Metadata-Version: 2.4
Name: jingu-trust-gate
Version: 0.1.2
Summary: A deterministic trust gate for LLM systems
Project-URL: Homepage, https://github.com/ylu999/jingu-trust-gate-py
Project-URL: Repository, https://github.com/ylu999/jingu-trust-gate-py
Project-URL: Issues, https://github.com/ylu999/jingu-trust-gate-py/issues
Author: ylu999
License: MIT
Keywords: admission-control,audit,grounding,hallucination,llm,policy,rag,trust-gate
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Software Development :: Libraries
Requires-Python: >=3.11
Description-Content-Type: text/markdown

# jingu-trust-gate

**LLM output is untrusted input. jingu-trust-gate decides what is allowed to become trusted system state.**

Python SDK for [jingu-trust-gate](https://github.com/ylu999/jingu-trust-gate) — deterministic admission control layer for LLM systems.

## Install

```bash
pip install jingu-trust-gate
```

## Quick start

```python
from jingu_trust_gate import create_trust_gate, GatePolicy, Proposal, SupportRef

class MyPolicy(GatePolicy[MyClaim]):
    def validate_structure(self, proposal): ...
    def bind_support(self, unit, pool): ...
    def evaluate_unit(self, unit_with_support, ctx): ...
    def detect_conflicts(self, units, pool): ...
    def render(self, admitted_units, pool, ctx): ...
    def build_retry_feedback(self, unit_results, ctx): ...

gate = create_trust_gate(policy=MyPolicy())
result = await gate.admit(proposal, support_pool)
context = gate.render(result)
```

See `examples/medical_symptom_policy.py` for a complete working example.

## License

MIT
