mirror of
https://github.com/onestardao/WFGY.git
synced 2026-04-28 11:40:07 +00:00
0d8a6437a0
Replace over-broad on_chain_start with on_retriever_start to capture the query embedding as late and as safely as possible in the RetrievalQA flow, minimizing framework footprint. Addresses maintainer review feedback for minimal & safe implementation. |
||
|---|---|---|
| .. | ||
| firewall.py | ||
| README.md | ||
WFGY LangChain Adapter (Semantic Firewall)
This adapter provides a WFGYSemanticFirewall callback handler for LangChain, enabling real-time monitoring of semantic tension (ΔS) to prevent hallucinations and perception drift.
Quick Start
from wfgy.adapters.langchain.firewall import WFGYSemanticFirewall
from langchain_openai import OpenAIEmbeddings
from langchain_openai import ChatOpenAI
from langchain.chains import RetrievalQA
# Initialize embeddings (must be consistent)
embeddings = OpenAIEmbeddings()
# Initialize WFGY Firewall
firewall = WFGYSemanticFirewall(embedding_model=embeddings)
# Plug into any LangChain object or chain
llm = ChatOpenAI(callbacks=[firewall])
# or use it in a chain
qa_chain = RetrievalQA.from_chain_type(
llm=llm,
chain_type="stuff",
retriever=your_retriever,
callbacks=[firewall]
)
qa_chain.run("How does WFGY work?")
Metrics Tracked
- ΔS (Query <-> Context): Measures how well the retrieved documents match the user's intent. High ΔS indicates "Perception Drift".
- ΔS (Final Answer): Measures how well the LLM's response stays within the semantic bounds of the query.
Zones
- Safe: ΔS < 0.4
- Risk: 0.6 < ΔS < 0.85
- Danger: ΔS ≥ 0.85 (Triggers alerts)