Python

juergenbarth/templates

Homelab setup files and scripts behind a YouTube channel — A personal collection of Docker templates, scripts and notes the author uses to run his homelab, tied to accompanying tutorials on his YouTube channel. Useful as a starting point for self-hosters, but the author warns the resources may not be kept regularly updated.

Templates for all kinds of projects, e.g. Docker

pvjbuilds/delegate-router

Claude Code and Codex CLI skill: an LLM router that delegates bulk, already-decided coding tasks to another model you already pay for (GPT via Codex, Gemini via agy, a Claude Opus subagent). Task-based model routing, cross-model code review, new-model alerts. Saves tokens and quota; the host verifies every diff.

AmadeusITGroup/Edison

A Python EDIFACT Parser That Builds Real Syntax Trees — Edison models EDIFACT messages as hierarchical syntax trees instead of flat segment lists, converting them into a state machine for parsing, validation, and pretty-printing. It supports conditional and repeatable segment groups, version-based tree selection, escaped separators, and EBCDIC handling—useful for anyone building EDI integrations for travel, logistics, or trade data in Python.

Edison is an EDIFACT parser in Python, will real syntax trees

menloresearch/menlo-sdk

A Python SDK for controlling Menlo's Asimov humanoid robot — This SDK lets Python developers connect to Menlo's Asimov 1 robot over UDP, read its state, and command standing, balancing, and walking with velocity control. It also exposes camera frame and clip utilities (JPEG, NumPy, MP4) via optional Pillow, NumPy, and OpenCV support, aimed at robotics developers building on Asimov hardware.

Atotti/ja-speech-llm

A Full Training Pipeline for a Japanese Speech-Understanding LLM — This repo provides an end-to-end pipeline that connects a frozen Whisper-large-v3 encoder to an LLM-jp 8B decoder via a trainable adapter, enabling Japanese speech chat and ASR. It covers ASR pretraining, multi-task SFT, DPO alignment, and a Gradio demo, with released checkpoints and multiple training modes (adapter-only, LoRA, full).

FT-LLM2026で開発した日本語SpeechLLMの学習コード

pixelartistry/PixelArtistry-MoCapAnything

A Zero-Hassle Local Installer for MoCapAnything V2's Video-to-Animation Magic — This project wraps MoCapAnything V2 into a one-click installer for Windows, Linux and macOS, letting anyone turn a video into animation for 72 creatures and objects without manual setup. It handles Python env creation, PyTorch/CUDA install, model downloads, and even fixes upstream Windows bugs, making local motion capture accessible to non-developers with an NVIDIA GPU.

One-click installer for MoCapAnything V2: turn a video into animation for 72 creatures, animals and objects, locally on your PC (Windows/Linux/macOS).

scienthoon/jev-ood-calibration

Testing Whether Jev's Confidence Scores Stay Honest Out-of-Distribution — An independent, reproducible (~$0.06) calibration audit of TypeSafe's Jev model, comparing its probability estimates on public benchmarks it likely trained on versus 900 novel rule-based tickets it cannot have seen. Finds Jev stays well-calibrated in-distribution but grows confidently wrong on unseen rules, with the miscalibration's sign flipping by question type — useful for anyone evaluating Jev's 'epistemically honest probabilities' claim before routing decisions to it.

Independent calibration test of TypeSafe's Jev on a task it cannot have seen: 900 rule-generated support tickets (choice / score / boolean) plus 3 public benchmarks via Vercel AI Gateway. Raw responses, ECE with noise floor, temperature refit, per-type sign of miscalibration. Reproducible for ~$0.06.

dojicreates/taga-edit

A local Claude Code skill that auto-edits short vertical videos — taga-edit bundles 11 Claude Code skills into one pipeline: drop raw clips in INPUT/, and it joins, trims silence, captions (local Whisper), animates text, grades color, zooms and adds music, SFX and your logo. Built for Reels/TikTok/Shorts, runs fully offline with nothing uploaded.

nempro/development-cleanup

A Codex Skill for Safely Auditing Git Worktrees on Windows — This is a Codex skill for Windows that inspects leftover Git worktrees and temp build artifacts without touching files by default, only removing items when the user explicitly confirms. It needs just Python's standard library plus Git and PowerShell, with no pip packages, models, or API keys required, making it a low-risk cleanup tool for developers juggling many worktrees.

