[sessions][langchain] capture session and user ids passed into langchain metadata

[axiomofjoy]

LangChain now supports sessions via a key-value pair passed into their metadata. We should parse this data in our LangChainInstrumentor and assign it to the session.id semantic convention. They have three different keys this can be passed in as, so we need to check to make sure we parse all of them. They might also have support for user IDs via the same mechanism.

Code snippet:

from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.messages import HumanMessage
from langchain_core.runnables import RunnableConfig
import uuid

import os


from openinference.instrumentation.langchain import LangChainInstrumentor

LangChainInstrumentor().instrument()


os.environ["LANGCHAIN_PROJECT"] = "sessions"

model = ChatOpenAI(model="gpt-4")
prompt = ChatPromptTemplate.from_messages([
    ("human", "Hello, how are you?"),
    ("ai", "I'm doing well, thanks!"),
    ("human", "That's good to hear."),
])
print(prompt)
chain = prompt | model
messages = [HumanMessage(content="hi! I'm bob")]
config = RunnableConfig(metadata={"conversation_id": str(uuid.uuid4())})
response = chain.invoke({"messages": messages}, config=config)
print(response)
messages = messages + [response, HumanMessage(content="whats my name")]
response = chain.invoke({"messages": messages}, config=config)
print(response)

[axiomofjoy]

@anticorrelator Where did we land on how to handle the interaction between LangChain's metadata and our own context managers?

[mikeldking]

Need to support thread_id, session_id, conversation_id

[mikeldking]

@fjcasti1 closing