All Documents
3,528 documents available
Version 1 Mental Model
Documents the evolution of a resume chatbot across seven versions, detailing components, data flow, and design decisions.
Traul — Personal Intelligence Engine
Defines a CLI tool that syncs, indexes, and searches messages from multiple communication platforms with hybrid vector+FTS search.
Qwen3-TTS — Model Reference
Documents the architecture, weights, tokenizer, and inference algorithm for implementing Qwen3-TTS from scratch.
Product Requirements Document — `codex-prompt-refinery
Ingests OpenAI Codex CLI history, deduplicates and groups prompts, then synthesizes atomic and workflow prompts via the Responses API with Structured Outputs.
Dynamic Coattention Networks For Question Answering
Explains the Dynamic Coattention Network model for question answering, covering encoder, co-attention, and iterative pointer network.
Agent Learnings - Papr Memory Python SDK
Documents 12 real-world learnings from converting a 4B-parameter embedding model to Core ML for on-device Apple Silicon inference.
Part 2: Latent Action Prediction Model
Builds a VQ-VAE model to infer discrete latent actions between consecutive Atari game frames for next-frame prediction.
MORE
Reports NER model evaluation results (F1 scores, processing times) for QRNN, Transformer, BiLSTM, and BERT variants on CoNLL-2003.
Implementation Plan: codeindex
We're building a local semantic index for codebases that augments Claude Code's Glob/Grep/Read tools with embedding-based search. The design is informed by OpenViking (AST skeletons, bottom-up directory summaries, score propagation), Augment Context Services (single retrieval tool for agents), sigma-ralph-grindset (`claude --print` as LLM backend), and db-harness (Bun-native dual-database skill pattern).
Criterion MCP Server
Documents a hosted MCP server that exposes three semantic search tools for Quran verses and Hadith collections to AI assistants.
Embedding-Based Memory Recall — Implementation Plan
Replaces LIKE keyword search with semantic vector search for memory recall, using sqlite-vec and OpenAI embeddings with keyword fallback.
Memory Compass - 记忆罗盘 ⭐⭐⭐⭐⭐
Defines a three-tier memory system for AI agents with vector search, file-based WAL logging, and checkpoint recovery.
Building SupportX AI Assist: A Multi-Agent IT Support System
Describes building a multi-agent IT support system with AutoGen, Azure AI Search, and Gemini embeddings for instant issue resolution and automatic escalation.
FastEmbed-rb Roadmap
Tracks unimplemented features from the Python FastEmbed library and documents completed refactoring for the Ruby port.
Loop #1 – RAG Pipeline
Documents a two-loop system: a YouTube transcript RAG pipeline and a Pydantic AI agent with a FastAPI streaming API for querying the ingested content.
Usage
Shows how to use the zai Zig library for chat, streaming, embeddings, and managing providers via code and CLI.
vlite Documentation
Documents the vlite vector database library: installation, CRUD operations, and the custom CTX file format for storing embeddings.
Go-Attention API Documentation
Documents all exported APIs for the go-attention library, covering core types, attention mechanisms, transformer layers, and performance features.
The Annotated Encoder-Decoder with Attention
Implements an encoder-decoder with Bahdanau attention using GRUs, closely following the 2015 neural machine translation paper.
Model Documentation
Documents the transfer learning strategy, architecture, training procedure, and explainability method for a CodeBERT-based human-vs-LLM code classifier.
ChronoMind Development TODO
Maps the current state, immediate fixes, and a phased roadmap for a temporal vector store project in Rust.
uwot (development version)
Changelog for the uwot R package, documenting new features, bug fixes, and parameter changes across versions.
Deployment Guide — Hetzner via Kamal
Walks through provisioning a Hetzner VPS, configuring Kamal 2, and deploying a Rails app with PostgreSQL and S3.
Chloe 部署文档
Guides you through deploying an enhanced SillyTavern fork with direct, Docker, and reverse-proxy setups, plus security hardening and troubleshooting.