LangChain

Install

pip install veldrixai langchain

Callback handler

Attach the VeldrixAI callback to any chain or agent. It evaluates each LLM output as it is produced and logs the verdict to your audit trail.

from veldrixai.adapters.langchain import VeldrixCallbackHandler

handler = VeldrixCallbackHandler(api_key="vx-live-...", block_on_verdict=["BLOCK"])

chain.invoke({"input": user_query}, config={"callbacks": [handler]})

Blocking on verdict

With block_on_verdict=["BLOCK"] the handler raises when an output is blocked, halting the chain so the unsafe text never reaches the next step or the user. Catch the exception to return a safe fallback.