A reading list for open source
GitHub trends, explained through repositories
Track GitHub trends through trending repositories, repo profiles, daily open-source news, and weekly topic reports.
Repository profiles
-
tw93/Pake: turn any website into a tiny Tauri desktop app with one command
Pake wraps any webpage into a lightweight Rust/Tauri desktop app, around 5MB and far smaller than Electron. Install, the GPL output exception, and where the webview approach stops.
-
Kilo-Org/kilocode: an open-source AI coding agent for VS Code, JetBrains, and the CLI
Kilo Code is an open-source AI coding agent across VS Code, JetBrains, and a CLI, with 500+ models and no-markup pricing. How its open-core model and autonomous mode compare to Cline and Continue.
-
DeusData/codebase-memory-mcp: a C MCP server that turns your repo into a queryable graph
codebase-memory-mcp is a single-binary MCP server that indexes any codebase into a knowledge graph for AI coding agents: 158 languages, sub-ms queries, no API key.
-
EpicGames/lore: Epic's open-source, content-addressed version control for big binary projects
Lore is Epic Games' open-source version control system in Rust: centralized, content-addressed, built for code plus terabytes of binary game assets. How it differs from Git and Perforce.
-
Omnigent is an open-source Python meta-harness: run and supervise Claude Code, Codex, Cursor, Pi, and custom agents in one session, with governance policies, sandboxing, and cross-device collab.
-
google-research/timesfm: zero-shot time-series forecasting from a pretrained model
Google Research's TimesFM does zero-shot time-series forecasting from a pretrained model. What version 2.5 changed, the batch-composition gotcha, install, and how it compares to Chronos and TimeGPT.
-
Hello Algo is a free, beginner-friendly data structures and algorithms book with animated illustrations and runnable code in Python, Java, C++, Go and more.
-
OpenHands/OpenHands: the open-source autonomous software-engineering agent (formerly OpenDevin)
OpenHands is an open-source autonomous coding agent that runs in a sandbox, as an SDK, CLI, local GUI, or cloud. Formerly OpenDevin; how it compares to cline, plus Docker and cost gotchas.
-
tesseract-ocr/tesseract: the classic open-source OCR engine, and how to get accuracy out of it
Tesseract is the long-running open-source OCR engine with LSTM recognition and 100+ languages. Why accuracy hinges on preprocessing, and how it compares to EasyOCR and PaddleOCR.
-
cline/cline: the open-source coding agent for VS Code, JetBrains, and the terminal
Cline is an open-source AI coding agent for VS Code, JetBrains, and the CLI: Plan/Act modes, edit checkpoints, bring-your-own model, and how its token cost compares to Aider.
-
docling-project/docling: parse PDFs, DOCX, and PPTX into clean Markdown for RAG
docling converts PDF, DOCX, PPTX, HTML, and images into structured Markdown or JSON with layout and tables preserved, for LLM and RAG pipelines. How it compares to unstructured and markitdown, and its real limits.
-
virattt/ai-hedge-fund: an educational multi-agent stock-analysis simulator (not for real trading)
ai-hedge-fund runs a team of LLM agents modeled on famous investors to debate buy and sell signals. Why it is educational only with no real trading, its API costs, and how it compares to TradingAgents.
-
mem0ai/mem0: a long-term memory layer for AI agents
Mem0 gives AI agents long-term memory: it extracts, stores, and retrieves facts across sessions, with an open-source library and a managed platform. How it compares to Letta, the vector-store setup, and its limits.
-
remotion-dev/remotion: make videos programmatically in React (mind the license)
Remotion lets you build videos with React components and code, ideal for data-driven and personalized video at scale. Its company-license requirement, render costs, and how it compares to Revideo.
-
Aider-AI/aider: AI pair programming in your terminal, with git baked in
Aider is an open-source AI pair programmer that runs in your terminal, commits every change to git, and maps large repos for context. How it compares to cline, and where token cost bites.
-
microsoft/qlib: a full quant research platform, and its steep learning curve
qlib is Microsoft's AI quant investment platform covering the pipeline from data to backtest, with many ML models and RD-Agent. Why the learning curve is steep, the data-setup catch, and how it compares.