開発フォルダお掃除 — Windows上のCodexでGitワークツリーと一時作業物を安全に点検・整理するスキル。初期動作は読み取り専用。MIT License。

ericchansen/kosync-cwa-bridge

Mirror KOReader reading progress from sync.koreader.rocks into Calibre-Web-Automated

aws-samples/sample-automate-remediation-post-devops-agent-investigation

Turning AWS DevOps Agent reports into approved auto-fixes — This sample CDK project builds a remediation pipeline that reacts to AWS DevOps Agent investigations: EventBridge triggers a durable Lambda orchestrator that queries Bedrock, runs allowlisted tools, and pauses for human approval before any mutating fix. It's aimed at teams wanting a concrete starting point for human-in-the-loop, agentic incident remediation on AWS.

Sample CDK project that demonstrates an automated remediation pipeline for AWS DevOps Agent.

simplynadaf/aws-lean-mcp

Same AWS MCP server, 15 tools vs 4, to show what good tool design saves — A side-by-side AWS cost-triage MCP server built twice: a bloated 15-tool version and a lean 4-tool rebuild following AWS's tool-design guidance. Same account, same question, 7 tool calls drop to 2, making the context-window and token cost of sloppy MCP design concrete rather than theoretical. Useful for anyone designing or auditing MCP servers.

A read-only AWS cost-triage MCP server, shipped bloated (15 tools) and lean (4 tools), to demonstrate MCP tool design and cut token cost. Same account, same question: 7 tool calls down to 2.

AnthusAI/Jev-Calibration

Fixing Jev's Overconfidence: Isotonic Regression Beats Platt Scaling — This repo tests whether Jev's stated confidence matches real accuracy on 8,801 sentiment examples. Raw probabilities had an expected calibration error of 0.117; Platt scaling cut it to 0.052, but isotonic regression dropped it to 0.008 and worked with as few as 20-100 calibration examples. It's a practical follow-up for anyone building LLM-based classifiers who need trustworthy confidence scores, not just correct answers.

Does Jev's confidence mean what it says? Calibrating Jev (TypeSafe System One) with Platt scaling and isotonic regression.

nikisge/claude-motion-kit

Motion Graphics in verschiedenen Stilen mit Claude Code + HyperFrames: Prozess, Stilkarten, Referenzfilme, Prüfwerkzeuge, Sound, 3D/Blender

ashtonckj/Purrrfessorrr-Slothhh

A Pixel-Art Office of Cats Automates Your EDA Boilerplate — This hackathon project turns routine exploratory data analysis into a themed workflow: upload a CSV/TSV/XLSX and five 'specialist cats' handle quality checks, charts, plain-English summaries, and reusable Python/R code. Statistics are computed deterministically; an optional Gemma model only explains results from a summary, never raw data or the math itself.

A playful, pixel-inspired data analysis app that makes exploring, cleaning, and visualising datasets simple, interactive, and fun! 🐈✨

crasyimperior/creative-destruction-reborn

Play Creative Destruction PC offline: local server + client mod (fan preservation project, no game files)

jbulger82/Fullstack_MCP_hub

Fullstack MCP Hub — Complete MCP toolkit with built in RAG, GUI and 50+ free tools (no paid APIs) that runs out-of-the-box. Works with Gemini CLI, OpenAI/Codex CLI, ChatGPT (dev mode), Claude, local LLMs, and anything that can connect to an MCP server—basically “Claude-level” tool power for local models.

agentic-suite/agentsec-agentwatch

Agent security telemetry via OpenTelemetry GenAI traces — agentwatch records AI agent sessions through Claude Code hooks, normalizing events into validated, redacted records with a hash-chained store you can verify and replay. Now at v0.3.0 (a rename-only release), it previously passed 92/94 field-test cases, making it a concrete tool for diagnosing agent loops, tool overuse, or budget burn.

OTel GenAI telemetry and the open agent security-event schema

piyushsachdeva/dataimpulse-job-agent

A job-scraping MCP server routed through residential proxies — This repo pairs a job search scraper with an MCP server for Claude Code, routing every request through DataImpulse residential IPs to avoid blocks. It includes a before/after demo script and proxy test tool, aimed at developers building AI agents that need reliable web scraping access.

AI job search agent + MCP server on DataImpulse residential proxies (demo)

AnmolCanCodes/snaprecall

Search your screenshots by describing what you remember — SnapRecall watches your screenshots folder, has an AI model read each image in the background, and builds a searchable hybrid index. Search in English, Hindi or Hinglish, and choose hosted mode (Gemini API) or fully local mode (Ollama) depending on your privacy needs.

