Skip to content

API.md

Latest commit

 

History

History
382 lines (241 loc) · 8.48 KB

API.md

File metadata and controls

382 lines (241 loc) · 8.48 KB

API Reference

📚 Table of Contents

🔧 Core Classes

RAGPDFChatbot

The main application class that orchestrates the entire RAG pipeline.

Constructor

RAGPDFChatbot()

Returns: RAGPDFChatbot instance

Description: Initializes the chatbot with all required components.

Methods

initialize(rebuild_vector_store: bool = False) -> None

Initialize the application and prepare for question answering.

Parameters:

  • rebuild_vector_store (bool, optional): Whether to rebuild vector store from scratch. Defaults to False.

Raises:

  • RuntimeError: If document processing or vector store initialization fails
ask(question: str) -> str

Ask a question using the RAG pipeline.

Parameters:

  • question (str): Question to answer

Returns: str - Answer to the question

Raises:

  • RuntimeError: If application is not initialized
interactive_mode() -> None

Run the application in interactive mode allowing multiple questions in a session.

DocumentProcessor

Handles loading and processing of PDF documents.

Constructor

DocumentProcessor()

Returns: DocumentProcessor instance

Methods

discover_pdf_files() -> List[str]

Discover all PDF files in the dataset directory.

Returns: List[str] - List of file paths to PDF documents

Raises:

  • FileNotFoundError: If no PDF files are found
load_documents() -> List[Dict[str, Any]]

Load all PDF documents from the dataset.

Returns: List[Dict[str, Any]] - List of document objects

Raises:

  • RuntimeError: If no documents are successfully loaded
chunk_documents(documents: List[Dict[str, Any]]) -> List[Dict[str, Any]]

Split documents into chunks for embedding.

Parameters:

  • documents (List[Dict[str, Any]]): List of document objects

Returns: List[Dict[str, Any]] - List of document chunks

process_documents() -> List[Dict[str, Any]]

Complete document processing pipeline.

Returns: List[Dict[str, Any]] - List of processed document chunks

VectorStoreManager

Manages document embedding and vector storage.

Constructor

VectorStoreManager()

Returns: VectorStoreManager instance

Methods

create_vector_store(documents: List[Dict[str, Any]]) -> FAISS

Create and populate vector store with document embeddings.

Parameters:

  • documents (List[Dict[str, Any]]): List of document chunks to embed

Returns: FAISS - Populated vector store

get_retriever() -> Any

Get a retriever configured with current settings.

Returns: Any - Configured retriever object

Raises:

  • RuntimeError: If vector store is not initialized
save_vector_store(path: Optional[str] = None) -> None

Save vector store to local storage.

Parameters:

  • path (str, optional): Path to save vector store

Raises:

  • RuntimeError: If vector store is not initialized
load_vector_store(path: Optional[str] = None) -> FAISS

Load vector store from local storage.

Parameters:

  • path (str, optional): Path to load vector store from

Returns: FAISS - Loaded vector store

vector_store_exists(path: Optional[str] = None) -> bool

Check if vector store exists at specified path.

Parameters:

  • path (str, optional): Path to check

Returns: bool - True if vector store exists

RAGChain

Implements the Retrieval-Augmented Generation pipeline.

Constructor

RAGChain(retriever: Any)

Parameters:

  • retriever (Any): Document retriever object

Returns: RAGChain instance

Methods

ask_question(question: str) -> str

Ask a question using the RAG pipeline.

Parameters:

  • question (str): Question to answer

Returns: str - Answer to the question

Raises:

  • RuntimeError: If RAG chain is not initialized
get_chain() -> Any

Get the RAG chain object.

Returns: Any - RAG chain object

⚙️ Configuration Classes

AppConfig

Main application configuration dataclass.

