| name | langchain-common-errors |
| description | Diagnose and fix common LangChain errors and exceptions. Use when encountering LangChain errors, debugging failures, or troubleshooting integration issues. Trigger with phrases like "langchain error", "langchain exception", "debug langchain", "langchain not working", "langchain troubleshoot". |
| allowed-tools | Read, Write, Edit, Grep |
| version | 1.0.0 |
| license | MIT |
| author | Jeremy Longshore <jeremy@intentsolutions.io> |
LangChain Common Errors
Overview
Quick reference for diagnosing and resolving the most common LangChain errors.
Prerequisites
- LangChain installed and configured
- Access to application logs
- Understanding of your LangChain implementation
Error Reference
Authentication Errors
openai.AuthenticationError: Incorrect API key provided
# Cause: Invalid or missing API key
# Solution:
import os
os.environ["OPENAI_API_KEY"] = "sk-..." # Set correct key
# Verify key is loaded
from langchain_openai import ChatOpenAI
llm = ChatOpenAI() # Will raise error if key invalid
anthropic.AuthenticationError: Invalid x-api-key
# Cause: Anthropic API key not set or invalid
# Solution:
os.environ["ANTHROPIC_API_KEY"] = "sk-ant-..."
# Or pass directly
from langchain_anthropic import ChatAnthropic
llm = ChatAnthropic(api_key="sk-ant-...")
Import Errors
ModuleNotFoundError: No module named 'langchain_openai'
# Cause: Provider package not installed
# Solution:
pip install langchain-openai
# For other providers:
pip install langchain-anthropic
pip install langchain-google-genai
pip install langchain-community
ImportError: cannot import name 'ChatOpenAI' from 'langchain'
# Cause: Using old import path (pre-0.2.0)
# Old (deprecated):
from langchain.chat_models import ChatOpenAI
# New (correct):
from langchain_openai import ChatOpenAI
Rate Limiting
openai.RateLimitError: Rate limit reached
# Cause: Too many API requests
# Solution: Implement retry with backoff
from langchain_openai import ChatOpenAI
from tenacity import retry, wait_exponential, stop_after_attempt
@retry(wait=wait_exponential(min=1, max=60), stop=stop_after_attempt(5))
def call_with_retry(llm, prompt):
return llm.invoke(prompt)
# Or use LangChain's built-in retry
llm = ChatOpenAI(max_retries=3)
Output Parsing Errors
OutputParserException: Failed to parse output
# Cause: LLM output doesn't match expected format
# Solution 1: Use with_retry
from langchain.output_parsers import RetryOutputParser
parser = RetryOutputParser.from_llm(parser=your_parser, llm=llm)
# Solution 2: Use structured output (more reliable)
from pydantic import BaseModel
class Output(BaseModel):
answer: str
llm_with_structure = llm.with_structured_output(Output)
ValidationError: field required
# Cause: Pydantic model validation failed
# Solution: Make fields optional or provide defaults
from pydantic import BaseModel, Field
from typing import Optional
class Output(BaseModel):
answer: str
confidence: Optional[float] = Field(default=None)
Chain Errors
ValueError: Missing required input keys
# Cause: Input dict missing required variables
# Debug:
prompt = ChatPromptTemplate.from_template("Hello {name}, you are {age}")
print(prompt.input_variables) # ['name', 'age']
# Solution: Provide all required keys
chain.invoke({"name": "Alice", "age": 30})
TypeError: Expected mapping type as input
# Cause: Passing wrong input type
# Wrong:
chain.invoke("hello")
# Correct:
chain.invoke({"input": "hello"})
Agent Errors
AgentExecutor: max iterations reached
# Cause: Agent stuck in loop
# Solution: Increase iterations or improve prompts
agent_executor = AgentExecutor(
agent=agent,
tools=tools,
max_iterations=20, # Increase from default 15
early_stopping_method="force" # Force stop after max
)
ToolException: Tool execution failed
# Cause: Tool raised an exception
# Solution: Add error handling in tool
@tool
def my_tool(input: str) -> str:
"""Tool description."""
try:
# Tool logic
return result
except Exception as e:
return f"Tool error: {str(e)}"
Memory Errors
KeyError: 'chat_history'
# Cause: Memory key mismatch
# Solution: Ensure consistent key names
prompt = ChatPromptTemplate.from_messages([
MessagesPlaceholder(variable_name="chat_history"), # Match this
("human", "{input}")
])
# When invoking:
chain.invoke({
"input": "hello",
"chat_history": [] # Must match placeholder name
})
Debugging Tips
Enable Verbose Mode
import langchain
langchain.debug = True # Shows all chain steps
# Or per-component
agent_executor = AgentExecutor(verbose=True)
Trace with LangSmith
# Set environment variables
os.environ["LANGCHAIN_TRACING_V2"] = "true"
os.environ["LANGCHAIN_API_KEY"] = "your-langsmith-key"
os.environ["LANGCHAIN_PROJECT"] = "my-project"
# All chains automatically traced
Check Version Compatibility
pip show langchain langchain-core langchain-openai
# Ensure versions are compatible:
# langchain >= 0.3.0
# langchain-core >= 0.3.0
# langchain-openai >= 0.2.0
Resources
Next Steps
For complex debugging, use langchain-debug-bundle to collect evidence.