LlamaIndex LangChain LLMs Integration

0.8.0 · active · verified Thu Apr 16

The `llama-index-llms-langchain` library provides a seamless integration layer, allowing users to leverage LangChain's extensive collection of Large Language Models (LLMs) within the LlamaIndex framework. It acts as a bridge, enabling LangChain LLM instances to conform to LlamaIndex's LLM interface. The current version is `0.8.0`, and it receives regular updates aligned with LlamaIndex core releases.

Common errors

```
ModuleNotFoundError: No module named 'llama_index.llms.langchain'
```
cause The `llama-index-llms-langchain` package is not installed or the import path is incorrect.

fix Install the package: `pip install llama-index-llms-langchain`. If installed, verify the import path is `from llama_index.llms.langchain import LangChainLLM`.
```
ImportError: cannot import name 'LangChainLLM' from 'llama_index.integrations.llms.langchain'
```
cause Using an outdated import path for LlamaIndex versions v0.10 and later.

fix Update your import statement to `from llama_index.llms.langchain import LangChainLLM`.
```
AttributeError: 'ChatOpenAI' object has no attribute 'complete' (or 'chat')
```
cause A raw LangChain LLM object is being used where LlamaIndex expects its own `LLM` interface. The `LangChainLLM` wrapper was not applied.

fix You must wrap the LangChain LLM instance with `LangChainLLM` before passing it to LlamaIndex components. Example: `lc_llm = ChatOpenAI(...); llm = LangChainLLM(llm=lc_llm)`.

Warnings

breaking LlamaIndex v0.10+ introduced a significant package restructuring. Integration packages like `llama-index-llms-langchain` moved from the `llama_index.integrations` namespace to dedicated namespaces like `llama_index.llms` or `llama_index.embeddings`.
Fix: Update import statements. For `LangChainLLM`, change `from llama_index.integrations.llms.langchain import LangChainLLM` to `from llama_index.llms.langchain import LangChainLLM`.
gotcha This package (`llama-index-llms-langchain`) only provides the wrapper. You still need to install the specific LangChain provider packages (e.g., `langchain-openai`, `langchain-anthropic`, `langchain-google-genai`) for the LLMs you intend to use.
Fix: After installing `llama-index-llms-langchain`, run `pip install langchain-<provider>` for your chosen LLM provider (e.g., `pip install langchain-openai`).
gotcha Ensure compatible versions of `llama-index-core` and `langchain` (and its sub-packages like `langchain-core`, `langchain-openai`). Version mismatches can lead to `ImportError` or `AttributeError`.
Fix: Refer to the `pyproject.toml` or `setup.py` of `llama-index-llms-langchain` on GitHub for exact dependency version constraints, or update all related packages: `pip install -U llama-index-llms-langchain llama-index-core langchain langchain-core langchain-openai`.

Install

pip install llama-index-llms-langchain llama-index-core Basic Install
pip install llama-index-llms-langchain llama-index-core langchain-openai Install with OpenAI (LangChain provider)

Imports

LangChainLLM
wrong:
```
from llama_index.integrations.llms.langchain import LangChainLLM
```
correct:
```
from llama_index.llms.langchain import LangChainLLM
```
Import path changed with LlamaIndex v0.10+ package restructuring. Older versions used the `integrations` namespace.

Quickstart

This quickstart demonstrates how to instantiate a LangChain LLM (using `ChatOpenAI` as an example), wrap it using `LangChainLLM` from `llama-index-llms-langchain`, and then configure it as the default LLM within LlamaIndex's `Settings`. It shows both completion and chat interactions.

import os
from langchain_openai import ChatOpenAI
from llama_index.llms.langchain import LangChainLLM
from llama_index.core import Settings

# 1. Initialize a LangChain LLM instance
# Make sure to install 'langchain-openai' (pip install langchain-openai)
# and set your OPENAI_API_KEY environment variable.
# Using os.environ.get for safe execution in environments without the key.
lc_llm = ChatOpenAI(temperature=0.0, model="gpt-3.5-turbo", api_key=os.environ.get("OPENAI_API_KEY", "test_key"))

# 2. Wrap the LangChain LLM with LlamaIndex's LangChainLLM wrapper
llm = LangChainLLM(llm=lc_llm)

# 3. Use the wrapped LLM with LlamaIndex
# You can either set it globally or pass it directly to components.
Settings.llm = llm

# Example: Generate a completion
response = Settings.llm.complete("Tell me a short story about a magical cat.")
print(response.text)

# Example: Generate a chat response
from llama_index.core.llms import ChatMessage, MessageRole
chat_response = Settings.llm.chat([
    ChatMessage(role=MessageRole.USER, content="What is the capital of France?")
])
print(chat_response.message.content)

view raw JSON →