QuerySUTRA 0.3.3

SUTRA: Structured-Unstructured-Text-Retrieval-Architecture

QuerySUTRA

SUTRA: Structured-Unstructured-Text-Retrieval-Architecture

Professional Python library for AI-powered data analysis with automatic entity extraction, natural language querying, and intelligent caching.

Installation

pip install QuerySUTRA

# Optional features
pip install QuerySUTRA[embeddings]  # Smart caching
pip install QuerySUTRA[mysql]       # MySQL support
pip install QuerySUTRA[postgres]    # PostgreSQL support
pip install QuerySUTRA[all]         # All features

Key Features

1. Automatic Multi-Table Creation

Upload PDFs, Word documents, or text files and automatically extract structured entities.

from sutra import SUTRA

sutra = SUTRA(api_key="your-openai-key")
sutra.upload("employee_data.pdf")

# Automatically creates:
# - employee_data_people (20 rows, 6 columns)
# - employee_data_contacts (20 rows, 4 columns)
# - employee_data_events (15 rows, 4 columns)

2. Natural Language Querying

result = sutra.ask("Show me all people from New York")
print(result.data)

# With visualization
result = sutra.ask("Show sales by region", viz="pie")

3. Load Existing Databases

# Load SQLite database
sutra = SUTRA.load_from_db("sutra.db", api_key="your-key")

# Connect to MySQL
sutra = SUTRA.connect_mysql("localhost", "root", "password", "database")

# Connect to PostgreSQL
sutra = SUTRA.connect_postgres("localhost", "postgres", "password", "database")

4. Custom Visualizations

result = sutra.ask("Sales by region", viz="pie")       # Pie chart
result = sutra.ask("Trends", viz="line")               # Line chart
result = sutra.ask("Compare", viz="bar")               # Bar chart
result = sutra.ask("Correlation", viz="scatter")       # Scatter plot
result = sutra.ask("Data", viz="table")                # Table view
result = sutra.ask("Analysis", viz="heatmap")          # Heatmap
result = sutra.ask("Auto", viz=True)                   # Auto-detect

5. Smart Fuzzy Matching

sutra = SUTRA(api_key="your-key", fuzzy_match=True)

# "New York City" matches "New York" automatically
result = sutra.ask("Who are from New York City?")

6. Intelligent Caching with Embeddings

sutra = SUTRA(api_key="your-key", use_embeddings=True)

result = sutra.ask("Show sales")           # Calls API
result = sutra.ask("Display sales data")   # Uses cache (no API call)

7. Irrelevant Query Detection

sutra = SUTRA(api_key="your-key", check_relevance=True)

result = sutra.ask("What is the weather?")
# Warns: "This question seems irrelevant to your database"

8. Direct SQL Access (Free)

result = sutra.sql("SELECT * FROM people WHERE city='New York'")
print(result.data)

Complete Configuration

sutra = SUTRA(
    api_key="your-openai-key",
    db="database.db",              # SQLite path
    use_embeddings=True,           # Smart caching (saves API calls)
    check_relevance=True,          # Detect irrelevant queries
    fuzzy_match=True,              # Better NLP
    cache_queries=True             # Simple caching
)

Supported Formats

CSV, Excel, JSON, SQL, PDF, Word, Text, Pandas DataFrame

Usage Examples

Basic Workflow

sutra = SUTRA(api_key="your-key")
sutra.upload("data.pdf")
sutra.tables()                    # View tables
sutra.schema()                    # View schema
sutra.peek("table_name", n=10)    # Preview data
result = sutra.ask("Your question?")

Database Export

sutra.export_db("backup.db", format="sqlite")
sutra.export_db("schema.sql", format="sql")
sutra.save_to_mysql("localhost", "root", "pass", "db")
sutra.save_to_postgres("localhost", "postgres", "pass", "db")
sutra.backup("./backups")

How It Works

Entity Extraction Example

Input PDF:

John Doe lives at 123 Main St, Dallas. Email: [e-mail].
Sarah Smith lives at 456 Oak Ave, Boston. Email: [e-mail].

Output Tables:

people

id	name	address	city	email
1	John Doe	123 Main St	Dallas	[e-mail]
2	Sarah Smith	456 Oak Ave	Boston	[e-mail]

Embeddings for Smart Caching

Uses all-MiniLM-L6-v2 model (80MB, runs locally):

• Query 1: "Show sales" → API call
• Query 2: "Display sales" → 92% similar → Cached (no API call)

Fuzzy Matching

• Query: "New York City"
• Database: ["New York", "Dallas", "Boston"]
• Match: "New York City" → "New York" (85% similar)

API Reference

Class Methods

SUTRA.load_from_db(db_path, api_key, **kwargs) - Load existing SQLite database

SUTRA.connect_mysql(host, user, password, database, ...) - Connect to MySQL

SUTRA.connect_postgres(host, user, password, database, ...) - Connect to PostgreSQL

Instance Methods

upload(data, name=None) - Upload data

ask(question, viz=False, table=None) - Natural language query

sql(query, viz=False) - Raw SQL query

tables() - List all tables

schema(table=None) - Show schema

peek(table=None, n=5) - Preview data

export_db(path, format) - Export database

save_to_mysql(...) - Export to MySQL

save_to_postgres(...) - Export to PostgreSQL

backup(path=None) - Create backup

close() - Close connection

Performance Tips

1. Use load_from_db() to avoid re-uploading
2. Use sql() for complex queries (no API cost)
3. Enable use_embeddings=True for caching
4. Enable cache_queries=True for exact matches

Troubleshooting

No API key error: sutra = SUTRA(api_key="sk-...")

PDF fails: pip install PyPDF2

MySQL error: pip install QuerySUTRA[mysql]

Embeddings error: pip install QuerySUTRA[embeddings]

Requirements

• Python 3.8+
• OpenAI API key
• 100MB disk space (if using embeddings)

License

MIT License

Changelog

v0.3.1

• Semantic embeddings for smart caching
• Fuzzy matching for better NLP
• Irrelevant query detection
• Load existing databases
• MySQL/PostgreSQL connectivity
• Custom visualizations
• All features optional

---

Made by Aditya Batta