vecforge 1.0.0

Forge your vector database. Own it forever. Local-first, encrypted, quantum-inspired.

VecForge

Forge your vector database. Own it forever.

Local-first · Encrypted · Hybrid Search · Zero Cloud Dependency

---

VecForge is a universal, local-first Python vector database with enterprise security, multimodal ingestion, and optional quantum-inspired acceleration.

Built by Suneel Bose K — Founder & CEO, ArcGX TechLabs Private Limited

---

⚡ 5-Line Quickstart

from vecforge import VecForge

db = VecForge("my_vault")
db.add("Patient admitted with type 2 diabetes", metadata={"ward": "7"})
results = db.search("diabetic patient")
print(results[0].text)

That's it. No API keys. No cloud. No config files. Your data stays on your machine.

---

🔥 Why VecForge?

Feature	Pinecone	ChromaDB	VecForge
Local-first	❌ Cloud-only	✅	✅ Always
Encryption at rest	❌	❌	✅ AES-256
Hybrid search	✅	❌	✅ Dense + BM25
Namespace isolation	✅ Cloud	❌	✅ Local
RBAC	✅ Cloud	❌	✅ Built-in
Audit logging	❌	❌	✅ JSONL
Price	$$$$	Free	✅ Free

---

📦 Install

pip install vecforge

From source (development)

git clone https://github.com/bosekarmegam/vecforge.git
cd vecforge
pip install -e ".[dev]"

System Requirements

Windows users: VecForge uses PyTorch under the hood, which requires the Microsoft Visual C++ Redistributable. Install it before running VecForge.

📖 See the full Installation Guide for GPU, encryption, and platform-specific options.

---

🔐 Encrypted Vault

import os
from vecforge import VecForge

db = VecForge(
    "secure_vault",
    encryption_key=os.environ["VECFORGE_KEY"],
    audit_log="audit.jsonl",
    deletion_protection=True,
)
db.add("Top secret patient data", namespace="ward_7")

---

🔍 Hybrid Search

results = db.search(
    "elderly diabetic hip fracture",
    top_k=5,
    alpha=0.7,        # 70% semantic, 30% keyword
    rerank=True,       # cross-encoder precision boost
    namespace="ward_7",
    filters={"year": {"gte": 2023}},
)

📖 See the Search Guide for alpha tuning, metadata operators, and reranking strategies.

---

📄 Auto-Ingest Documents

# Ingest entire directories — auto-detects format
db.ingest("medical_records/")  # PDF, DOCX, TXT, MD, HTML

📖 See the Ingestion Guide for chunking configuration and supported formats.

---

🛡️ Multi-Tenant Namespaces

db.create_namespace("hospital_a")
db.create_namespace("hospital_b")

db.add("Patient data A", namespace="hospital_a")
db.add("Patient data B", namespace="hospital_b")

# Tenant isolation — hospital_a never sees hospital_b's data
results = db.search("patient", namespace="hospital_a")

---

🖥️ CLI

vecforge ingest my_docs/ --vault my.db
vecforge search "diabetes" --vault my.db --top-k 5
vecforge stats my.db
vecforge export my.db -o data.json
vecforge serve --vault my.db --port 8080

📖 See the CLI Reference for all commands and options.

---

🌐 REST API

vecforge serve --vault my.db --port 8080

# Add document
curl -X POST http://localhost:8080/api/v1/add \
  -H "Content-Type: application/json" \
  -d '{"text": "Patient record", "namespace": "default"}'

# Search
curl -X POST http://localhost:8080/api/v1/search \
  -H "Content-Type: application/json" \
  -d '{"query": "diabetes", "top_k": 5}'

📖 See the REST API Reference for all endpoints with request/response schemas.

---

🧪 Examples

Ready-to-run example scripts demonstrating real-world use cases:

Example	Description
🏥 Hospital Search	Medical record search with namespace isolation per ward
⚖️ Legal Documents	NDA and contract search with type/year filtering
🌍 GIS Data Search	Geospatial dataset discovery with USGS, Sentinel, OSM
🤖 RAG Pipeline	Retrieval-Augmented Generation with VecForge as backend
🏢 Multi-Tenant SaaS	Namespace isolation, RBAC, and audit logging demo
💻 Codebase Assistant	Code documentation semantic search

# Run any example
python examples/hospital_search.py
python examples/gis_data_search.py
python examples/rag_pipeline.py

---

📚 Documentation

Getting Started

• ⚡ Quickstart — Get running in 5 minutes
• 📦 Installation — All install options & system requirements

User Guides

• 🧠 Core Concepts — Vaults, namespaces, hybrid search explained
• 🔍 Search Guide — Alpha tuning, filters, reranking
• 🔐 Security Guide — Encryption, RBAC, audit logging
• 📄 Ingestion Guide — PDF, DOCX, HTML, TXT ingestion & chunking

Reference

• 📖 API Reference — Full Python API documentation
• 🖥️ CLI Reference — All CLI commands & options
• 🌐 REST API — FastAPI server endpoints
• ⚙️ Configuration — All config options in one place

---

📊 Benchmarks

Verified on Phase 2 benchmark suite (benchmarks/bench_search.py)

Operation	VecForge (Actual)	North Star Target	Pinecone	ChromaDB
Search 1k docs	0.04ms p50	—	~80ms	~200ms
Search 10k docs	1.63ms p50	—	~80ms	~200ms
Search 100k docs	11.31ms p50 ✅	<15ms	~80ms	~200ms
Ingest 100k docs	2.9M docs/sec	—	Manual	Manual
BM25 Search 10k	9.40ms p50	—	N/A	N/A
Encrypted search	<20ms overhead	<20ms	N/A	N/A

Quality Gates

Check	Result
Ruff lint	✅ All checks passed
Mypy type check	✅ 0 errors (27 files)
Pytest	✅ 128/128 tests pass
Coverage	89% (core modules 85-100%)

---

⚖️ License

Business Source License 1.1 (BSL)

• ✅ Free for personal, research, open-source, and non-commercial use
• ✅ Read, modify, and share freely
• 📋 Commercial use requires a license from ArcGX TechLabs

Contact: [e-mail]

---

Built with ❤️ by Suneel Bose K · ArcGX TechLabs Private Limited