-
2noise/ChatTTS: conversational Chinese-English TTS, with research-only licensing
ChatTTS is a text-to-speech model tuned for dialogue, with natural prosody, laughter, and pauses in Chinese and English. Its research-only license, anti-misuse design, and how it compares to F5-TTS.
-
continuedev/continue: the open-source coding assistant, now at a final read-only release
Continue is an open-source, configurable AI coding assistant for VS Code, JetBrains, and the CLI. As of 2026-06 the repo is read-only at a final 2.0.0; what that means and where to go.
-
fishaudio/fish-speech: multilingual TTS and voice cloning under a non-commercial research license
fish-speech (now Fish Audio S2) is an open multilingual text-to-speech and voice-cloning model. Why its custom research license blocks commercial use, and how it compares to F5-TTS, ChatTTS, and VoxCPM.
-
JaidedAI/EasyOCR: one-line multilingual OCR in Python, and its maintenance question
EasyOCR is a deep-learning OCR library that reads 80+ languages with a single readtext() call. What it is great at, its slowing maintenance, and how it compares to Tesseract and PaddleOCR.
-
eyaltoledano/claude-task-master: turn a PRD into tracked tasks for Cursor, Windsurf, and Claude Code
Task Master parses a PRD into tracked, dependency-aware tasks and runs as an MCP server in Cursor, Windsurf, Claude Code, and VS Code. Its setup, API-key needs, license, and real gotchas.
-
vercel/ai: the TypeScript AI SDK for building LLM apps across providers
The Vercel AI SDK is a unified TypeScript toolkit for LLM apps: generateText, streamText, and useChat across many providers, with streaming and tool calling. Breaking changes between versions, and how it compares to LangChain.
-
letta-ai/letta: stateful agents with memory, formerly MemGPT
Letta (formerly MemGPT) is a framework for stateful agents that keep long-term memory across sessions, with core and archival memory, a CLI, an API, and a visual editor. How it compares to Mem0 and its self-hosting trade-offs.
-
AI4Finance-Foundation/FinGPT: open financial LLMs and fine-tuning recipes, not a trading system
FinGPT is a collection of open financial language models and LoRA fine-tuning recipes for sentiment and financial NLP. Why it is a research resource not a trading system, its reproducibility friction, and how it compares.
-
Unstructured-IO/unstructured: preprocess any document into elements for RAG
unstructured turns PDFs, DOCX, PPTX, HTML, and email into structured elements for LLM and RAG pipelines. The open-source library versus the paid platform, its heavy dependencies, and how it compares to docling.
-
SWivid/F5-TTS: flow-matching voice cloning with MIT code but non-commercial weights
F5-TTS is a flow-matching, zero-shot voice-cloning TTS with MIT-licensed code. Why the pretrained weights are CC-BY-NC, how cloning depends on the reference clip, and how it compares to fish-speech.
-
FujiwaraChoki/MoneyPrinter: a local, Ollama-first short-video generator
MoneyPrinter generates short-form videos from a topic using local Ollama models and a DB-backed queue. How it differs from the larger MoneyPrinterTurbo, its API and platform-policy risks, and when to use it.
-
jina-ai/reader: turn any URL into LLM-friendly Markdown with one prefix
Jina Reader (r.jina.ai) converts any web page into clean Markdown by prefixing the URL, with a search endpoint and self-hosting. How it compares to Firecrawl, its rate limits, and when to use it.
-
voxel51/fiftyone: curate and debug computer-vision datasets
FiftyOne is an open-source tool for visualizing, curating, and debugging computer-vision datasets: find label errors, dedupe by embeddings, evaluate models. Its MongoDB dependency, the open-source vs Teams split, and how it compares to supervision.
-
assistant-ui/assistant-ui: composable React primitives for AI chat interfaces
assistant-ui provides composable React primitives for AI chat UIs: streaming, tool-call rendering, and runtime adapters for the Vercel AI SDK and LangGraph. How it complements the AI SDK and when to use it.
Trends: This week's fastest-growing · GitHub news · GitHub repo profiles · Popular topics