Daily Digest – 2026-08-03

ClickHouse/ClickHouse

Real-time analytics database that doesn't choke when you have millions of rows to query. (C++, 49k★)

★ 49,026

logseq/logseq

Open-source note-taking with bidirectional links and local-first storage. Notion killer if you hate cloud lock-in. (Clojure, 44.2k★)

★ 44,231

codecrafters-io/build-your-own-x

Guided tutorials to rebuild Redis, Docker, Git from scratch. Learn by shipping real implementations. (Markdown, 535.6k★ ⚡992★/day)

★ 535,552

n0-computer/iroh

QUIC + NAT traversal library. Connect to peers directly without messing with IP addresses. (Rust, 12k★ ⚡75★/day)

★ 12,027

paperless-ngx/paperless-ngx

Self-hosted document management that OCRs PDFs and lets you actually find stuff you filed away. (Python, 43.9k★)

★ 43,936

PostHog/posthog

All-in-one analytics, session replay, feature flags, A/B tests. Replaces six SaaS tools with one self-hosted option. (Python, 37.5k★)

★ 37,465

sync/lipsync-2-pro

Professional lipsync synthesis on Replicate. Useful for video avatars or dubbed content without weeks of work. (59K runs)

digital-prairie-labs/cinnabar-agora

cinnabar-agora (206 runs)

starling-lm

7B LLM trained on RLHF feedback. Runs locally via Ollama, reasonable reasoning for its size. (950.8K pulls)

phind-codellama

34B Code Llama variant optimized for programming tasks. Overkill for small scripts, solid for real projects. (949K pulls)

solar

10.7B compact LLM focused on single-turn conversation. Faster inference than Llama 2 at similar quality. (944.9K pulls)

yarn-llama2

Extended Llama 2 with 128k context window. Use when you need long-document analysis locally. (944.8K pulls)

paraphrase-multilingual

Multilingual sentence embeddings for semantic search and clustering. Lightweight, works in 40+ languages. (932.1K pulls)

@appium/logger

Logging library for Appium test automation. Structured logs across Android, iOS, web drivers. (2.1M/wk)

appium-android-driver

Appium driver for Android. Run automated UI tests against real Android apps and Chrome. (980K/wk)

@genkit-ai/core

JavaScript/TypeScript framework for building AI apps. Structured prompts, model routing, built-in tracing. (250K/wk)

@genkit-ai/ai

Genkit's generative AI APIs. Use models from OpenAI, Gemini, Claude through a unified interface. (244K/wk)

genkit

Full Genkit AI framework. Build agents, workflows, and RAG apps without boilerplate. (238K/wk)

openai/whisper

Speech-to-text model that actually works without being cloud-dependent. Gradio demo included. (gradio, 2816 likes)

★ 106,030

coqui/xtts

Text-to-speech that clones voices in multiple languages. Actually sounds natural, no robotic vibes. (gradio, 2766 likes)

tonyassi/voice-clone

Voice cloning space. Feed it audio, get voice model. Useful for podcast production or testing. (gradio, 2665 likes)

multimodalart/qwen-image-multiple-angles-3d-camera

Generate 3D product views from a single image using Qwen. For e-commerce or 3D asset creation. (gradio, 2643 likes)

fffiloni/diffusers-image-outpaint

Image extension tool. Expand canvas and fill new areas intelligently. Useful for product photography. (gradio, 2540 likes)

Qwen/Qwen2.5-0.5B-Instruct

Tiny 500M LLM that runs on a phone. Trade-off: not smart enough for complex tasks, but inference is instant. (text-generation, 6.4M downloads)

mixedbread-ai/mxbai-embed-large-v1

Text embedding model optimized for semantic search. MTEB scores are solid for local RAG setups. (feature-extraction, 5.0M downloads)

google/vit-base-patch16-224

Vision transformer for image classification. Fast baseline that beats ResNets on most benchmarks. (image-classification, 4.7M downloads)

Qwen/Qwen2.5-7B-Instruct-AWQ

7B LLM quantized to 4-bit. Runs on modest GPUs, good balance of speed and reasoning. (text-generation, 4.6M downloads)

dima806/fairface_age_image_detection

Fine-tuned Vision Transformer for age detection from photos. Useful for moderation or personalization flows. (image-classification, 3.3M downloads)

genrobot2025/10Kh-RealOmin-OpenData

Robot manipulation dataset with 1TB+ of video. For training embodied AI models at scale. (800K downloads)

mvp-lab/LLaVA-OneVision-1.5-Mid-Training-85M

Vision-language training dataset. Use this if you're fine-tuning multimodal models. (699K downloads)

XDOF/ABC-130k

Large robotic manipulation dataset. For training models that need to understand object interaction. (549K downloads)

Isopolis – Isometric pixel map of SF

Show HN: Isopolis – Isometric pixel map of SF (305 upvotes) [hackernews_showhn]

AgentSky

Any harness, any LLM — cloud-hosted agents on demand. (305 upvotes) [producthunt]

Ctruh Studio

Create interactive 3D and XR experiences without coding (248 upvotes) [producthunt]

Kakehashi – Experimental userspace to run macOS binaries on Linux ARM

Show HN: Kakehashi – Experimental userspace to run macOS binaries on Linux ARM (240 upvotes) [hackernews_showhn]

Airtop for Google Ads Automation

Build campaigns, optimize spend, and create reports. (191 upvotes) [producthunt]

Qwen3.8-Max

Qwen’s most capable model for coding and cowork (184 upvotes) [producthunt]

Bor – Open-source policy management for Linux desktops

Show HN: Bor – Open-source policy management for Linux desktops (181 upvotes) [hackernews_showhn]

ssh ssh.place

Show HN: ssh ssh.place (164 upvotes) [hackernews_showhn]