Finlytic/FinlyticSentiment/Project.md

Finlytic Sentiment Service

Finlytic Sentiment is a C# microservice dedicated to real-time AI sentiment analysis of financial news. It consumes pending news articles, runs FinBERT neural model evaluations, and aggregates sentiment scores at the asset (ISIN) and sector level.

---

Core Features & Architecture

1. FinBERT AI Integration:

   • Evaluates positive, negative, and neutral sentiment probabilities (positiveProbability, negativeProbability, neutralProbability, compoundScore).

2. ISIN & Sector Sentiment Aggregation:

   • Aggregates sentiment scores per financial asset (IsinSentimentSummaryDto) and sector (SectorSentimentSummaryDto).

3. Background Worker Engine:

   • Runs a 5-minute background loop (SentimentBackgroundService) polling pending articles, processing FinBERT evaluations, and publishing results over MQTT.

4. MQTT Event Channels:

   • Publishes updates to finlytic/sentiment/result and responds to RPC requests on services/request/sentiment_GetArticle/# and services/request/sentiment_GetIsin/#.

---

Feature Status

Implemented Features

• FinBERT AI Sentiment Evaluation Service (IFinBertAnalyzerService).
• Sentiment Storage & In-Memory Aggregation (ISentimentStorageService).
• 5-minute Background Worker loop (SentimentBackgroundService).
• Zero-Allocation MQTT serialization via FinlyticJsonSerializerContext.
• Pure Worker Service architecture (Host.CreateApplicationBuilder, no Kestrel HTTP server).

Planned Features

• Historical sentiment trend charting (multi-month sentiment drift per asset).
• Financial entity sentiment impact correlation model.