> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pulsr.news/llms.txt
> Use this file to discover all available pages before exploring further.

# Relevance, Sentiment & Summarization Logic

> How Pulsr determines relevance, sentiment, and creates summaries

Pulsr's AI agent operates on a modular pipeline:

* **Entity Extraction:** Identifies markets/assets involved
* **Relevance Scoring:** Determines whether a tweet should be processed
* **Sentiment Modeling:** Classifies tone (bullish, bearish, neutral)
* **Impact Prediction:** Scores the tweet's likely influence on the market using a proprietary hybrid model based on market volatility data and backtested reactions
* **Summarization:** Generates a concise, readable output for Twitter and the dashboard

All AI components are custom-integrated — this is **not** a ChatGPT wrapper. There's no prompt/response model involved. Pulsr uses autonomous AI agents built for real-time decision-making.

Training is ongoing — we feed the system new labeled news every day to improve accuracy and context awareness.