Skip to content

Yashaswini-V21/Multi-Agent-Analysts

BranchesTags

Repository files navigation


Python Streamlit Groq Agents License Stars


🚀 Upload any CSV or Excel · Five AI agents fire in sequence

Clean · Analyze · Visualize · Report · Quality Gate — Fully automated.


🎬 Watch the Working Demo

Watch Demo

AnalytIQ Demo

💡 Tip: For the best experience, watch this demo in 720p HD — click the gear icon ⚙️ on YouTube and select Quality → 720p.


Get Started   Roadmap

📌 Table of Contents

🧠 What is AnalytIQ?

AnalytIQ is a fully automated, multi-agent data analysis platform built with Streamlit and powered by Groq's LLaMA 3.3 70B.

You upload a file — the platform does the rest.

A crew of five specialized AI agents fire in sequence, each passing a shared memory block to the next. By the time the Quality Gate agent finishes, you have:

  • ✅ A cleaned & profiled dataset
  • ✅ Statistical insights and patterns
  • ✅ Auto-generated interactive charts
  • ✅ A domain-framed executive report
  • ✅ A quality audit score with verdict
  • ✅ An SQL/DAX query optimizer (Groq-powered)
  • ✅ Full vs Lightning mode benchmarking and leaderboard scores
  • ✅ Async API jobs for long-running analysis
  • ✅ Versioned run artifacts for reproducibility

No data science experience required. No complex setup. Just upload and go.

🏗 System Architecture

╔══════════════════════════════════════════════════════════════════╗
║                    AnalytIQ  v9.0                                ║
║               Streamlit Dark-Mode Web Application                ║
╚══════════════════════╦═══════════════════════════════════════════╝
                       ║
       ╔═══════════════╬══════════════════════╗
       ▼               ▼                      ▼
┌─────────────┐  ┌─────────────────┐  ┌──────────────────────┐
│  Data Layer │  │  Agent Layer    │  │     Query Lab        │
│             │  │  (Groq LLaMA)   │  │    (Groq LLaMA)      │
│ · DataProc  │  │                 │  │                      │
│   load CSV/ │  │  ① Data Eng.   │  │  · SQL Optimizer     │
│   Excel     │  │        ↓        │  │  · DAX Optimizer     │
│   clean     │  │  ② Analyst     │  │  · Auto-Generate     │
│   profile   │  │        ↓        │  │    from schema       │
│   stats     │  │  ③ Visualizer  │  └──────────────────────┘
│             │  │        ↓        │
│ · DataViz   │  │  ④ Reporter    │
│   8 charts  │  │        ↓        │
│   builder   │  │  ⑤ Quality Gate│
└─────────────┘  └────────┬────────┘
                          │  memory-linked pipeline
                          ▼
             ┌────────────────────────┐
             │      Output Layer      │
             │  · Analysis insights   │
             │  · Executive report    │
             │  · QG verdict + score  │
             │  · Agent memory log    │
             │  · Plotly charts       │
             │  · JSON / TXT export   │
             └────────────────────────┘

Data Flow

CSV / Excel  (or built-in demo dataset)
    │
    ├──▶  utils/data_processor.py   ──▶  cleaned DataFrame + text profile
    ├──▶  agents/crew_agents.py     ──▶  5-agent Groq pipeline
    │          Agent ① → ② → ③ → ④ → ⑤ Quality Gate
    ├──▶  utils/visualizer.py       ──▶  Plotly interactive figures
    └──▶  app.py  (Streamlit)       ──▶  rendered UI + download exports

🤖 Agent Pipeline

Five agents run in strict sequence. Each receives the full output of all prior agents via a shared memory block — enabling chained, context-aware reasoning.

# Agent Role
⬡ 1 Data Engineer Validates, cleans nulls, detects types, writes data profile
◎ 2 Data Analyst Statistical analysis — patterns, correlations, outliers, anomalies
△ 3 Visualizer Generates Python/Plotly chart code and recommendations
▫ 4 Report Writer Domain-framed executive report with key findings
🛡 5 Quality Gate Audits every claim · issues PASS / WARN / FAIL verdict + score

Domain contexts: Finance · HR · Marketing · Healthcare · General

Each domain tunes every agent's system prompt for field-specific vocabulary, priority metrics, and framing.

