GitHub 2025-12-26 Daily Trending Projects

RenderCV turns YAML-written CVs into Typst-rendered, versionable, and highly customizable PDF resumes—designed for academics and engineers who prioritize typography and maintainability.

Yuxi-Know is an open-source agent platform centered on LangChain/LangGraph that merges RAG with knowledge graphs, suited for file-driven retrieval, graph visualization and enterprise agent deployments.

X's open recommendation repo exposes end-to-end candidate recall, embeddings, ranking and filtering for real-time feeds; complete but AGPLv3 and internal infra ties increase productionization cost.

Vendure delivers an extensible headless commerce framework centered on TypeScript and GraphQL with a plugin-based model for enterprise customization; however, beware of GPLv3 licensing implications and verify actual repository activity.

LEANN combines graph-based selective recomputation and high-degree-preserving pruning to deliver an ultra-small on-device vector index. It claims ~97% storage savings without retrieval quality loss, enabling private local RAG across personal files and agent memories—suitable for privacy-focused users, researchers, and developers evaluating on-device retrieval.

Chatterbox is a production-oriented open-source multilingual TTS offering 23-language zero-shot voice cloning, emotion exaggeration control, and built-in PerTh watermarking—well suited for content creation, interactive agents and prototyping, but users should consider maintenance activity and training-data provenance for compliance.

Plane is an open-source project management platform for product and engineering teams, combining work items, cycles, modular views and real-time analytics; available as a Cloud service or self-hosted, it suits organizations seeking a controllable, JIRA-like alternative.

vLLM-Omni is a high-throughput, distributed serving framework for multimodal models—combining KV cache, pipelined execution, and OmniConnector for resource disaggregation and dynamic scheduling—suited for low-latency, high-concurrency, and mixed-output production deployments.

LightX2V is a production-oriented lightweight video-generation inference framework that leverages distillation, quantization and multi-hardware optimizations to deliver significant speedups and low-VRAM operation—suited for engineering teams seeking high throughput and rapid deployment.