Cognify: An AI-Powered Intelligent Content Analysis Platform to Combat Information Overload

Cognify is an open-source AI-driven content intelligence platform developed by harshitnub077. It aims to proactively analyze users' digital content consumption habits using neural network and machine learning technologies, helping them establish healthy information intake patterns. It adopts an Agentic RAG system, with the core being Drift Coach (a drift coach) rather than a simple filtering tool. The project is open-sourced on GitHub, with its homepage at https://cognify-beige.vercel.app. It was released in December 2025 and last updated in June 2026.

In an era dominated by algorithmic recommendations, humans face unprecedented cognitive overload. Short videos, social media, etc., compete for attention in milliseconds, while brain resources are limited. Studies show that modern people spend more than 6 hours per day on digital content, but the proportion of effective value is low. Cognify was born to solve this problem.

Cognify uses a microservices architecture:

1. Browser Extension Layer: The Chrome extension extracts web page DOM and transcribes information (e.g., YouTube/Twitter content) as the starting point for analysis.
2. FastAPI NLP Engine: Integrates local embeddings (sentence-transformers/all-MiniLM-L6-v2), Qdrant vector database, and intelligent LLM routing (primary model: Groq API Llama3, backup: Gemini 2.0 Flash).
3. Drift Coach: Establishes a baseline by retrieving the user's latest 20 pieces of content, generates a drift score (0-100), dominant emotion, a one-sentence reality check, and provides personalized interventions.

Cognify implements a zero-cost architecture: Local embedding models avoid third-party APIs; Qdrant semantic caching (directly returns results if similarity >0.95); uses free tiers of Groq/Gemini. For performance, asyncio.gather asynchronous pipeline avoids lag; data persistence uses PostgreSQL + Prisma ORM, with models including user profiles, content logs, drift scores, etc.

Typical scenarios for Cognify:

1. YouTube Learning Management: Identifies videos with Python in the title but are actually funny clips, prompting a deviation degree of 78%.
2. Twitter Quality Control: Identifies incendiary content and prompts the risk of emotional language.
3. Long-term Habit Tracking: Intuitively views weekly changes in content consumption quality through drift score trends.

Current limitations: Only optimizes English content; only supports Chrome browser and specific websites; GitHub Star count is zero, and the community needs to be expanded. Prospects: The modular architecture facilitates adding new content sources or language support; developers are welcome to contribute.

Cognify represents a reverse AI paradigm—using algorithms to protect rather than hijack attention. For users, it helps establish healthy information habits; for developers, it is an excellent resource for learning end-to-end Agentic RAG and zero-cost LLM applications. It is recommended for users who are troubled by information overload or want to learn AI architecture to try it out. Project link: https://github.com/harshitnub077/Cognify.