moss-langchain 0.1.0

MOSS signing integration for LangChain via callback handler

moss-langchain

MOSS signing integration for LangChain via callback handler.

Installation

pip install moss-langchain

Usage

from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from moss_langchain import SignedCallbackHandler

# Create callback handler
cb = SignedCallbackHandler("moss:bot:summary")

# Create your chain
chain = ChatPromptTemplate.from_template("Summarize: {text}") | ChatOpenAI()

# Invoke with callback
result = chain.invoke(
    {"text": "Long document..."},
    config={"callbacks": [cb]}
)

# Access the signature
envelope = cb.envelope

Verification

from moss import Subject

# Verify the output
result = Subject.verify(cb.envelope)
assert result.valid

Multiple Outputs

The handler tracks all outputs during a session:

cb = SignedCallbackHandler("moss:bot:pipeline")

# Run multiple operations
chain1.invoke(input1, config={"callbacks": [cb]})
chain2.invoke(input2, config={"callbacks": [cb]})

# Access all envelopes
for envelope in cb.envelopes:
    print(f"Seq {envelope.seq}: {envelope.payload_hash}")

# Clear for next session
cb.clear()

Async Chains

from moss_langchain import AsyncSignedCallbackHandler

cb = AsyncSignedCallbackHandler("moss:bot:async")
result = await chain.ainvoke(input, config={"callbacks": [cb]})

Signed Events

The handler signs outputs from:

• on_llm_end - LLM generation complete
• on_chain_end - Chain execution complete
• on_tool_end - Tool execution complete
• on_agent_finish - Agent finished
• on_retriever_end - Retriever returned documents