Attributes

  • embedding: EmbeddingConfig - Embedding model configuration
  • llm: LLMConfig - LLM model configuration
  • vector_store: VectorStoreConfig - Vector store configuration
  • retrieval: RetrievalConfig - Retrieval configuration
  • document_processing: DocumentProcessingConfig - Document processing configuration

EmbeddingConfig

Configuration for embedding models.

Attributes

  • model_name: str - Model name (default: "nomic-embed-text")
  • base_url: str - Ollama base URL (default: "http://localhost:11434")
  • dimension: Optional[int] - Embedding dimension (default: None)

LLMConfig

Configuration for LLM models.

Attributes

  • model_name: str - Model name (default: "llama3.2:3b")
  • base_url: str - Ollama base URL (default: "http://localhost:11434")
  • temperature: float - Generation temperature (default: 0.7)
  • max_tokens: int - Maximum tokens (default: 512)

VectorStoreConfig

Configuration for vector store.

Attributes

  • index_type: str - Index type (default: "flat")
  • metric: str - Distance metric (default: "L2")
  • save_local: bool - Save locally (default: True)
  • local_path: str - Local path (default: "health_supplemets")

RetrievalConfig

Configuration for document retrieval.

Attributes

  • search_type: str - Search type (default: "mmr")
  • k: int - Number of documents (default: 3)
  • fetch_k: int - Number to fetch (default: 100)
  • lambda_mult: float - MMR lambda (default: 1.0)

DocumentProcessingConfig

Configuration for document processing.

Attributes

  • chunk_size: int - Chunk size (default: 1000)
  • chunk_overlap: int - Chunk overlap (default: 100)
  • dataset_path: str - Dataset path (default: "rag-dataset")

🎯 Main Interface

CLI Interface

The application can be run via command line with the following options:

python -m src.main [OPTIONS]

Options:
  --rebuild           Rebuild vector store from scratch
  --interactive       Run in interactive mode
  --question TEXT     Ask a specific question
  --help              Show help message

Programmatic Interface

from src import RAGPDFChatbot

# Initialize chatbot
chatbot = RAGPDFChatbot()
chatbot.initialize()

# Ask questions
answer = chatbot.ask("What are the benefits of BCAA supplements?")
print(answer)

# Interactive mode
chatbot.interactive_mode()

🚨 Exception Handling

RuntimeError

  • Raised when application is not properly initialized
  • Raised when vector store operations fail
  • Raised when document processing fails

FileNotFoundError

  • Raised when no PDF files are found in dataset directory

ValueError

  • Raised for invalid configuration values
  • Raised for invalid input parameters

📝 Examples

Basic Usage

from src import RAGPDFChatbot

# Create and initialize chatbot
chatbot = RAGPDFChatbot()
chatbot.initialize()

# Ask a question
question = "What are the benefits of BCAA supplements?"
answer = chatbot.ask(question)
print(f"Answer: {answer}")

Configuration

import os
from src.config import config

# Set environment variables before importing
os.environ["LLM_MODEL"] = "llama3.2:1b"
os.environ["CHUNK_SIZE"] = "500"

# Config will use the new values
print(f"Model: {config.llm.model_name}")
print(f"Chunk size: {config.document_processing.chunk_size}")

Custom Configuration

from src.config import EmbeddingConfig, LLMConfig, AppConfig

# Create custom config
custom_config = AppConfig(
    embedding=EmbeddingConfig(model_name="custom-embedding"),
    llm=LLMConfig(model_name="custom-llm", temperature=0.5),
    # ... other configs
)

# Use with components
processor = DocumentProcessor()
processor.dataset_path = custom_config.document_processing.dataset_path

Error Handling

from src import RAGPDFChatbot

chatbot = RAGPDFChatbot()

try:
    # This will fail if not initialized
    answer = chatbot.ask("Test question")
except RuntimeError as e:
    print(f"Error: {e}")
    # Initialize first
    chatbot.initialize()
    answer = chatbot.ask("Test question")

🔗 Related Documentation