vectordbpipe 0.1.3

A modular text embedding and vector database pipeline for local and cloud vector stores.

vectorDBpipe

A Modular, End-to-End RAG Pipeline for Production-Ready Vector Search.

vectorDBpipe is a robust framework designed to automate the heavy lifting of building RAG (Retrieval-Augmented Generation) systems. It seamlessly handles data ingestion, text cleaning, semantic embedding, and storage in modern vector databases.

---

🎯 Project Objectives

Building a vector search system often involves writing the same "glue code" over and over again:

1. Parsing PDFs, Word docs, and Text files.
2. Cleaning funny characters and whitespace.
3. Chunking long text so it fits into context windows.
4. Batching embeddings to avoid OOM (Out-of-Memory) errors.
5. Creating and managing database indexes.

vectorDBpipe solves this. It is a "download-and-go" architected solution that reduces weeks of boilerplate work into a standardized config.yaml file.

Ideal for:

• AI Engineers building internal RAG tools.
• Developers needing to "chat with their data" instantly.
• Researchers testing different embedding models or databases (switch from Chroma to Pinecone in 1 line).

---

🛠️ Tech Stack & Architecture

This project utilizes best-in-class open-source technologies:

• Ingestion: PyMuPDF (PDF), python-docx (DOCX), pandas (CSV), BeautifulSoup (HTML).
• Vectorization: sentence-transformers (HuggingFace compatible).
• Vector Database:
  • ChromaDB (Local, persistent, file-based).
  • Pinecone (Serverless, Cloud-native v3.0+).
  • FAISS (Via underlying libraries or custom adapters).
• Orchestration: Custom batch-processing Pipeline.

🏗️ Architecture Flow

graph LR
    A[Raw Data Folder] --> B(DataLoader);
    B --> C{Cleaner & Chunker};
    C --Batching--> D[Embedder Model];
    D --> E[(Vector Database)];
    E --> F[Semantic Search API];
    F --> G[RAG Application / Chatbot];

---

💡 Use Cases

1. Enterprise Knowledge Base

Company wikis, PDFs, and policy documents are scattered.

• Solution: Point vectorDBpipe to the shared drive. It indexes 10,000+ docs into Pinecone.
• Result: Employees get instant, semantic answers ("What is the travel policy?") instead of keyword search.

2. Legal / Medical Document Search

Long documents need to be split intelligently.

• Solution: Use the standardized chunker (e.g., 512 tokens with overlap).
• Result: Retrieval finds the exact paragraph containing the clause or diagnosis.

3. Rapid Prototype for RAG

You have a hackathon idea but don't want to spend 4 hours setting up FAISS.

• Solution: pip install vectordbpipe -> pipeline.run().
• Result: Working MVP in 5 minutes.

---

📦 Installation

Install the package directly from PyPI:

pip install vectordbpipe

🔧 Windows Users (DLL Error Constraints)

If you encounter WinError 1114 or DLL initialization errors with Torch, install the CPU-optimized binaries:

pip install -r requirements-cpu.txt

(This forces intel-openmp and CPU-only libraries to ensure stability on non-CUDA machines).

---

⚙️ Configuration

Control your entire pipeline via config.yaml. No need to touch the code.

# vectorDBpipe/config/config.yaml

paths:
  data_dir: "data/"  # Folder containing your .pdf, .txt, .docx files

model:
  name: "sentence-transformers/all-MiniLM-L6-v2" # Any HF model
  batch_size: 32

vector_db:
  type: "pinecone"   # Options: "chroma" or "pinecone"
  index_name: "my-knowledge-base"
  # For Chroma, use:
  # persist_directory: "data/chroma_store"

🔐 Credentials

Do NOT hardcode API keys. The system looks for environment variables:

Linux/Mac:

export PINECONE_API_KEY="your-secret-key"

Windows PowerShell:

$env:PINECONE_API_KEY="your-secret-key"

---

🚀 Step-by-Step Demo: The "10-Line" RAG Pipeline

This script detects all files in your data/ folder, processes them in memory-safe batches, and makes them searchable.

from vectorDBpipe.pipeline.text_pipeline import TextPipeline

# ---------------------------------------------------------
# STEP 1: Initialize Pipeline
# ---------------------------------------------------------
# Reads config.yaml, sets up logging, connects to DB (Pinecone/Chroma)
pipeline = TextPipeline()

# ---------------------------------------------------------
# STEP 2: Ingest Data (The "Magic" Step)
# ---------------------------------------------------------
# Loops through all files in 'data_dir', cleans text, splits into 
# 512-token chunks, embeds them using HuggingFace model, 
# and uploads to DB in batches of 100 to save RAM.
pipeline.process(batch_size=100)

# ---------------------------------------------------------
# STEP 3: Semantic Search
# ---------------------------------------------------------
query = "How does vectorDBpipe reduce workload?"
results = pipeline.search(query, top_k=3)

print("--- Search Results ---")
for match in results:
    # Metadata contains the original text chunk and source file name
    print(f"Source: {match.get('metadata', {}).get('source', 'unknown')}")
    print(f"Content: {match.get('metadata', {}).get('text', '')[:200]}...\n")

---

🧠 Deep Dive: How It Reduces Work

Before vectorDBpipe vs. After

Feature	The "Hard Way" (Manual)	The vectorDBpipe Way
PDF Parsing	Write fitz loops, handle exceptions, merge pages.	loader.load_data() handles PDF, DOCX, TXT, HTML auto-magically.
Chunking	Write regex wrappers, handle overlaps, off-by-one errors.	chunk_text(text, chunk_size=512) built-in utility.
Embeddings	Manually loop model.encode(), manage tensors.	Embedder class abstracts this away (Mock fallback included).
Scalability	"Out of Memory" when loading 1000 PDFs.	Batch Processing built-in. Flushes data every 100 chunks.
DB Switching	Rewrite insert logic for Pinecone connection vs Chroma.	Change type: pinecone in YAML. Done.

Code Snippet: Scalable Batch Processing

Use the new process(batch_size=N) method introduced in v0.1.3 to handle massive datasets.

# Even if you have 10GB of text files, this won't crash your RAM.
pipeline.process(batch_size=50) 

---

📁 Project Structure

vectorDBpipe/
├── config/             # YAML configuration
├── data/               # Drop your raw files here
├── vectorDBpipe/
│   ├── data/           # Loader logic (PDF/DOCX/TXT parsers)
│   ├── embeddings/     # SentenceTransformer wrapper
│   ├── pipeline/       # The "Brain" (Process & Search flow)
│   └── vectordb/       # Store adapters (Chroma/Pinecone)
└── requirements.txt    # Production deps

---

🤝 Contributing

We welcome issues and PRs!

• Report Bugs: Create an issue on GitHub.
• Updates: We are working on adding Qdrant and Weaviate support in v0.2.0.

Author: Yash Desai
License: MIT