LlamaIndex

0.14.15 · active · verified Sat Feb 28

LlamaIndex is a data framework for building LLM-powered agents over your data. Specializes in RAG pipelines, document parsing, and agent workflows. Core package is llama-index-core. Integrations are separate packages installed from LlamaHub.

Warnings

breaking ServiceContext is fully removed. All code using ServiceContext.from_defaults() will fail.
Fix: Use Settings global object: from llama_index.core import Settings
breaking AgentRunner, AgentWorker, FunctionCallingAgent, OpenAIAgent, StructuredAgentPlanner all removed.
Fix: Migrate to AgentWorkflow, FunctionAgent, CodeActAgent, or ReActAgent (new workflow-based)
breaking QueryPipeline class removed.
Fix: Use Workflows-based approach instead
breaking llama-index-legacy package deprecated and removed from repository.
Fix: Migrate fully to llama-index-core and current integration packages
breaking GPTVectorStoreIndex, GPTSimpleKeywordTableIndex and all GPT-prefixed index names removed.
Fix: Use VectorStoreIndex, SimpleKeywordTableIndex etc.
gotcha pip install llama-index alone installs OpenAI integrations by default. For other providers install llama-index-core plus the specific integration package.
Fix: pip install llama-index-core llama-index-llms-anthropic for Claude access
gotcha index.as_chat_engine() default changed to CondensePlusContextChatEngine in 0.14.x.
Fix: Explicitly pass chat_mode if you relied on previous default behaviour

Install

pip install llama-index Python (starter — includes core + default integrations)
pip install llama-index-core Python (core only — add integrations separately)
pip install llama-index-core llama-index-llms-openai llama-index-embeddings-openai Python (core + OpenAI)

Imports

VectorStoreIndex
```
from llama_index.core import VectorStoreIndex
```
GPTVectorStoreIndex renamed to VectorStoreIndex in 0.10. Old name removed.
OpenAI (LLM)
```
from llama_index.llms.openai import OpenAI
```
LLMs moved to integration packages in 0.10. Install llama-index-llms-openai.
AgentWorkflow
```
from llama_index.core.agent.workflow import AgentWorkflow
```
AgentRunner, AgentWorker, FunctionCallingAgent, ReActAgent (old) all removed. Use AgentWorkflow.
Settings
```
from llama_index.core import Settings
Settings.llm = OpenAI(model='gpt-4o')
```
ServiceContext fully removed. Use Settings global object or pass params directly.

Quickstart

Minimal RAG pipeline using VectorStoreIndex with OpenAI in LlamaIndex 0.14.x.

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings
from llama_index.llms.openai import OpenAI
from llama_index.embeddings.openai import OpenAIEmbedding

Settings.llm = OpenAI(model='gpt-4o')
Settings.embed_model = OpenAIEmbedding(model='text-embedding-3-small')

documents = SimpleDirectoryReader('data').load_data()
index = VectorStoreIndex.from_documents(documents)
query_engine = index.as_query_engine()
response = query_engine.query('What did the author do growing up?')
print(response)

view raw JSON →