LlamaIndex Legacy

0.9.48.post4 · deprecated · verified Wed Apr 15

LlamaIndex Legacy is a compatibility package that contains the pre-v0.10.0 monolithic codebase of LlamaIndex, a data framework designed to build LLM applications by connecting large language models with external data sources. It is currently at version 0.9.48.post4 and is maintained for users who have not yet migrated to the modular LlamaIndex v0.10+ architecture. While the broader LlamaIndex ecosystem has a rapid release cadence, this specific legacy package receives minimal updates as it is a bridge for older codebases.

Warnings

breaking The primary LlamaIndex library underwent a significant modularization in version 0.10.0, splitting into `llama-index-core` and numerous integration-specific packages. `llama-index-legacy` is provided as a temporary bridge for pre-v0.10.0 code, but new development should use the modular `llama-index` packages.
Fix: Migrate your codebase to use `llama-index` (v0.10+) and its modular components, importing from `llama_index.core` and specific integration packages (e.g., `llama_index.llms.openai`). A CLI tool is available for automated import updates.
deprecated The `ServiceContext` object has been deprecated in favor of a global `Settings` object or local configurations passed directly to APIs. Many agent-related classes (e.g., `AgentRunner`, `FunctionCallingAgent`) are also deprecated in favor of `AgentWorkflow`.
Fix: Replace `ServiceContext` usage with `Settings` for global configuration or pass parameters directly. Update deprecated agent classes to use the `AgentWorkflow` abstraction.
deprecated The `llama-index-legacy` package itself is deprecated and has been removed from the main LlamaIndex repository. This implies that it will receive minimal or no new features and potentially limited maintenance.
Fix: Users are strongly encouraged to migrate their applications to the latest `llama-index` package and its modular components to ensure access to new features, bug fixes, and ongoing support.
gotcha While `llama-index-legacy` explicitly supports Python >=3.8.1 and <4.0, the broader `llama-index` ecosystem has begun deprecating Python 3.9 in recent releases of its modular packages. Users planning a migration should be aware of future Python version requirements.
Fix: When migrating to the latest LlamaIndex, ensure your environment uses a supported Python version, preferably 3.10 or newer, to align with ongoing development.

Install

pip install llama-index-legacy Install LlamaIndex Legacy

Imports

VectorStoreIndex
```
from llama_index.legacy import VectorStoreIndex
```
For users sticking to the 0.9.x API, import from `llama_index.legacy`. The new `llama_index` package (v0.10+) uses `llama_index.core` for core abstractions.
SimpleDirectoryReader
```
from llama_index.legacy import SimpleDirectoryReader
```
Used for loading documents from a local directory.
Replicate
```
from llama_index.legacy.llms import Replicate
```
Example for importing a specific LLM integration from the legacy package structure.

Quickstart

This quickstart demonstrates how to load documents from a local directory, create a vector store index, and query it using the legacy LlamaIndex API. Ensure you have an `OPENAI_API_KEY` set as an environment variable and a `data` directory with at least one text file.

import os
from llama_index.legacy import VectorStoreIndex, SimpleDirectoryReader

# Set your OpenAI API key as an environment variable
os.environ["OPENAI_API_KEY"] = os.environ.get("OPENAI_API_KEY", "YOUR_OPENAI_API_KEY")

# Create a 'data' directory and place some text files inside it
# e.g., echo "The quick brown fox jumps over the lazy dog." > data/example.txt

# Load documents from the specified directory
documents = SimpleDirectoryReader("data").load_data()

# Create a vector store index from the loaded documents
index = VectorStoreIndex.from_documents(documents)

# Create a query engine and query the index
query_engine = index.as_query_engine()
response = query_engine.query("What did the fox do?")

print(response.response)

view raw JSON →