-
GitHub Spotlight: OpenCodex – Run Any LLM in Claude Code, Codex, and Beyond
Break free from vendor lock-in with OpenCodex, a lightweight local proxy that lets you plug DeepSeek, Gemini, Ollama, and Grok right into Claude Code, Codex CLI, and Claude Desktop.
-
GitHub Spotlight: ppt-master – Turn Docs Into Native PowerPoint Decks with AI
Stop wrestling with image-heavy AI slide generators. PPT Master converts PDFs, docs, and raw ideas into fully editable, native PowerPoint files complete with vector shapes, charts, and voice narration.
-
GitHub Spotlight: Langfuse – Open Source LLM Engineering Platform
Langfuse brings open-source LLM observability, prompt management, evaluations, and cost tracking directly to your AI stack without vendor lock-in.
-
GitHub Spotlight: ComfyUI – Master Visual AI Generation with Modular Workflows
Take total control over your diffusion models. ComfyUI gives developers and creators a node-based architecture for rapid visual AI execution, memory optimization, and custom API pipelines.
-
GitHub Spotlight: Graft – Turbocharge Claude Code and AI Coding Agents
Discover Graft, the open-source context layer that slashes AI coding token costs by 42% and boosts accuracy across Claude Code, Cursor, and Codex.
-
GitHub Spotlight: NVIDIA OpenShell – The Safe Runtime for Autonomous AI Agents
Autonomous AI agents can execute arbitrary commands and call external endpoints. NVIDIA OpenShell delivers zero-trust sandboxes to keep agent execution safe, private, and strictly policy-governed.
-
GitHub Spotlight: Obscura – The Headless Browser for AI Agents and Web Scraping
Traditional browser automation frameworks break when paired with modern AI agents. Obscura fixes this with a high-performance, stealth-first headless browser built specifically for LLM-driven web navigation and intelligence.
-
GitHub Spotlight: LibreChat – The All-in-One Open-Source AI Workspace
Break free from single-provider lock-in. LibreChat brings OpenAI, Anthropic, Gemini, DeepSeek, local models, MCP tools, and human-in-the-loop AI agents into one self-hosted web interface.
-
GitHub Spotlight: MiMo-Code – Where Models and Agents Co-Evolve
Xiaomi’s MiMo-Code brings long-horizon reasoning, multi-provider LLM support, and cross-session persistent project memory straight to your command line.
-
GitHub Spotlight: TencentDB Agent Memory – Building Smarter, Low-Cost AI Agents
TencentDB Agent Memory solves the AI context window bottleneck by replacing flat vector dumps with symbolic short-term memory and structured long-term assets, slashing token consumption up to 61% while boosting accuracy.