Kayce001/learn_pi
记录我对 Pi Agent 的学习
记录我对 Pi Agent 的学习
Ground Pi's answers in the codebase, not assumptions. Read-only research mode with automatic file:line citation verification.
Unofficial source-oriented reconstruction and extension of Grok Bot 0.18.0 for macOS
将参考声音和中文文案自动生成白板动画视频的本地 AI 工具。
Mipmaps for LLM memory. A fixed context budget always covers the entire history — the distant past as one line, this morning verbatim. For agents, assistants, and long-running conversations. One SQLite file, zero dependencies.
Evidence-first verification for AI agents.Don't trust “done”. Verify it.
An interactive line-art room in pure front-end code. Monochrome — the only thing with colour is light.
所有项目的基本原则
A community-maintained Python framework for creating mathematical animations.
FrameMind: reusable visual primitives for AI explainer keyframes.
Agentic report generation service for parking payment analysis.
微信聊天情报看板:聚合群聊信号、话题、链接和趋势
Optimize prompts, code, and more with AI-powered Reflective Text Evolution
Claude Code Snapshot for Research. All original source code is the property of Anthropic.
对 Janlaywss/cloud-code 源码的分析实现
RetrievalLLMSynthesis combines LLMs with RAG to improve text generation relevance and accuracy. It features a Gradio-based chat system for efficient querying and conversation management, using FAISS for fast text retrieval. Simple tools make text generation and data management easy.
PromptFlux is an innovative tool designed for creatives, artists, and developers who want to transform textual ideas into stunning visuals. By combining the power of a large language model for prompt refinement with the Flux text-to-image model, PromptFlux provides a solution for generating high-quality images from refined text descriptions.
这个练手项目花了一天时间写完,并成功完成了训练,训练效果也能看得过去
embed + rerank
源码阅读
My learning notes/codes for ML SYS.
Hands-On Practical MLIR Tutorial
This repository provides example usage of SGLang, a versatile framework for efficient language and vision-language model inference.
AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
A tiny yet powerful LLM inference system tailored for researching purpose. vLLM-equivalent performance with only 2k lines of code (2% of vLLM).
Cross-platform, customizable ML solutions for live and streaming media.
Using modified BiSeNet for face parsing in PyTorch