grepctl 0.2.0

One-command orchestration for multimodal semantic search in BigQuery

grepctl - BigQuery Semantic Search Orchestrator

🚀 One-command multimodal semantic search across your entire data lake using BigQuery ML and Google Cloud AI.

📦 Installation

# Install from PyPI
pip install grepctl

🎯 Quick Start - From Zero to Search in One Command

# Complete setup with automatic data ingestion
grepctl init all --bucket your-bucket --auto-ingest

# Start searching immediately
grepctl search "find all mentions of machine learning"

That's it! The system automatically:

• ✅ Enables all required Google Cloud APIs
• ✅ Creates BigQuery dataset and tables
• ✅ Deploys Vertex AI embedding models
• ✅ Ingests all 8 data modalities from your GCS bucket
• ✅ Generates 768-dimensional embeddings
• ✅ Configures semantic search with VECTOR_SEARCH

📊 What is grepctl?

grepctl is a powerful command-line orchestration tool that transforms your Google Cloud Storage data lake into a searchable knowledge base. It provides a unified interface for searching across 8 different data types:

• 📄 Text & Markdown - Direct content extraction
• 📑 PDF Documents - OCR with Document AI
• 🖼️ Images - Vision API analysis (labels, text, objects, faces)
• 🎵 Audio Files - Speech-to-Text transcription
• 🎬 Video Files - Video Intelligence analysis
• 📊 JSON & CSV - Structured data parsing

All searchable through semantic understanding, not just keywords!

🏗️ Architecture Overview

🛠️ Installation & Setup

Prerequisites

1. Google Cloud Project with billing enabled
2. Python 3.11+
3. gcloud CLI authenticated with appropriate permissions

Install from PyPI

# Install the package
pip install grepctl

# Verify installation
grepctl --help

Install from Source

# Clone repository
git clone https://github.com/gregorymulla/grepctl.git
cd grepctl

# Install with uv (recommended)
uv sync
uv run grepctl --help

# Or install with pip
pip install -e .
grepctl --help

Complete System Setup

Option 1: Fully Automated (Recommended)

# One command does everything!
grepctl init all --bucket your-bucket --auto-ingest

# This single command:
# 1. Enables 7 Google Cloud APIs
# 2. Creates BigQuery dataset and 3 tables
# 3. Deploys 3 Vertex AI models
# 4. Ingests all files from GCS
# 5. Generates embeddings
# 6. Sets up semantic search

Option 2: Step-by-Step Control

# Enable APIs
grepctl apis enable --all

# Initialize BigQuery
grepctl init dataset
grepctl init models

# Ingest data
grepctl ingest all

# Generate embeddings
grepctl index update

# Start searching
grepctl search "your query"

🔍 Using the System

Web User Interface

Start the web server with a beautiful, responsive search interface:

# Start the web server (default port 8000)
grepctl serve

# Start with custom port
grepctl serve --port 3000

# Start with auto-reload for development
grepctl serve --reload

Then open your browser to http://localhost:8000 to access:

• 🎨 Clean, modern interface with your company logo
• 📱 Mobile-friendly responsive design - Works seamlessly on phones, tablets, and desktops
• ⚡ Real-time search across all modalities
• 📊 Live system status and document count
• 🔍 Relevance scoring and result highlighting
• 🌐 Touch-optimized interface for mobile devices

Command Line Interface

# Search across all data
grepctl search "machine learning algorithms"

# Search specific modalities
grepctl search "error handling" -k 20 -m pdf -m markdown

# Check system status
grepctl status

# View all available commands
grepctl --help

grepctl search "bird"

SQL Interface

-- Direct semantic search
WITH query_embedding AS (
  SELECT ml_generate_embedding_result AS embedding
  FROM ML.GENERATE_EMBEDDING(
    MODEL `your-project.mmgrep.text_embedding_model`,
    (SELECT 'machine learning' AS content),
    STRUCT(TRUE AS flatten_json_output)
  )
)
SELECT doc_id, source, text_content, distance AS score
FROM VECTOR_SEARCH(
  TABLE `your-project.mmgrep.search_corpus`,
  'embedding',
  (SELECT embedding FROM query_embedding),
  top_k => 10
)
ORDER BY distance;

Python API (When installed from source)

from grepctl.search.vector_search import SemanticSearch
from grepctl.bigquery.connection import BigQueryClient
from grepctl.config import load_config

# Load configuration
config = load_config()
client = BigQueryClient(config)

# Initialize searcher
searcher = SemanticSearch(client, config)

# Search across all modalities
results = searcher.search(
    query="neural networks",
    top_k=20,
    source_filter=['pdf', 'images'],
    use_rerank=True
)

📈 System Capabilities

Current Status (Production Ready)

• ✅ 425+ documents indexed across 8 modalities
• ✅ 768-dimensional embeddings for semantic understanding
• ✅ Sub-second query response times
• ✅ 100% embedding coverage for all documents
• ✅ 5 Google Cloud APIs integrated
• ✅ Auto-recovery from embedding issues

Supported Operations

Operation	Command	Description
Setup	grepctl init all --auto-ingest	Complete one-command setup
Ingest	grepctl ingest all	Process all file types
Index	grepctl index update	Generate embeddings
Fix	grepctl fix embeddings	Auto-fix dimension issues
Search	grepctl search "query"	Semantic search
Status	grepctl status	System health check

