How to debug LangChain chain
LangChain chain, enable verbose logging by setting LANGCHAIN_HANDLER to langchain and use chain.debug = True to trace execution. Inspect intermediate outputs and exceptions by wrapping chain calls in try-except blocks and printing chain.run() results step-by-step.code_error chain.debug = True and enable LangChain logging to trace chain execution and identify errors.Why this happens
LangChain chains can fail silently or produce unexpected outputs due to misconfigured inputs, missing environment variables, or incorrect prompt templates. For example, calling chain.run() without proper input keys or with invalid data types triggers errors or empty results. Without debug info, it’s hard to pinpoint the failure cause.
Typical error output includes exceptions like KeyError, ValueError, or unexpected empty strings.
from langchain.chains import LLMChain
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(model_name="gpt-4o", temperature=0)
chain = LLMChain(llm=llm, prompt=some_prompt_template)
# Missing required input key 'question'
result = chain.run({}) # Raises KeyError or returns empty outputKeyError: 'question'
The fix
Enable debugging by setting chain.debug = True to log intermediate steps and outputs. Also, configure Python logging for LangChain to see detailed tracebacks and prompt expansions. Wrap calls in try-except blocks to catch and print exceptions.
This approach reveals exactly where the chain fails and what data it processes, making it easier to fix input or prompt issues.
import logging
from langchain.chains import LLMChain
from langchain_openai import ChatOpenAI
logging.basicConfig(level=logging.DEBUG)
llm = ChatOpenAI(model_name="gpt-4o", temperature=0)
chain = LLMChain(llm=llm, prompt=some_prompt_template)
chain.debug = True
try:
result = chain.run({"question": "What is LangChain?"})
print("Result:", result)
except Exception as e:
print("Error during chain execution:", e)DEBUG:langchain:Prompt after formatting: What is LangChain? Result: LangChain is a framework for building AI applications.
Preventing it in production
Use structured error handling with retries and fallbacks around chain calls to handle transient API errors gracefully. Validate inputs before passing them to chains to avoid runtime exceptions. Enable logging in staging environments to catch issues early. Consider unit testing chains with mock LLMs to verify logic without incurring API costs.
Key Takeaways
- Set
chain.debug = Trueto enable detailed LangChain execution logs. - Wrap chain calls in try-except blocks to catch and diagnose errors.
- Validate all inputs before running chains to prevent common runtime errors.