📰 sentinews

AI-powered financial & political news sentiment analysis API

sentinews aggregates news from major RSS feeds (Reuters, BBC, NPR, Al Jazeera, TechCrunch, HN), performs real-time sentiment analysis using a financial/political lexicon engine, and serves the results through a clean REST API.

Features

• 🔍 Sentiment Analysis Engine — Custom lexicon with 100+ financial & political terms, negation handling, and intensifier support
• 📡 RSS Feed Aggregation — Fetches from 8+ major news sources with deduplication
• 🌐 REST API — Hono-powered API with digest, articles, and custom text analysis endpoints
• 💻 CLI Tool — Terminal-based news digest with color-coded sentiment visualization
• ⚡ Built with Bun — Fast startup, fast execution
• 🧪 Tested — Comprehensive test suite for sentiment engine and caching

Quick Start

# Install dependencies
bun install

# Start the API server
bun run dev

# Or use the CLI
bun run cli digest
bun run cli digest finance 10
bun run cli analyze "Markets surge on peace talks"

API Endpoints

Endpoint	Method	Description
/api/v1/health	GET	Health check
/api/v1/digest	GET	News sentiment digest
/api/v1/articles	GET	List analyzed articles
/api/v1/analyze	POST	Analyze custom text
/api/v1/cache/stats	GET	Cache statistics
/api/v1/cache/clear	POST	Clear cache

Query Parameters

/api/v1/digest

• category — Filter by: politics, finance, tech, general
• limit — Max articles (default: 50, max: 200)
• refresh — Force cache refresh: true

/api/v1/articles

• category — Filter by category
• sentiment — Filter by: very_negative, negative, neutral, positive, very_positive
• page — Page number (default: 1)
• limit — Per page (default: 50)

Examples

# Get financial news digest
curl http://localhost:3000/api/v1/digest?category=finance

# Get negative political articles
curl http://localhost:3000/api/v1/articles?category=politics&sentiment=negative

# Analyze custom text
curl -X POST http://localhost:3000/api/v1/analyze \
  -H "Content-Type: application/json" \
  -d '{"text": "Markets crash amid recession fears"}'

Sentiment Scoring

Scores range from -1 (very negative) to +1 (very positive):

Range	Label
≤ -0.5	very_negative
-0.5 to -0.15	negative
-0.15 to 0.15	neutral
0.15 to 0.5	positive
≥ 0.5	very_positive

The engine considers:

• 60+ positive keywords (surge, rally, growth, peace, reform...)
• 60+ negative keywords (crash, recession, war, crisis...)
• Negation (not, never, without...) flips sentiment with dampening
• Intensifiers (very, extremely, slightly...) scale scores
• Title weighting — titles count 60%, content 40%

Development

# Run tests
bun test

# Type check
bun run typecheck

# Build
bun run build

Tech Stack

• Runtime: Bun
• Framework: Hono (ultrafast web framework)
• Language: TypeScript
• Feed Parser: rss-parser
• Validation: Zod v4
• Testing: Bun built-in test runner

License

MIT

---

Built during a late-night dev session 🌙