SnapRecall:Find any screenshot by typing what you remember about it..Built for the Hacktoberfest DEV Challenge, Week 3 (theme: "Slow and Steady")

ab2030909/NEXI-focus-companion

A Windows Focus Companion That Coaches Instead of Blocks — NEXI tracks active apps and browser domains on Windows via native Win32 hooks and true idle detection, then nudges you with supportive interventions rather than blocking software. Its Intent Mode lets you set time-budgeted bounded-browsing sessions, paired with Pomodoro sprints and local visual analytics. Built with AdaL, it's for Windows users wanting mindful, local-first focus coaching over hard restrictions.

NEXI is a local-first Windows desktop productivity companion built using AdaL. It bridges system-level activity tracking with non-intrusive focus coaching: rather than forcefully blocking software or closing browser tabs, NEXI brings mindfulness to how you spend your workday.

stefanbosak/codex-cli

OpenAI codex CLI and tooling

motchie/mortgage-forecast

Local-first mortgage forecasting dashboard for variable-rate mortgages

furuya02/sagemaker-embedding-ft-visual-inspection

Fine-tuning loses to better lighting in zero-defect visual inspection — An experiment measuring anomaly detection for visual inspection trained on zero defect images, using a 67-photo rubber duck dataset (CC BY 4.0) bundled in the repo. It finds fine-tuning actually worsens the usable threshold range, while controlling lighting and switching to DINOv2 embeddings gives far bigger gains, even though AUROC looks identical at 1.000 in all cases.

不良画像を1枚も使わない外観検査で、ファインチューニングは割に合うのか実測した記録。アヒル67枚のデータセット(CC BY 4.0)を同梱

takagakiryuheiCM/claude-code-japanese-reply

A Claude Code plugin that hooks replies back into Japanese — This plugin detects when Claude Code's replies drift into English and forces a rewrite in Japanese using hooks on UserPromptSubmit, MessageDisplay, PreToolUse, and Stop events. It judges English by character ratio, ignoring code blocks, URLs, and quotes, and rewrites only once to avoid loops. Useful for users whose tools output English docs that leak into Claude's Japanese responses.

Claude Code の返事が英語になったら、フックで止めて日本語で書き直させるプラグイン

bipul724/availability-ledger

A local-only AI tool that turns chat messages into volunteer availability sheets — Availazy (formerly Availability Ledger) lets volunteer coordinators paste chat messages and a roster; a local Gemma model via Ollama proposes availability statuses while Python code verifies each suggestion against the actual message text and flags unsupported ones. Everything runs on 127.0.0.1, keeping message content off any network by default, and produces a reconciled sheet plus a review log.

simplynadaf/unpile-read-later-triage

A local two-model pipeline that triages your read-later backlog — Unpile pairs TabPFN-v2 for keep/skim/drop prediction with Gemma 3 for a morning digest, running fully offline on your own laptop. It learns from your reopen history to rank saved articles by actual likelihood you'll read them, turning a pile of 1,000+ into a handful with plain-language reasons. Useful for anyone drowning in read-later apps who wants privacy and zero API cost.

A 100% local agent that triages your read-later backlog with TabPFN-v2 and Gemma 3, keeping only the few articles worth your morning. Open-weight, offline, $0. Hacktoberfest 2026 Week 3.

cm-nakamura-shogo/blog-deepseek-v4p1-1017-self-hosting

Self-hosting DeepSeek V4.1 Flash on AWS G7e with SGLang and DSpark — A reproducible verification log for running DeepSeek-V4.1-Flash-1017 on AWS G7e instances via SGLang/ECS/Terraform. It documents DSpark speculative decoding boosting throughput roughly 2x (up to ~55 tok/s), an Engram offload comparison, a failed vLLM attempt, and capacity shortages across regions—useful for anyone self-hosting large MoE models on new AWS GPU instances.

DeepSeek V4.1 Flash 1017 on AWS G7e with SGLang (ECS, Terraform) - blog verification code

Labreo/slothform

An AI agent that fills job and contractor forms for you — SlothForm automates tedious multi-page job applications (Greenhouse, Lever, Ashby) and contractor onboarding forms like W-8BEN. It pairs Browserbase for cloud browser automation with an LLM for tailored answers, and uses Auth0's CIBA flow for a human approval push before any submission goes through.

nyankotaro/sample-automate-remediation-post-devops-agent-investigation

Turning AWS DevOps Agent plans into approved, automated fixes — This CDK sample pipes AWS DevOps Agent mitigation plans through Lambda Durable Functions and Bedrock, validating each step against an allowlist before a single human approval via Slack, CLI, or console. Once approved, steps run in order and a short report is logged automatically. It's a modified version of an aws-samples project, aimed at teams wanting human-in-the-loop remediation.

AWS DevOps Agent の緩和計画を、計画単位の1回の承認（Slack / CLI）で Lambda Durable Functions から実行する。aws-samples の改変版