GitHub 2026-09-25 Daily Trending Projects

An AI engineering course for developers, moving from Python math foundations to LLMs, ReAct, and MCP with an agent tutor.

Hindsight is an agent memory system that organizes long-term memories with biomimetic structures and offers SDKs and an LLM wrapper for quick integration; it targets agents that must learn and adapt over time, but license, community activity, and compliance considerations require careful evaluation.

An Office SDK embedding spreadsheets, documents, and presentations into SaaS or AI apps with one plugin architecture for browsers and Node.js.

AX lets Kubernetes teams declaratively run isolated agent tasks, with Workspace and Gateway managing state and egress.

It compresses Hugging Face, PyTorch, and ONNX models, then connects them directly to TensorRT-LLM, vLLM, and other inference backends.

It fixes X and Bluesky embeds for Discord and Telegram while adding videos, polls, and translations.

Provides industry-tailored Claude agents, skills, and data connectors for financial workflows, deployable as Cowork plugins or via Managed Agents; useful for IB, research, PE and wealth teams for modeling, reporting and ops automation, but requires license and compliance review.

Makes software agent-accessible via a CLI ecosystem with a central Hub, skill registry, and multi-source installers for automation and agent orchestration

A command-line tool for investigators to inspect Android/iOS acquisitions for spyware traces using IOCs.

Adds TDD, planning, and subagent review to coding agents instead of coding immediately.

An AI-agent SDK for Python and TypeScript teams that avoids hand-written loops and hosted control planes.

An offline desktop photo manager for large local libraries, keeping AI and indexes on-device instead of locking files into the cloud.

A tool-routing layer for agent teams: one token reaches 3,000+ endpoints without separate provider signups.

stable-diffusion.cpp implements lightweight, cross-platform Diffusion inference in pure C/C++, compatible with multiple models and backends; it is suitable for teams seeking local high-performance inference and edge/embedded deployment, but be mindful of license and maintenance uncertainties.