GitHub - Fyxod/AnalytiQ: Natural-language analytics for Excel and CSV with structured ingestion, text-to-SQL, interactive visualizations, source tracing, and generated reports.

Natural language in. Decision-ready Excel out.

What is AnalytiQ?

AnalytiQ is an AI-powered data assistant that lets you interact with your data using simple language.

You upload files, ask questions, and get:

Clear answers
Insights you can trust
Interactive charts and visualizations
Downloadable Excel reports

No SQL. No formulas. No dashboards.

Upload → Ask → Visualize → Understand → Act

The Problem We Solve

Pain Point	What Users Face
Complex Excel files	Hard to analyze without technical skills
Slow insights	Manually querying data takes hours
Tool overload	Most BI tools require training
Decision paralysis	Too much data, too little clarity

Most users just want answers - not complex tools.

The Solution

AnalytiQ simplifies the entire process:

	Step	Description
1st	Upload	Add your Excel, CSV, or image files
2nd	Ask	Type your question in plain English
3rd	AI processes	Understands intent, runs SQL queries
4th	Get results	Answers, charts, insights, or a full Excel report

Core Capabilities

1. Natural Language Queries

Ask your data anything - no query language required:

"Which region is underperforming?"
"Compare this month with last month"
"What are the top 5 products by revenue?"
"Why are sales declining in Q3?"

The system understands your intent and finds the answer.

2. Data Processing Engine

The system:

Converts questions into SQL queries
Runs them on your data
Handles large datasets
Combines multiple results when needed

The output is explained in simple language.

3. Chart Builder

Build rich, interactive visualizations - visually or just by asking in the chat.

Two ways to create charts:

Chat: Simply describe a chart in your message - "Show me a 3D scatter plot of 10th, 12th grade and CGPA" - and AnalytiQ generates it inline.
Studio Panel: Open the Chart Builder from the Studio to configure and generate charts with full control.

Chart types supported:

Chart Type	Best For
Auto	Let AnalytiQ pick the best fit
Bar / Line / Area	Trends and comparisons
Pie	Part-to-whole breakdowns
Scatter / 3D Scatter	Correlations and distributions
Bubble	Three-variable relationships
Radar	Multi-metric comparisons
Heatmap	Density and matrix patterns
Treemap	Hierarchical data
Composed	Multi-layer custom charts

Quick Actions let you pick a visualization pattern instantly - Monthly trend, Category compare, Part to whole, Metric dashboard, Heatmap matrix, 3D relationship - no description needed.

Chart Gallery: All generated charts are saved in the Chart Builder panel. You can browse, re-open, download (JSON/CSV), or delete any previously created chart at any time.

Chart Analysis: Each chart comes with an AI-written summary below it - explaining the data, key patterns, ranges, and anomalies in plain English.

Export: Every chart can be exported as JSON or CSV for further use.

_{Chart Builder - create new charts}	_{Chart gallery - browse saved charts}

_{Interactive 3D scatter with axis ranges}	_{AI chart analysis below every visualization}

---

4. SQL Transparency

Every answer and chart shows exactly what happened behind the scenes.

Expand SQL details on any message or chart to see:

Query - the exact SQL that was executed
Tables used - which internal tables contributed
Spreadsheet sources - which uploaded files were involved
Execution status - SUCCESS or error details

-- Example: auto-generated query for a correlation chart
SELECT col_10th_percentage_cgpa, col_12th_percentage_cgpa, current_cgpa
FROM arista_1_ece_other_branches
UNION ALL
SELECT col_10th_percentage_cgpa, col_12th_percentage_cgpa, current_cgpa
FROM arista_1_cs_branch_students

You always know what data was used and how the answer was derived - no black boxes.

5. Excel Report Generation

You can generate full Excel reports by describing what you need.

Example:

"Create a report with summary, grouped data, and charts"

The system creates:

Multi-sheet workbooks
Calculations and aggregations
Charts (bar, line, pie)
Proper formatting

The result is a ready-to-use .xlsx file.

6. Background Processing

Excel generation runs in the background:

No waiting screen
Progress can be tracked
File is available when ready

7. Thread Instructions

Customize how AnalytiQ behaves - per thread.

Open Thread Instructions from the settings panel to add persistent instructions that are automatically injected alongside the system prompt for every message in that thread.

Examples:

"Give short answers"
"Always respond in bullet points"
"Focus only on revenue metrics"
"Assume currency is INR"

Instructions can be added, edited, toggled on/off, or deleted at any time. This lets you shape AnalytiQ's behavior for your specific dataset or use case - without repeating yourself every message.

8. Text-to-Speech

Every AI response includes a speaker button that reads the answer aloud.

Click the speaker icon next to any agent message to hear it read out
Useful for reviewing long reports hands-free or accessibility purposes
Works directly in the browser - no setup required

9. Data Handling

Supported inputs:

Excel (.xls, .xlsx)
CSV (.csv)
Images (.jpg, .png, etc.)

Uploaded data is:

Structured automatically
Stored for fast querying
Ready for analysis

10. Spreadsheet Intelligence

Each sheet becomes a queryable table
Multiple sheets are handled automatically
Columns are cleaned and understood
Works with real-world messy data

11. Trust and Transparency

Every answer clearly shows:

Data source - which file was used
Tables used - which sheets contributed
Columns involved - what data was analyzed
SQL queries - exactly how the result was computed

This helps users verify results end-to-end.

12. Flexible Query Modes & Smart Context

Choose how AnalytiQ should answer:

Mode	Behavior
Internal	Uses only your uploaded data
External	Enriches answers with web knowledge
Context	Maintains chat history for follow-ups
Self-Knowledge	Falls back to general AI knowledge

13. Advanced Query Intelligence

The system can:

Break complex questions into smaller parts
Run queries in parallel
Combine results into one answer

If direct answers are not possible, it adapts and still provides useful output.

14. Image Understanding (OCR)

Extracts text from images
Works with scanned documents
Converts visual data into usable information

15. Smart Hybrid Retrieval

Combines keyword search (BM25) + semantic search (vector embeddings)
Finds the most relevant data across all your uploads
Dramatically improves answer accuracy on large datasets

16. Workspace Management

Work is organized into threads
Each thread has its own data and chat
Multiple datasets can be managed easily

17. Insights Tools

Document Summaries - instant overview of any uploaded file
Global Summaries - cross-file insights across your workspace
Mind Map Generation - visual knowledge mapping from your data

18. Export Options

Format	Use Case
Markdown	Clean, shareable chat exports
HTML	Web-ready formatted exports
Excel `.xlsx`	Structured, chart-rich reports
JSON / CSV	Raw chart data export

19. Security

JWT-based user authentication
Controlled, isolated data access per user
Safe, sandboxed file handling

20. Real-Time Experience

Live WebSocket updates via Socket.IO
Fast responses
Smooth interaction

How It Works

Upload file
    ↓
Choose mode 
    ↓
Ask question (or open Chart Builder)
    ↓
System processes data & runs SQL
    ↓
Get answer, chart, or Excel report
    ↓
Inspect SQL · Export data · Listen aloud

_{Clean, minimal dashboard}	_{Name thread & upload files}

_{View all thread documents}	_{Create thread in seconds}

_{Select docs to summarize}	_{Auto-generated document summary}

_{Mind map generating in real-time}	_{Interactive, expandable mind map}

_{Export filtered data as Excel}	_{Studio panel with advanced settings}

System Architecture

flowchart LR
    U[User] --> FE[React Frontend]
    FE --> API[FastAPI + Socket.IO]

    API --> AUTH[JWT Middleware]
    API --> AGENT[Agent Graph Runtime]
    API --> THREADS[Thread/User State]

    THREADS --> MONGO[(MongoDB)]

    AGENT --> RETR[Hybrid Retriever]
    RETR --> CHROMA[(Chroma Vector Store)]
    RETR --> BM25[(BM25 Index)]
    AGENT --> TRIPLES[(Triple Store)]

    AGENT --> SQL[Spreadsheet SQL Engine]
    SQL --> SQLITE[(Per-thread SQLite)]

    AGENT --> EXCEL[Excel Skill Pipeline]
    EXCEL --> XLSX[(Generated .xlsx Exports)]

