lexora 0.1.1

A production-ready, plug-and-play Python SDK for building intelligent RAG systems

Lexora Agentic RAG SDK

Production-ready Agentic RAG SDK with minimal configuration

📖 Full Documentation • 🚀 Getting Started • 💬 Discussions

---

🚀 What is Lexora?

Lexora is a production-ready Agentic RAG (Retrieval-Augmented Generation) SDK that makes it easy to build intelligent applications with semantic search and AI-powered reasoning. With just a few lines of code, you can:

• 📚 Create and manage document corpora
• 🔍 Perform semantic search across your documents
• 🤖 Build AI agents that reason over your data
• 🛠️ Extend functionality with custom tools
• 🎯 Deploy to production with confidence

✨ Key Features

• Zero-Config Setup: Get started in minutes with sensible defaults
• Multiple Vector Databases: Support for FAISS, Pinecone, and Chroma
• Flexible Embeddings: OpenAI, HuggingFace, Gemini, or custom providers
• Flexible LLM Integration: Works with any LLM via LiteLLM
• Built-in RAG Tools: 10+ pre-built tools for document management
• Custom Tool Support: Easily add your own tools
• Production-Ready: Comprehensive error handling, logging, and testing
• Type-Safe: Full type hints and Pydantic validation
• Cost-Effective: Free embedding options available

---

📦 Installation

Prerequisites

• Python 3.8 or higher
• pip or conda package manager

Install from Source

# Clone the repository
git clone https://github.com/yourusername/lexora.git
cd lexora

# Create a virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Install in development mode
pip install -e .

Install from PyPI (Coming Soon)

pip install lexora

---

🎯 Quick Start

Basic Usage

from lexora import RAGAgent

# Initialize the agent with defaults
agent = RAGAgent()

# Create a document corpus
await agent.tool_registry.get_tool("create_corpus").run(
    corpus_name="my_docs",
    description="My document collection"
)

# Add documents
documents = [
    {"content": "Python is a programming language.", "metadata": {"topic": "python"}},
    {"content": "Machine learning is a subset of AI.", "metadata": {"topic": "ml"}}
]

await agent.tool_registry.get_tool("add_data").run(
    corpus_name="my_docs",
    documents=documents
)

# Query your documents
result = await agent.tool_registry.get_tool("rag_query").run(
    corpus_name="my_docs",
    query="What is Python?",
    top_k=5
)

print(result.data["results"])

Using the Agent for Reasoning

# Ask questions and get AI-powered answers
response = await agent.query("Explain machine learning in simple terms")

print(f"Answer: {response.answer}")
print(f"Confidence: {response.confidence}")
print(f"Sources: {len(response.sources)}")

---

📖 Documentation

Table of Contents

1. Installation Guide
2. Configuration
3. Core Concepts
4. RAG Tools
5. Custom Tools
6. Vector Databases
7. LLM Integration
8. Error Handling
9. Best Practices
10. API Reference

---

⚙️ Configuration

Lexora supports multiple configuration methods:

1. Default Configuration (Easiest)

from lexora import RAGAgent

# Uses mock LLM and FAISS vector database
agent = RAGAgent()

2. YAML Configuration

# config.yaml
llm:
  provider: "openai"
  model: "gpt-4"
  api_key: "${OPENAI_API_KEY}"
  temperature: 0.7

vector_db:
  provider: "faiss"
  embedding_model: "text-embedding-ada-002"
  dimension: 1536
  connection_params:
    storage_path: "./vector_storage"

agent:
  max_iterations: 5
  enable_reasoning: true
  log_level: "INFO"

from lexora import RAGAgent

agent = RAGAgent.from_yaml("config.yaml")

3. Environment Variables

# .env file
LEXORA_LLM_PROVIDER=openai
LEXORA_LLM_MODEL=gpt-4
LEXORA_LLM_API_KEY=your-api-key
LEXORA_VECTORDB_PROVIDER=faiss
LEXORA_VECTORDB_EMBEDDING_MODEL=text-embedding-ada-002

from lexora import RAGAgent

agent = RAGAgent.from_env()

