Open-source encrypted messaging platform EnterChat released as self-hostable alternative

·5 views

Developer Shivay00001 has released EnterChat, a complete open-source messaging platform on GitHub. The system includes a Flutter mobile app and Node.js backend with real end-to-end encryption using X25519 key exchange and AES-256-GCM. Unlike tutorial projects that rely on services like Firebase, EnterChat offers a fully self-hostable architecture where the server cannot access message content. The platform implements a simplified X3DH-inspired key exchange protocol for forward secrecy and authentication. All components including PostgreSQL database migrations and Docker configurations are available in the public repository.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Tokenization becomes bottleneck for AI systems as GPU speeds increase

Hugging Face has rebuilt its tokenizer library after identifying tokenization as a new performance bottleneck in AI systems. Tokenization, the process of converting text into numerical tokens, occurs on CPUs before GPUs can begin processing. As GPU speeds have increased significantly, the slower CPU-based tokenization step now causes GPUs to sit idle, delaying responses. This issue is particularly pronounced in high-demand scenarios like serving many simultaneous requests or processing long conversational histories. Studies from Georgia Tech and the PyTorch blog confirm that improving CPU allocation or removing tokenization from Python can dramatically reduce response latency.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Open-source WhatsApp business agent qualifies leads in three languages, runs locally

A developer has released an open-source Python application that functions as an automated business agent on WhatsApp. The software can converse in English, Hindi, and Hinglish to qualify sales leads by extracting key information like name, need, budget, and timeline. It operates with a fully scripted, rule-based system for reliability and offers an optional upgrade layer using free large language models. Designed for small businesses, the agent aims to handle repetitive customer inquiries without requiring expensive platforms or constant human monitoring. The entire system runs locally without external dependencies or API keys in its basic mode.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

ML Zoomcamp Student Highlights Feature Selection Insights from Classification Module

A student has completed Module 3 of the ML Zoomcamp 2026, focusing on machine learning for classification. They noted that correlation analysis and mutual information scoring can yield different rankings for feature importance. In their exercise, a categorical feature was top-ranked by mutual information but invisible to standard correlation matrices. The student also emphasized verifying small performance differences in models before dismissing them as noise. Their next step in the course is Module 4, which covers evaluation metrics for classification.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Open-source n8n automation pack analyzed for core engineering choices

VisionQuantech's open-source n8n automation pack provides 20 workflow JSONs and Python builder scripts. The core workflow uses four nodes to process inquiries through keyword-based intent detection, prioritizing complaints and urgent messages for human escalation. It employs deterministic routing without LLMs, making it fast and debuggable. Advanced workflows add features like Gmail integration, Google Sheets logging, and calendar-based reminder systems. The pack emphasizes programmatically generated workflows for consistency and maintainability over manual creation.

0 comments Read more at DEV Community