SentiBot

A sentiment analysis tool powered by Claude (claude-sonnet-4-6). Classify text as positive, negative, or neutral with a confidence score — via a web UI or the command line.

---

Features

• Single-text analysis — paste any text and get an instant sentiment result
• CSV batch processing — upload a CSV and analyze every row in one go
• Streamlit UI — color-coded result cards with confidence progress bars
• CLI — scriptable via main.py for pipeline use

---

Setup

1. Clone and install dependencies

git clone <repo-url>
cd SentiBot
pip install -r requirements.txt

2. Configure your API key

cp .env.example .env
# Open .env and set your key:
# ANTHROPIC_API_KEY=sk-ant-...

Get a key at console.anthropic.com.

---

Usage

Web UI

streamlit run app.py

Open the URL printed in the terminal (usually http://localhost:8501).

• Analyze Text tab — type or paste text, click Analyze
• Batch CSV tab — upload a CSV (first column = text), click Analyze Batch, download results

CLI

Single text:

python main.py --text "This is absolutely fantastic!"

{
  "text": "This is absolutely fantastic!",
  "label": "positive",
  "confidence": 0.99,
  "tokens_used": 268,
  "model": "claude-sonnet-4-6"
}

Batch CSV:

python main.py --file data/sample.csv --output results.csv

The input CSV must have a header row; text is read from the first column. The output CSV contains all five fields: text, label, confidence, tokens_used, model.

---

Output format

Every result — from the UI, CLI, or API — has the same shape:

Field	Type	Description
text	string	Original input text
label	string	positive, negative, or neutral
confidence	float	0.0 – 1.0
tokens_used	int	Total tokens consumed by the API call
model	string	Model used (claude-sonnet-4-6)

---

Project structure

SentiBot/
├── analyzer.py        # All Anthropic API interaction
├── main.py            # CLI entry point (argparse)
├── app.py             # Streamlit UI
├── data/
│   └── sample.csv     # 10 sample texts for testing
├── tests/
│   ├── test_main.py   # CLI tests (8 cases, all mocked)
│   └── test_app.py    # UI tests (6 cases, AppTest)
├── .env.example       # API key template
└── requirements.txt

---

Running tests

pytest

All tests mock the Anthropic API — no real calls are made and no key is required.

---

Architecture

                    ┌─────────────┐
                    │  analyzer.py │  ← all Claude API calls live here
                    └──────┬──────┘
                           │
              ┌────────────┴────────────┐
              │                         │
       ┌──────▼──────┐         ┌────────▼──────┐
       │   main.py   │         │    app.py     │
       │  (CLI)      │         │  (Streamlit)  │
       └─────────────┘         └───────────────┘

analyzer.py is the only file that talks to the Anthropic API. main.py and app.py are thin layers that handle input/output and delegate everything else.