Skip to main content

​
Overview

Mini RAG provides a modular, Pythonic API for building RAG (Retrieval-Augmented Generation) applications. This reference documentation covers all classes, methods, and configuration options.

AgenticRAG

Main RAG orchestrator class

Configuration

Configuration dataclasses

Vector Store

Milvus vector storage

Rerankers

Reranking strategies

​
Installation

uv add mini-rag

​
Basic Usage

import os
from mini import (
    AgenticRAG,
    LLMConfig,
    RetrievalConfig,
    EmbeddingModel,
    VectorStore
)

# Initialize components
embedding_model = EmbeddingModel()
vector_store = VectorStore(
    uri=os.getenv("MILVUS_URI"),
    token=os.getenv("MILVUS_TOKEN"),
    collection_name="my_docs",
    dimension=1536
)

# Create RAG instance
rag = AgenticRAG(
    vector_store=vector_store,
    embedding_model=embedding_model,
    llm_config=LLMConfig(model="gpt-4o-mini"),
    retrieval_config=RetrievalConfig(
        top_k=10,
        rerank_top_k=3,
        use_query_rewriting=True,
        use_reranking=True
    )
)

# Index documents
rag.index_document("document.pdf")

# Query
response = rag.query("What is this about?")
print(response.answer)

​
Core Concepts

​
Configuration-Based API

Mini RAG uses a clean, configuration-based API with four main configuration classes:
  • LLMConfig: Language model settings
  • RetrievalConfig: Retrieval behavior
  • RerankerConfig: Reranking strategy
  • ObservabilityConfig: Monitoring and tracing

​
Modular Components

Use individual components or the complete pipeline:
  • DocumentLoader: Multi-format document loading
  • Chunker: Smart text chunking with Chonkie
  • EmbeddingModel: OpenAI-compatible embeddings
  • VectorStore: Milvus vector storage
  • AgenticRAG: Complete RAG orchestrator

​
Import Reference

# Main classes
from mini import AgenticRAG, EmbeddingModel, VectorStore

# Configuration
from mini import (
    LLMConfig,
    RetrievalConfig,
    RerankerConfig,
    ObservabilityConfig
)

# Individual components
from mini.loader import DocumentLoader
from mini.chunker import Chunker
from mini.reranker import (
    CohereReranker,
    SentenceTransformerReranker,
    LLMReranker,
    create_reranker
)

​
Type Hints

Mini RAG is fully typed with Pydantic for validation and IDE support: 
from typing import Optional, List, Dict, Any
from mini import AgenticRAG, RAGResponse

def process_query(rag: AgenticRAG, question: str) -> RAGResponse:
    """Type-safe query processing."""
    response: RAGResponse = rag.query(question)
    return response

​
Error Handling

All methods raise standard Python exceptions: 
from mini import AgenticRAG

try:
    rag = AgenticRAG(vector_store=store, embedding_model=model)
    response = rag.query("question")
except FileNotFoundError:
    print("Document not found")
except ValueError:
    print("Invalid configuration")
except Exception as e:
    print(f"Unexpected error: {e}")

​
Next Steps

Core Classes

Explore the main classes

Configuration

Learn about configuration options

Examples

See practical examples