✨ Features

Module Details
📂 Upload CSV / Excel · auto-detect types · instant preview · session state
🔍 Profile Column stats · null analysis · type breakdown · distributions
📊 Analytics Correlation matrix · segmentation · distribution · summary stats
📈 Visualization 8 chart types · Chart Builder · Auto-Generate · Gallery
🤖 AI Engine 5-agent Groq pipeline · domain selector · memory log · QG verdict
⚡ Lightning Mode Faster run mode for low-latency analysis workflows
🧪 Evaluation Lab Full vs Lightning benchmarking with composite score leaderboard
🧵 Async Jobs API Submit analysis jobs, poll status, fetch final result
📡 Metrics API Run completion rate, verdict distribution, and latency summaries
🔬 Query Lab SQL optimizer · DAX optimizer · auto-generate queries (Groq)
💾 Export .txt report · .json full export · audit .md · pipeline download
🎮 Demo Data HR Analytics · E-commerce · Finance Report (built-in, no upload needed)

🎨 Design Highlights

  • Dark-mode only — deep #0d0f14 background with glassmorphism cards
  • Inter + JetBrains Mono typography
  • Purple #7c6aff accent · Cyan #38bdf8 secondary
  • Page-transition flash animations on every navigation
  • Dot-bounce animated processing spinners
  • Ripple effects on nav buttons
  • Glowing animated progress bar during AI pipeline

⚡ Quick Start

1. Clone & Install

git clone https://github.com/Yashaswini-V21/Multi-Agent-Analysts.git
cd Multi-Agent-Analysts
pip install -r requirements.txt

2. Configure API Keys

Create local env from template:

copy .env.example .env

Then edit .env:

# Required — AI Engine + Query Lab
GROQ_API_KEY=gsk_...        # https://console.groq.com  (free)

# Optional — model override
GROQ_MODEL=llama-3.3-70b-versatile

# Optional — API protection
API_AUTH_ENABLED=false
API_AUTH_KEY=replace_with_strong_key

# Optional — key rotation audit
KEY_ROTATED_AT=YYYY-MM-DDTHH:MM:SS+00:00

No API key? Upload → Profile → Analytics → Visualization all work without any key.

3. Run

python -m streamlit run app.py

Open http://localhost:8501 in your browser.

4. Run Backend API

python -m uvicorn api.server:app --host 0.0.0.0 --port 8000 --reload

Open http://localhost:8000/docs for API docs.

5. Docker (UI + API)

docker compose up --build

🔌 API Endpoints

  • GET /health
  • GET /health/keys
  • POST /analyze
  • POST /evaluate
  • POST /jobs/analyze
  • GET /jobs/{job_id}
  • GET /jobs/{job_id}/result
  • GET /metrics/summary

If API auth is enabled, pass:

X-API-Key: <your_api_auth_key>

🧪 Evaluation & Benchmark

Run mode comparison (Full vs Lightning):

python main.py evaluate sample_data.csv --domain General --repeats 2 --modes full lightning

Generate benchmark evidence file:

python scripts/generate_benchmark_report.py --repeats 2 --modes full lightning

📁 Project Structure

Multi-Agent-Analysts/
│
├── app.py                   # Streamlit app (~2800 lines)
│                            #  · CSS design system (Inter + JetBrains Mono)
│                            #  · All 7 page functions
│                            #  · Sidebar + navigation
│                            #  · Query Lab (Groq-powered)
│                            #  · Animation helpers
│
├── agents/
│   ├── crew_agents.py       # 5 GroqAgent classes + DataAnalysisCrew
│   └── prompts.py           # System prompts & DOMAIN_CONTEXTS
│
├── api/
│   └── server.py            # FastAPI endpoints
│
├── services/
│   ├── analysis_service.py  # Analysis orchestration
│   ├── evaluation_service.py# Evaluation and scoring
│   ├── job_service.py       # Async job queue and polling
│   ├── metrics_service.py   # Metrics summary generation
│   └── benchmark_service.py # Benchmark report generation
│
├── utils/
│   ├── data_processor.py    # DataProcessor — load, clean, profile
│   ├── visualizer.py        # DataVisualizer — auto chart generation
│   ├── config.py            # Environment config
│   ├── telemetry.py         # Step timing telemetry
│   ├── artifacts.py         # Artifact persistence
│   ├── reproducibility.py   # Dataset fingerprint/schema snapshot
│   └── security.py          # Key rotation status checks
│
├── tests/                   # Unit tests (API, pipeline, evaluation, security)
├── scripts/
│   └── generate_benchmark_report.py
│
├── outputs/
│   ├── charts/              # Saved chart exports
│   ├── data/                # Processed data exports
│   └── reports/             # Generated report files
│
├── sample_data.csv          # Small sample dataset
├── requirements.txt         # Python dependencies
└── .env                     # API keys  ← gitignored

