Insightly.ai - API

Professional Sentiment Analysis & Strategic Business Insights API

Insightly.ai is a high-performance backend service designed to transform raw customer feedback into actionable business strategies. By combining a fine-tuned IndoBERT model for precision sentiment classification and Gemini 2.5 Flash for strategic intelligence, Insightly.ai provides deep-dive analytics that go beyond simple "Positive/Negative" labels.

Key Features

High-Precision Sentiment Engine: Fine-tuned IndoBERT-base-p2 with 0.79 Macro F1-Score.
Strategic AI Insights: Automated business recommendations and issue categorization powered by Gemini 2.5 Flash.
Efficient Batch Processing: Optimized batch inference handling 1,500+ records in under 45 seconds.
Multi-Format Support: Seamlessly process .csv, .xlsx, and .xls files.
Smart Column Detection: Automatically identifies feedback columns (e.g., 'review', 'content', 'text').
Production Ready: Equipped with rate limiting (SlowAPI), professional logging, and Docker containerization.

Model Performance

The core classification engine was trained on a curated dataset of Indonesian customer reviews, achieving industry-standard benchmarks:

Metric	Score
Accuracy	0.7730
Macro F1-Score	0.7726
Precision (Neutral)	0.8325

Strategic Edge: Unlike standard models, Insightly.ai excels at identifying "Neutral" sentiments and "Authentication Issues" with high precision, which are critical for fintech and service-based applications.

Tech Stack

Framework: FastAPI (Python)
Deep Learning: PyTorch, Hugging Face Transformers
LLM Integration: Google Generative AI (Gemini 2.5 Flash)
Data Handling: Pandas, OpenPyXL
Rate Limiting: SlowAPI
Deployment: Docker, Hugging Face Spaces

Getting Started

Prerequisites

Python 3.12+
Gemini API Key (Google AI Studio)

Installation

Clone the repository:

git clone https://github.com/abidrizmii/insightly.ai.git
cd insightly.ai

Create and activate virtual environment:

python -m venv venv
source venv/bin/activate  # Windows: venv\\Scripts\\activate

Install dependencies:
```
pip install -r requirements.txt
```
Setup environment variables:
Create a .env file and add:
```
GEMINI_API_KEY=your_api_key_here
```

Running the API

uvicorn main:app --reload

API Endpoints

1. Single Prediction

POST /predict
Predict sentiment for a single string.

2. Batch Analysis (JSON)

POST /analyze-batch
Analyze a list of strings and generate strategic insights.

3. File Upload Analysis

POST /analyze-upload (Form-Data)
Upload a CSV or Excel file for mass analysis and full strategic reporting.

Example Output

{
    "status": "success",
    "results": {
        "statistics": { ... },
        "strategic_insights": [
            {
                "topic": "Authentication Issues",
                "urgency": "Critical",
                "evidence": "...",
                "recommendation": "..."
            }
        ]
    }
}

warizmy/insightly.ai-BE