🚀 AI Cryptocurrency Prediction Service with Gemini 2.0 Flash

This service provides real-time, LLM-driven cryptocurrency price predictions via a FastAPI backend. It is designed to be consumed by a frontend application to display current predictions and historical analysis for popular cryptocurrencies.

Core Technology Overview

Component	Technology	Role
Backend Framework	FastAPI (Python)	High-performance API server.
AI Model	Google Gemini 2.0 Flash	Generates the price prediction and commentary based on market data.
Data Source	CoinGecko API	Fetches comprehensive, real-time market data (price, volume, change).
Database	SQLAlchemy (ORM)	Persistence layer for storing prediction history and metadata (defaulting to SQLite).

---

🧠 Prediction Process Explained

The service acts as a specialized AI financial analyst. For every prediction, it follows these steps:

1. Data Acquisition: Fetches current market metrics (Price, 24h Volume, 24h Change, Market Cap) for the target crypto asset (e.g., btc, eth) from CoinGecko.
2. Prompt Construction: The fetched data is inserted into an advanced, structured prompt optimized for Gemini 2.0 Flash's analytical capabilities.
3. LLM Analysis: The Gemini model processes the data and the prompt's analysis framework to generate a prediction (probability up/down) and a concise expert commentary.
4. Normalization & Storage: The model's probabilities are validated and normalized to sum to $100%$. The complete result is then saved to the database, ensuring data integrity and historical tracking.

The prediction is a 24-hour directional forecast.

---

⚙️ API Endpoints for Frontend Developers

All endpoints return a standard JSON format.

1. Generate a New Prediction (On-Demand)

This endpoint forces the full prediction workflow, including data fetch, LLM call, and database write.

Method	Endpoint	Description
POST	/predict	Triggers a new prediction for the specified symbol immediately.

Request Body (POST /predict)

Field	Type	Description
symbol	string	The cryptocurrency symbol (e.g., btc, eth, sol).

Request Body:

{
  "symbol": "btc"
}

Result

{
  "symbol": "btc",
  "timestamp": "2025-09-27T14:25:38.348272+00:00",
  "probability_up": 58,
  "probability_down": 42,
  "commentary": "Bitcoin shows a slight positive 24h change with substantial volume, indicating continued interest. The high market cap suggests stability, but the upside is tempered by potential profit-taking at this price level. Expect moderate volatility with a slight upward bias.",
  "source_data": {
    "symbol": "BTC",
    "coin_id": "bitcoin",
    "price_usd": 109363,
    "24h_change_pct": 0.21008529749368454,
    "volume_24h": 33864925655.1949,
    "market_cap": 2179208938875.3438,
    "last_updated": "2025-09-27T14:25:12+00:00",
    "analysis_timestamp": "2025-09-27T14:25:32.779565+00:00"
  },
  "model_meta": {
    "provider": "gemini",
    "model": "gemini-2.0-flash-exp",
    "prompt_version": 2,
    "timestamp": "2025-09-27T14:25:38.348272+00:00",
    "confidence": "medium"
  }
}

2. Retrieve Latest Prediction

Method	Endpoint	Description
GET	/latest/{symbol}	Retrieves the most recent prediction for the given symbol from the database.

{
  "symbol": "btc",
  "timestamp": "2025-09-27T14:25:38.348272+00:00",
  "probability_up": 58,
  "probability_down": 42,
  "commentary": "Bitcoin shows a slight positive 24h change with substantial volume, indicating continued interest. The high market cap suggests stability, but the upside is tempered by potential profit-taking at this price level. Expect moderate volatility with a slight upward bias.",
  "source_data": {
    "symbol": "BTC",
    "coin_id": "bitcoin",
    "price_usd": 109363,
    "24h_change_pct": 0.21008529749368454,
    "volume_24h": 33864925655.1949,
    "market_cap": 2179208938875.3438,
    "last_updated": "2025-09-27T14:25:12+00:00",
    "analysis_timestamp": "2025-09-27T14:25:32.779565+00:00"
  },
  "model_meta": {
    "provider": "gemini",
    "model": "gemini-2.0-flash-exp",
    "prompt_version": 2,
    "timestamp": "2025-09-27T14:25:38.348272+00:00",
    "confidence": "medium"
  }
}

3. Retrieve Prediction History