar.gz
- Upload date:
- Size: 79.4 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.7
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
9833470a2dc36be0a35bfd0ef62da142c34412c88ccd06478ab609abda9706bf
|
|
| MD5 |
914c6c7f1965ec321856feac9000cf08
|
|
| BLAKE2b-256 |
3bcb68474cdd3ca30c98e85b03526a578d90783ac6273d4f9ee5b801ce26e336
|
File details
Details for the file openclaw_alignment-1.0.0-py3-none-any.whl.
File metadata
- Download URL: openclaw_alignment-1.0.0-py3-none-any.whl
- Upload date:
- Size: 59.4 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.7
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
16f744d76a71d46a101e19d555eb07366bafad9d5ee57846e7c340dd160e2a60
|
|
| MD5 |
8ca9b3bb2671b681ab274396e01bd83b
|
|
| BLAKE2b-256 |
5e09d16211023a991ae6d5d373d3e767c97bc2edd7c25ed7627f7e19b814a59a
|