    AGENT --> CHARTS[Chart Builder Engine]
    CHARTS --> CHARTSTORE[(Chart Artifact Store)]

    API --> STUDIO[Summary + Mind Map Services]
    API --> STORAGE[(Thread File Storage)]
    API --> TTS[Text-to-Speech]
    API --> INSTRUCT[Thread Instructions]

Example Queries

"Show revenue breakdown by region"
"Why are sales decreasing?"
"Compare performance across months"
"Create an Excel report with summary and charts"
"Plot a 3D scatter of 10th, 12th grade and CGPA"
"Show student distribution by branch and gender as a treemap"

Example Output

_{Ask questions, get structured answers}	_{Branch-wise rankings in clean tables}

_{Flags data inconsistencies transparently}	_{Correlations broken into clear insights}

_{Patterns across multiple metrics}	_{Uncovers hidden trends automatically}

Tech Stack

Layer	Technology
Frontend	React + TypeScript
Backend	Python 3.10+ · FastAPI · Socket.IO
AI / LLM	LangChain · LangGraph (Agent Graph Runtime)
Data Processing	Pandas · NLTK
Query Engine	SQLite (per-thread)
Vector Search	ChromaDB
Keyword Search	BM25 Index
Database	MongoDB
Excel Output	openpyxl / xlsxwriter
Charts	Recharts / Plotly (interactive, exportable)
Text-to-Speech	Web Speech API
Auth	JWT Middleware

Setup & Installation

Tested on Python 3.11 · CUDA 12.4 · Windows / Linux

1. Clone the Repository

git clone https://github.com/Fyxod/NatWest-Code-for-Purpose.git
cd analytiq

2. Create a Virtual Environment

# Recommended: Python 3.11
py -3.11 -m venv venv

# Or if Python 3.11 is your default:
python -m venv venv

Activate it:

# Windows
venv\Scripts\activate

# macOS / Linux
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

⚠️ First-time install may take a while - includes heavy packages like torch, sentence-transformers, and easyocr.

4. Configure Environment Variables

cp .env.example .env

Then edit .env and fill in your values:

Variable	Description
`SECRET_KEY`	Any long random string
`GEMINI_API_KEYS`	JSON array of keys - more keys = higher rate limits
`TAVILY_API_KEY`	Your Tavily search API key
`OPENAI_API_KEY`	Only needed if OpenAI is enabled in `core/constants.py`
`USE_VISION_MODEL`	`True` to enable vision/OCR features, `False` to skip

💡 LLM Priority Order: Local LLM → Gemini → OpenAI Default config uses Gemini only. To switch providers, edit core/constants.py. Hit rate limits? Add more keys to GEMINI_API_KEYS - they rotate automatically.

5. Run the Application

Start the backend (first run downloads models - may take a few minutes):

py backend.py

In a separate terminal (venv activated), start the frontend:

py frontend.py

Frontend runs at http://localhost:5173 · Backend API at http://localhost:8000

Why AnalytiQ?

Most data tools are built for analysts. AnalytiQ is built for everyone who needs answers.

Traditional Tools	AnalytiQ
Requires SQL / coding	Plain English
Hours to build reports	Seconds
Static dashboards	Dynamic, queryable chat
Technical setup	Upload and go
Black-box results	Full SQL transparency
No chart customization	10+ chart types, auto or manual
Silent responses	Built-in text-to-speech
Fixed AI behavior	Per-thread custom instructions

AnalytiQ turns questions into insights, insights into reports, and reports into decisions.

Name		Name	Last commit message	Last commit date
.vite/deps		.vite/deps
agent		agent
app		app
core		core
frontend		frontend
scripts		scripts
.env.example		.env.example
.gitignore		.gitignore
.python-version		.python-version
README.md		README.md
backend.py		backend.py
check_imports.py		check_imports.py
frontend.py		frontend.py
glmocr_config.yaml		glmocr_config.yaml
main.py		main.py
reindex.py		reindex.py
requirements.txt		requirements.txt
update_glmocr_config.py		update_glmocr_config.py

Latest commit

History

Folders and files

Repository files navigation