QuerySUTRA 0.6.0

AI-powered data analysis for structured and unstructured data. Query PDF, Word, CSV, Excel with natural language.

QuerySUTRA

SUTRA: Structured-Unstructured-Text-Retrieval-Architecture

AI-powered data analysis. Upload any data (PDF, Word, Text, CSV, Excel), query with natural language, export to MySQL.

Installation

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

Quick Start

from sutra import SUTRA

sutra = SUTRA(api_key="your-openai-key")
sutra.upload("data.pdf")  # or .docx, .txt, .csv, .xlsx, .json
result = sutra.ask("Show me all people")
print(result.data)

Supported Formats

Structured Data:

• CSV (.csv)
• Excel (.xlsx, .xls)
• JSON (.json)
• SQL (.sql)
• Pandas DataFrame

Unstructured Documents (AI Extraction):

• PDF (.pdf)
• Word (.docx)
• Text (.txt)

Core Features

1. Upload Any Data Format

# Structured data
sutra.upload("sales.csv")
sutra.upload("report.xlsx")
sutra.upload("api_data.json")
sutra.upload("dump.sql")

# Unstructured documents (AI extracts entities)
sutra.upload("resume.pdf")
sutra.upload("meeting_notes.docx")
sutra.upload("transcript.txt")

# DataFrame
import pandas as pd
df = pd.DataFrame({'name': ['Alice'], 'score': [95]})
sutra.upload(df, name="scores")

2. Complete Data Extraction

Processes entire documents in chunks. No data loss.

# PDF - Extracts ALL pages
sutra.upload("50_page_report.pdf")  # Gets all 50 pages, all employees

# Word - Extracts ALL content
sutra.upload("large_document.docx")  # Full document processed

# Text - Processes ALL lines
sutra.upload("log_file.txt")  # Entire file analyzed

# All create multiple related tables
sutra.tables()

3. Automatic MySQL Export

One-line upload and export. Database auto-created.

sutra.upload("data.pdf", auto_export_mysql={
    'host': 'localhost',
    'user': 'root',
    'password': 'your_password',
    'database': 'my_database'  # Auto-creates if not exists
})

4. Natural Language Queries

result = sutra.ask("Show all people from California")
result = sutra.ask("Who has Python skills?", table="skills")
result = sutra.ask("Count employees by state", viz="pie")

5. Custom Visualizations

result = sutra.ask("Sales by region", viz="pie")       # Pie chart
result = sutra.ask("Trends over time", viz="line")     # Line chart
result = sutra.ask("Compare values", viz="bar")        # Bar chart
result = sutra.ask("Correlations", viz="scatter")      # Scatter
result = sutra.ask("Show table", viz="table")          # Table
result = sutra.ask("Heatmap", viz="heatmap")           # Heatmap
result = sutra.ask("Auto", viz=True)                   # Auto-detect

6. Load Existing Databases

# Load SQLite
sutra = SUTRA.load_from_db("data.db", api_key="key")

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

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

7. Fuzzy Matching

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

# "New York City" matches "New York" automatically
result = sutra.ask("Who are from New York City?")
# Fuzzy: 'City' -> 'New York'

Uses difflib.get_close_matches with 60% threshold.

8. Embeddings for Smart Caching

Save 90% on API costs.

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

result = sutra.ask("Show sales")            # API call
result = sutra.ask("Display sales data")    # Cached (92% similar)
result = sutra.ask("Give me sales info")    # Cached (88% similar)

How it works:

• Model: all-MiniLM-L6-v2 (80MB, runs locally)
• Converts queries to 384D vectors
• 85% similarity threshold
• No external API calls

Cost savings:

• 10 similar queries: 1 API call vs 10 = 90% savings

9. Irrelevant Query Detection

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

result = sutra.ask("What's the weather?")
# Warning: Query may be irrelevant

10. Direct SQL

result = sutra.sql("SELECT * FROM people WHERE state='CA'")

Complete Example

from sutra import SUTRA

# Initialize with all features
sutra = SUTRA(
    api_key="your-key",
    use_embeddings=True,
    fuzzy_match=True,
    check_relevance=True
)

# Upload any format
sutra.upload("employees.pdf")      # PDF
sutra.upload("skills.docx")        # Word
sutra.upload("projects.txt")       # Text
sutra.upload("sales.csv")          # CSV
sutra.upload("budget.xlsx")        # Excel

# View tables
sutra.tables()

# Query
result = sutra.ask("Show all people", viz="bar")

# Export to MySQL
sutra.save_to_mysql("localhost", "root", "pass", "my_db")

Import to MySQL Workflow

Colab:

sutra.upload("data.pdf")
sutra.export_db("data.db", "sqlite")
from google.colab import files
files.download("data.db")

Windows:

sutra = SUTRA.load_from_db("data.db", api_key="key")
sutra.save_to_mysql("localhost", "root", "pass", "my_db")

Export Options

sutra.export_db("backup.db", "sqlite")
sutra.export_db("schema.sql", "sql")
sutra.export_db("data.json", "json")
sutra.export_db("data.xlsx", "excel")
sutra.save_to_mysql("localhost", "root", "pass", "db")
sutra.save_to_postgres("localhost", "postgres", "pass", "db")

API Reference

Methods

• upload(data, name, auto_export_mysql) - Upload any format
• ask(question, viz, table) - Natural language query
• sql(query, viz) - Direct SQL
• tables() - List tables
• schema() - Show schema
• peek(table, n) - Preview
• save_to_mysql(...) - Export MySQL (auto-creates DB)
• export_db(path, format) - Export database
• load_from_db(path) - Load SQLite
• connect_mysql(...) - Connect MySQL

Requirements

Python 3.8+, OpenAI API key

License

MIT

---

Made by Aditya Batta