4. Programmatic Configuration

from lexora import RAGAgent, LLMConfig, VectorDBConfig, AgentConfig

agent = RAGAgent(
    llm_config=LLMConfig(
        provider="openai",
        model="gpt-4",
        api_key="your-api-key"
    ),
    vector_db_config=VectorDBConfig(
        provider="faiss",
        embedding_model="text-embedding-ada-002",
        dimension=1536
    ),
    agent_config=AgentConfig(
        max_iterations=5,
        enable_reasoning=True
    )
)

---

🎨 Embedding Options

Important: You are NOT limited to OpenAI embeddings! Lexora supports multiple embedding providers.

Available Options

Provider	Cost	Quality	Privacy	Best For
HuggingFace	Free	High	✅ Local	Production (recommended)
OpenAI	Paid	Highest	❌ Cloud	Enterprise
Gemini	Free tier	High	❌ Cloud	Gemini users
Mock	Free	Low	✅ Local	Testing

Quick Examples

1. Free Local Embeddings (Recommended)

# Install sentence-transformers
# pip install sentence-transformers

from lexora import RAGAgent
from lexora.models.config import VectorDBConfig

agent = RAGAgent(
    vector_db_config=VectorDBConfig(
        provider="faiss",
        dimension=384,  # all-MiniLM-L6-v2 dimension
        connection_params={
            "index_type": "Flat",
            "persist_directory": "./vector_db"
        }
    )
)

2. OpenAI Embeddings

from lexora import RAGAgent
from lexora.models.config import VectorDBConfig

agent = RAGAgent(
    vector_db_config=VectorDBConfig(
        provider="faiss",
        dimension=1536,  # OpenAI dimension
        connection_params={
            "embedding_model": "text-embedding-ada-002",
            "openai_api_key": "your-key"
        }
    )
)

3. Custom Embedding Provider

from lexora.utils.embeddings import BaseEmbeddingProvider
from sentence_transformers import SentenceTransformer

class HuggingFaceProvider(BaseEmbeddingProvider):
    def __init__(self, model_name="all-MiniLM-L6-v2"):
        self.model = SentenceTransformer(model_name)
    
    async def generate_embedding(self, text: str):
        return self.model.encode(text).tolist()
    
    def get_dimension(self) -> int:
        return self.model.get_sentence_embedding_dimension()

📚 Full Embedding Guide - Detailed documentation on all embedding options

---

🧩 Core Concepts

Document Corpus

A corpus is a collection of documents that can be searched semantically.

# Create a corpus
await agent.tool_registry.get_tool("create_corpus").run(
    corpus_name="knowledge_base",
    description="Company knowledge base",
    metadata={"department": "engineering"}
)

Documents

Documents are the basic unit of information in Lexora.

document = {
    "content": "Your document text here",
    "metadata": {
        "source": "documentation",
        "author": "John Doe",
        "date": "2024-01-01"
    }
}

Semantic Search

Search documents by meaning, not just keywords.

results = await agent.tool_registry.get_tool("rag_query").run(
    corpus_name="knowledge_base",
    query="How do I deploy to production?",
    top_k=5,
    min_score=0.7
)

---

🛠️ RAG Tools

Lexora comes with 10+ built-in tools:

Core Tools

Tool	Description
create_corpus	Create a new document corpus
add_data	Add documents to a corpus
rag_query	Search documents semantically
list_corpora	List all available corpora
get_corpus_info	Get detailed corpus information
delete_corpus	Delete a corpus
delete_document	Delete a specific document
update_document	Update an existing document
bulk_add_data	Add large batches of documents
health_check	Check system health

Tool Usage Examples

See examples/ directory for detailed examples of each tool.

