📊 Data Science Agent API

An API-powered Data Science Agent that performs reliable, grounded data analysis using LLMs.
To avoid hallucinations, the agent never guesses — it writes Python functions, executes them on the dataset, and uses the computed outputs to generate accurate insights and visualizations.

It accepts natural-language questions + optional datasets and returns structured, computation-backed answers.

---

🔗 API Endpoint

POST https://app.example.com/api/

---

🛠️ Example Usage

curl "https://app.example.com/api/" \
  -F "questions.txt=@questions.txt" \
  -F "data.csv=@data.csv" \
  -F "image.png=@image.png"

🚀 Key Features

🧠 Function-Based Reasoning (Low Hallucinations)

• LLM generates Python helper functions instead of making assumptions
• Functions are executed in a sandbox
• Answers are based strictly on computed results
• Ensures correctness, consistency, and reproducibility

---

📥 Flexible Input Formats

• CSV
• JSON
• Images
• Free-form natural-language instructions

---

📤 Structured Output Formats

• JSON
• Text-based insights
• Base64-encoded plots
• Computed metrics, correlations, and summaries

---

⚡ Fast End-to-End Analysis

• Performs data cleaning, EDA, scraping, and lightweight ML
• Typical response time: under 3 minutes

---

🧪 Example Task

Scrape the Wikipedia page of highest-grossing films and answer:

1. How many $2bn movies were released before 2000?
2. Which is the earliest film that grossed over $1.5bn?
3. What’s the correlation between Rank and Peak?
4. Generate a scatterplot of Rank vs Peak with a regression line.

---

✅ Sample Output

[
  1,
  "Titanic",
  0.485782,
  "data:image/png;base64,iVBORw0KG..."
]