A deterministic safety controller for agentic workflows.
Project description
DASC-Core
DASC (Deterministic Agentic Safety Controller) is an open-source middleware designed to intercept outputs from AI orchestration frameworks (like LangGraph, AutoGen, and CrewAI) and enforce deterministic safety checks before actions are committed.
The Paradigm: Cognitive Fault vs. Committed Fault
DASC operates on the principle that AI agents will fail. These are Cognitive Faults (hallucinations, logic errors). Our goal is to prevent these from becoming Committed Faults (actual state changes). DASC provides the deterministic boundary where these faults are intercepted.
Installation
pip install dasc-core
Features
- LangChain/LangGraph Integration: Use
DASCCommitToolto wrap agent actions. - AutoGen Support: Use
DASCGatekeeperto intercept and validate function calls. - Semantic OCC (Hashing): Use
hash:<sha256>in version vectors for automatic file content verification. - Structured Observability: Built-in logging with detailed evaluation stages.
- Programmable Rejections: Custom exceptions (
OCCConflictError, etc.) for robust error handling. - Bitemporal SQLite Ledger: Persistent audit log of every decision.
Quick Start
Using the Kernel with Exceptions
from dasc.kernel import Kernel
from dasc.exceptions import OCCConflictError
kernel = Kernel()
try:
kernel.evaluate(intent, raise_on_failure=True)
except OCCConflictError:
# Trigger agent retry or state refresh logic
pass
Using Semantic OCC (Hashing)
from dasc.utils import calculate_file_hash
file_hash = calculate_file_hash("data.json")
intent = Intent(
...,
state_version_vector={"data.json": f"hash:{file_hash}"}
)
Documentation
- Visual Architecture: docs/ARCHITECTURE.md - Detailed Mermaid diagrams of the DASC logic flow.
- Research Paper (PDF): docs/DASC_Academic_Paper.pdf
- Technical Overview: docs/DASC_Academic_Paper.md
DASC CLI
The project includes a command-line tool for inspecting the ledger and generating hashes:
# View the last 10 decisions
dasc inspect
# Generate a hash for a file for OCC
dasc hash my_data.csv
# Verify a JSON intent manually
dasc verify intent.json --state '{"my_data.csv": "v1.0"}'
Examples
Run the production features demo:
python -m examples.production_features
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
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 dasc_core-0.1.0.tar.gz.
File metadata
- Download URL: dasc_core-0.1.0.tar.gz
- Upload date:
- Size: 17.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.9.6
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
fec75d5c3a48b2bccfd5544c5ad3b7fd043e8c9dd66ada2263fec85baa232a6e
|
|
| MD5 |
307b2a5ac38b26111a3d960a6ff5f376
|
|
| BLAKE2b-256 |
b50f15095175c82814e72d4c1af1b9a1c91ed0d6670ab036966ab25dedb2a1e4
|
File details
Details for the file dasc_core-0.1.0-py3-none-any.whl.
File metadata
- Download URL: dasc_core-0.1.0-py3-none-any.whl
- Upload date:
- Size: 20.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.9.6
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
e6701f9ad377b46fb61d8750d3f941b445356b2bd940dde08d399493f521e811
|
|
| MD5 |
9411fbac8fc3ed8613c7222033731423
|
|
| BLAKE2b-256 |
84bbf066552fcc4a908caae94e013089e8ffdce2cd191a97a1e4f4a8bde8120f
|