---

🔧 Custom Tools

Extend Lexora with your own tools:

from lexora import BaseTool, ToolParameter

class WeatherTool(BaseTool):
    @property
    def name(self) -> str:
        return "get_weather"
    
    @property
    def description(self) -> str:
        return "Get current weather for a location"
    
    @property
    def version(self) -> str:
        return "1.0.0"
    
    def _setup_parameters(self) -> None:
        self._parameters = [
            ToolParameter(
                name="location",
                type="string",
                description="City name",
                required=True
            )
        ]
    
    async def _execute(self, location: str, **kwargs):
        # Your implementation here
        return {"temperature": 72, "condition": "sunny"}

# Register the tool
agent.add_tool(WeatherTool())

---

💾 Vector Databases

FAISS (Default)

from lexora import RAGAgent, VectorDBConfig

agent = RAGAgent(
    vector_db_config=VectorDBConfig(
        provider="faiss",
        embedding_model="text-embedding-ada-002",
        dimension=1536,
        connection_params={"storage_path": "./faiss_storage"}
    )
)

Pinecone

agent = RAGAgent(
    vector_db_config=VectorDBConfig(
        provider="pinecone",
        embedding_model="text-embedding-ada-002",
        dimension=1536,
        connection_params={
            "api_key": "your-pinecone-key",
            "environment": "us-west1-gcp"
        }
    )
)

Chroma

agent = RAGAgent(
    vector_db_config=VectorDBConfig(
        provider="chroma",
        embedding_model="text-embedding-ada-002",
        dimension=1536,
        connection_params={"persist_directory": "./chroma_storage"}
    )
)

---

🤖 LLM Integration

Lexora uses LiteLLM for universal LLM support:

OpenAI

from lexora import LLMConfig

llm_config = LLMConfig(
    provider="openai",
    model="gpt-4",
    api_key="your-api-key",
    temperature=0.7
)

Anthropic Claude

llm_config = LLMConfig(
    provider="anthropic",
    model="claude-3-opus-20240229",
    api_key="your-api-key"
)

Azure OpenAI

llm_config = LLMConfig(
    provider="azure",
    model="gpt-4",
    api_key="your-api-key",
    api_base="https://your-resource.openai.azure.com/"
)

---

🚨 Error Handling

Lexora provides structured error handling:

result = await agent.tool_registry.get_tool("rag_query").run(
    corpus_name="nonexistent",
    query="test"
)

if result.status == "error":
    print(f"Error: {result.error}")
    # Error includes context and suggestions

All errors include:

• Error code
• Descriptive message
• Context information
• Helpful suggestions

---

📚 Examples

Check out the examples/ directory for complete examples:

• 01_quick_start.py - Basic usage
• 02_custom_configuration.py - Configuration options
• 03_corpus_management.py - Managing corpora
• 04_custom_tools.py - Creating custom tools
• rag_tools_demo.py - All RAG tools
• rag_agent_with_real_embeddings.py - Production setup

---

🧪 Testing

Run the test suite:

# Run all tests
python run_tests.py

# Run specific test file
python tests/test_error_handling.py

# Run with pytest
pytest tests/ -v

---

📊 Performance

• Query Speed: < 1ms for small corpora
• Batch Processing: 12,000+ documents/second
• Concurrent Queries: 10 queries in 5ms
• Memory Efficient: Handles 200+ documents in batches

---

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

🆘 Support

• Documentation: Full docs
• Examples: Example code
• Issues: GitHub Issues
• Discussions: GitHub Discussions

---

🗺️ Roadmap

• PyPI package distribution
• Additional vector database support
• Streaming responses
• Multi-modal support (images, audio)
• Advanced caching strategies
• Distributed deployment support

---

🙏 Acknowledgments

Built with:

• LiteLLM - Universal LLM interface
• FAISS - Vector similarity search
• Pydantic - Data validation
• Chroma - Vector database

---

Made with ❤️ by the Lexora Team

⭐ Star us on GitHub • 📖 Read the Docs • 🐦 Follow us on Twitter