DataLlama

DataLlama or LlamaIndex (GPT Index) is a powerful data framework designed to augment your LLM (Large Language Model) applications. It simplifies building applications by providing essential tools to ingest, structure, and query your data efficiently.

Getting Started

There are two primary ways to use LlamaIndex:

1. Starter

The simplest way is to install the starter package, which includes core functionalities and essential integrations.

sh pip install llama-index

2. Customized

For greater flexibility, install the core package and select your integrations from LlamaHub.

```sh pip install llama-index-core

Add specific integrations as needed

```

Import Patterns

python from llama_index.core.llms import LLM # core module from llama_index.llms.openai import OpenAI # integration

Important Links

• Documentation
• Typescript/Javascript version
• Twitter
• LinkedIn
• Discord Community
• Reddit Community

Ecosystem

• LlamaHub: Community-driven library of data loaders (Explore)
• LlamaLab: Advanced AGI projects built using LlamaIndex (GitHub)

Overview

LlamaIndex provides:

• Data connectors for ingesting various formats (APIs, PDFs, SQL, etc.)
• Structures (indices, graphs) for efficient querying
• Advanced retrieval and query interfaces
• Easy integration with other frameworks (LangChain, Flask, Docker, etc.)

Whether youre a beginner or advanced developer, LlamaIndex offers tools to simplify or deeply customize your LLM applications.

Documentation

Find comprehensive guides, tutorials, and references in the official documentation.

Quick Examples

Basic Setup with OpenAI

```python import os os.environ["OPENAIAPIKEY"] = "YOUROPENAIAPI_KEY"

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader

documents = SimpleDirectoryReader("YOURDATADIRECTORY").loaddata() index = VectorStoreIndex.fromdocuments(documents) ```

Using Non-OpenAI Models (e.g., Llama 2 via Replicate)

```python import os os.environ["REPLICATEAPITOKEN"] = "YOURREPLICATEAPI_TOKEN"

from llamaindex.core import Settings, VectorStoreIndex, SimpleDirectoryReader from llamaindex.embeddings.huggingface import HuggingFaceEmbedding from llama_index.llms.replicate import Replicate from transformers import AutoTokenizer

Settings.llm = Replicate( model="meta/llama-2-7b-chat:version", temperature=0.01, additionalkwargs={"topp": 1, "maxnewtokens": 300}, )

Settings.tokenizer = AutoTokenizer.frompretrained("NousResearch/Llama-2-7b-chat-hf") Settings.embedmodel = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")

documents = SimpleDirectoryReader("YOURDATADIRECTORY").loaddata() index = VectorStoreIndex.fromdocuments(documents) ```

Query Example

python query_engine = index.as_query_engine() response = query_engine.query("YOUR_QUESTION")

Persisting and Reloading Data

Persist: python index.storage_context.persist()

Reload: ```python from llamaindex.core import StorageContext, loadindexfromstorage

storagecontext = StorageContext.fromdefaults(persistdir="./storage") index = loadindexfromstorage(storage_context) ```