🧰 grepctl Commands

Complete CLI Management

# System initialization
grepctl init all --bucket your-bucket --auto-ingest

# API management
grepctl apis enable --all
grepctl apis check

# Data ingestion
grepctl ingest pdf        # Process PDFs
grepctl ingest images     # Analyze images with Vision API
grepctl ingest audio      # Transcribe audio
grepctl ingest video      # Analyze videos

# Index management
grepctl index rebuild     # Rebuild from scratch
grepctl index update      # Update missing embeddings
grepctl index verify      # Check embedding health

# Troubleshooting
grepctl fix embeddings    # Fix dimension issues
grepctl fix stuck         # Handle stuck processing
grepctl fix validate      # Check data integrity

# Search
grepctl search "query" -k 20 -o json

# Web Interface (NEW!)
grepctl serve             # Start web UI at http://localhost:8000
grepctl serve --port 3000 --theme-config ./my-theme.yaml

Configuration

grepctl uses ~/.grepctl.yaml for configuration:

project_id: your-project
dataset: mmgrep
bucket: your-bucket
location: US
batch_size: 100
chunk_size: 1000

📊 Supported Data Types

Modality	Extensions	Processing Method	Google API Used
Text	.txt, .log	Direct extraction	—
Markdown	.md	Markdown parsing	—
PDF	.pdf	OCR extraction	Document AI
Images	.jpg, .png, .gif	Visual analysis	Vision API
Audio	.mp3, .wav, .m4a	Transcription	Speech-to-Text
Video	.mp4, .avi, .mov	Frame + audio analysis	Video Intelligence
JSON	.json, .jsonl	Structured parsing	—
CSV	.csv, .tsv	Tabular analysis	—

🌐 Web Interface

grepctl now includes a beautiful, customizable web UI for semantic search!

Starting the Web Interface

# Quick start - serves at http://localhost:8000
grepctl serve

# Custom port and theme
grepctl serve --port 3000 --theme-config ./my-company-theme.yaml

# Development mode with auto-reload
grepctl serve --reload

Features

• 🎨 Fully Customizable - Change colors, logo, and branding
• 🌓 Dark Mode - Built-in dark/light theme support
• 🔍 Google-like Interface - Familiar, intuitive search experience
• ⚡ Real-time Search - Search as you type with debouncing
• 📊 Advanced Filters - Filter by modality, date, and more
• 📱 Responsive Design - Works on desktop, tablet, and mobile
• ⌨️ Keyboard Shortcuts - Press / to focus search

Customization

See UI Customization Guide for detailed instructions on:

• Adding your company logo
• Changing color schemes
• Creating custom themes
• White-labeling the interface

Quick customization example:

# my-company-theme.yaml
branding:
  companyName: "Acme Corp"
  logo: "/static/acme-logo.svg"
colors:
  primary: "#FF5722"
  secondary: "#FFC107"

🚀 Advanced Features

Multimodal Search

Search across all data types simultaneously:

# Find mentions across PDFs, images, and videos
grepctl search "quarterly revenue" -m pdf -m images -m video

Automatic Processing

• Vision API extracts text, labels, objects from images
• Document AI performs OCR on scanned PDFs
• Speech-to-Text transcribes audio with punctuation
• Video Intelligence analyzes frames and transcribes speech

Error Recovery

# Automatic fix for common issues
grepctl fix embeddings    # Fixes dimension mismatches
grepctl fix stuck         # Clears stuck processing

🔧 Troubleshooting

Common Issues & Solutions

Issue	Solution
"Permission denied"	Run gcloud auth login and ensure BigQuery Admin role
"Dataset not found"	Run grepctl init dataset
"Embedding dimension mismatch"	Run grepctl fix embeddings
"No search results"	Check grepctl status and run grepctl index update
"API not enabled"	Run grepctl apis enable --all

Quick Diagnostics

# Check everything
grepctl status

# Verify APIs
grepctl apis check

# Check embeddings
grepctl index verify

# Fix any issues
grepctl fix embeddings

🎯 Example Use Cases

1. Code Search: Find code patterns across repositories
2. Document Discovery: Search PDFs for specific topics
3. Media Analysis: Find content in images and videos
4. Log Analysis: Semantic search through log files
5. Data Mining: Query structured data semantically

📈 Performance

• Ingestion: ~50 docs/second for text
• Embedding Generation: ~20 docs/second
• Search Latency: <1 second for most queries
• Storage: ~500MB for 425+ documents
• Accuracy: 768-dimensional embeddings for semantic precision

📦 Package Information

• PyPI: https://pypi.org/project/grepctl/
• Version: 0.1.0
• Requirements: Python 3.11+, Google Cloud Project
• License: MIT

🤝 Contributing

Contributions welcome! Please see CONTRIBUTING.md for guidelines.

Development Setup

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

# Install in development mode with uv
uv sync
uv add --dev pytest black flake8

# Run tests
uv run pytest

# Format code
uv run black .

📄 License

MIT License - see LICENSE for details.

🙏 Acknowledgments

Built with:

• Google BigQuery ML
• Vertex AI (text-embedding-004)
• Google Cloud Vision, Document AI, Speech-to-Text, Video Intelligence APIs
• Python, Click, and Rich CLI libraries

📊 Citation

If you use grepctl in your research or project, please cite:

@software{grepctl2024,
  title = {grepctl: One-Command Orchestration for Multimodal Semantic Search in BigQuery},
  author = {Mulla, Gregory},
  year = {2024},
  url = {https://github.com/yourusername/grepctl},
  version = {0.1.0}
}

---

Ready to transform your data lake into a searchable knowledge base?

pip install grepctl
grepctl init all --bucket your-bucket --auto-ingest

🎉 That's all it takes!