🗂 Demo Datasets

Three built-in datasets — no file upload needed. Click "Load Demo" on the Upload page.

Dataset Rows Domain Key Columns
👥 HR Analytics 200 HR Dept, Level, Age, Salary, Performance, Attrition
🛒 E-commerce Sales 200 Marketing Category, Region, Revenue, Status, Rating
📈 Finance Report 96 Finance Month, Revenue, COGS, GrossProfit, EBITDA

🧭 Navigation Flow

① Overview  →  ② Upload  →  ③ Profile  →  ④ Analytics
                                                  ↓
          ⑦ Query Lab  ←  ⑥ AI Engine  ←  ⑤ Visualize
  • Every page has ← Back and Forward → buttons at the bottom
  • Sidebar radio always reflects the current page and stays stable during widget interactions
  • Smooth slide-in animation on every page transition

🔑 API Keys

Key Used For Get it Free
GROQ_API_KEY AI Engine (5-agent pipeline) + Query Lab console.groq.com
API_AUTH_KEY Optional protection for backend endpoints Set locally in .env

🔮 What's Coming Next

These features are actively planned and will ship in upcoming releases. ⭐ Star & watch the repo to get notified!

🚀 Near-Term Releases

Feature Description
☁️ Streamlit Cloud Deploy One-click public deployment — try without any local setup
📊 PowerBI Export Push charts and auto-generated DAX directly into Power BI
🗣 Voice-to-Query Speak your data question — get SQL back instantly
📋 Multi-file Join Upload two CSVs and auto-merge on matching columns
📧 Email Delivery Send the full analysis report straight to your inbox

🌟 Big Features on the Horizon

Feature Description
🤝 Real-time Collaboration Share a session URL — multiple analysts, one dataset
🗄 Database Connector Connect PostgreSQL / MySQL / BigQuery directly
📱 Mobile Layout Full responsive design for phone & tablet
🔄 Scheduled Pipelines Run analysis automatically on a cron schedule
🧩 Plugin SDK Build and plug in your own custom agent

🔬 AI Upgrades Coming

Upgrade Details
💬 Agentic Chat Ask follow-up questions about your data in natural language
🔍 Anomaly Detection Agent Dedicated ML-based outlier flagging with explanations
📅 Forecasting Agent Time-series prediction with confidence intervals
🌍 Multi-language Reports Generate analysis reports in 10+ languages
🤖 Model Selection Choose between Groq, GPT-4o, Claude 3.5 in settings

📦 Requirements

streamlit       pandas          numpy
plotly          python-dotenv   groq
openpyxl        scipy

Full pinned versions in requirements.txt.

🤝 Contributing

Contributions are very welcome!

# 1. Fork the repo on GitHub
# 2. Create your feature branch
git checkout -b feat/your-feature

# 3. Commit your changes
git commit -m "feat: add your feature"

# 4. Push and open a Pull Request
git push origin feat/your-feature

Please open an Issue first for large changes so we can discuss the approach.

📄 License

MIT — free to use, modify, and distribute.

👩‍💻 Author

Yashaswini V · @Yashaswini-V21

If AnalytIQ saved you time or impressed you — please give it a ⭐

Five AI agents. One upload. Infinite insights.

About

AnalytIQ - is a fully automated, multi-agent data analysis platform

Resources

Readme
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages