🛠️ All DevTools

[Other] Track your OpenAI Codex sessions live from your Mac menu bar SessionWatcher for Codex keeps your OpenAI Codex sessions visible at a glance. See live token counts, time remaining, and usage stats directly from your macOS menu bar. So you never get cut off mid-flow again.

[Other] easily manage apple refund requests, track status This service is built on Laravel / Filament multi-tenant architecture, effectively helping developers prevent fraudulent refunds by instantly processing Apple's CONSUMPTION_REQUEST notifications and asynchronously returning consumption data.

[Other] Vultisig launches with a new facelift. Revolutionary seedless crypto wallet powered by TSS (Threshold Signature Scheme) technology. Vultisig eliminates seed phrase vulnerabilities while delivering enterprise-grade multi-sig security for than 30 chains like Bitcoin, Ethereum, Solana, and more.

[Other] Create Pixel Art and Sprite Sheets Instantly for Your Games Generate stunning sprites and pixel art instantly with this powerful tool for pixel art creation and sprite sheets. Perfect for game developers and artists seeking ready to use assets, it delivers high quality visuals tailored for games and creative projects

[Other] Show HN: Agentic Design Patterns – Python Edition, from the Codex Codebase While reading Agentic Design Patterns by Antonio Gulli, I wanted to see how these patterns look in real code. I cloned the OpenAI Codex repo (the open-source AI coding assistant that recently trended on HN) — but it was in Rust.<p>So, I used an Cursor to help me extract and translate 18+ agentic patterns from Codex’s codebase into Python. That small experiment turned into a full open-source guide: GitHub: Codex Agentic Patterns <a href="https:&#x2F;&#x2F;github.com&#x2F;artvandelay&#x2F;codex-agentic-patterns" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;artvandelay&#x2F;codex-agentic-patterns</a><p>Each pattern comes with:<p>A short explanation and code sample<p>A runnable exercise and agent snippet<p>A summary of how Codex used the pattern (e.g., prompt chaining, tool orchestration, reflection loops, sandbox escalation)<p>One full working Python agent that ties it all together<p>If you’ve read the agentic design patterns book or explored Codex, this is a bridge between theory and practice — focused on runnable, open examples instead of abstract diagrams.<p>It’s completely free and open-source. Would love feedback, ideas, or even new patterns from your own agent experiments.

[Other] Building a JavaScript Runtime from Scratch using C

[Database] ThalamusDB: Query text, tables, images, and audio

[Other] Show HN: Rethinking audit trails in Django (structured and database-free) We were using django-easy-audit to track CRUD and API events in Django. The problem: every action meant an extra database write, which slowed transactions, increased infra costs, and left us with audit tables that weren’t great for analytics.<p>I built django-activity-audit (PyPI) to fix this: - Extends Django’s logging system with custom AUDIT and API levels. - Captures CRUD + API request&#x2F;response events as structured JSON logs. - Vector tails the logs and ships them into ClickHouse. - Grafana makes them queryable and visual.<p>This removes the extra DB writes, gives structured data ready for analysis, and keeps costs down.<p>Curious — how are others handling audit logging in Django (or other frameworks)? Do you log it, write it to a DB, or something else entirely?

[Database] Show HN: Arc – high-throughput time-series warehouse with DuckDB analytics Hi HN, I’m Ignacio, founder at Basekick Labs.<p>Over the past months I’ve been building Arc, a time-series data platform designed to combine very fast ingestion with strong analytical queries.<p>What Arc does? Ingest via a binary MessagePack API (fast path), Compatible with Line Protocol for existing tools (Like InfluxDB, I&#x27;m ex Influxer), Store data as Parquet with hourly partitions, Query via DuckDB engine using SQL<p>Why I built it:<p>Many systems force you to trade retention, throughput, or complexity. I wanted something where ingestion performance doesn’t kill your analytics.<p>Performance &amp; benchmarks that I have so far.<p>Write throughput: ~1.88M records&#x2F;sec (MessagePack, untuned) in my M3 Pro Max (14 cores, 36gb RAM) ClickBench on AWS c6a.4xlarge: 35.18 s cold, ~0.81 s hot (43&#x2F;43 queries succeeded) In those runs, caching was disabled to match benchmark rules; enabling cache in production gives ~20% faster repeated queries<p>I’ve open-sourced the Arc repo so you can dive into implementation, benchmarks, and code. Would love your thoughts, critiques, and use-case ideas.<p>Thanks!

[Other] Show HN: We trained an MoE LLM built for developer tasks

[API/SDK] Show HN: Kalendis – Scheduling API (keep your UI, we handle timezones/DST) Kalendis is an API-first scheduling backend. You keep your UI; we handle the gnarly parts (recurrence, time zones, DST, conflict-safe bookings).<p>What it does:<p>• MCP tool: generates typed clients and API route handlers (Next.js&#x2F;Express&#x2F;Fastify&#x2F;Nest) so you can scaffold calls straight from your IDE&#x2F;agent tooling.<p>• Availability engine: recurring rules + one-off exceptions&#x2F;blackouts, returned in a clean, queryable shape.<p>• Bookings: conflict-safe endpoints for creating&#x2F;updating&#x2F;canceling slots.<p>Why we built it:<p>We kept rebuilding the same &quot;hard parts&quot; of scheduling: time zones&#x2F;DST edge cases, recurring availability, conflict-aware booking, etc. We wanted a boring, reliable backend so we could ship product features without adopting a hosted scheduling UI.<p>How it&#x27;s helped:<p>We stopped re-implementing DST&#x2F;recurrence math and shipped booking flows faster.<p>One small team (just 2 developers) built a robust booking platform for their business using Kalendis—they kept full control of their UX without spending lots of cycles on scheduling infrastructure.<p>The MCP generator cut the glue code: drop in a typed client or route, call the API, move on.<p>Some tech details:<p>• REST API with ISO-8601 timestamps and IANA time zones<p>• Recurring availability + one-off exceptions (designed to compose cleanly)<p>• Focused scope: users, availability, exceptions, bookings (not a monolithic suite)<p>MCP integration:<p>MCP server exposes tools like generate-frontend-client, generate-backend-client, generate-api-routes, and list-endpoints.<p>Add to your MCP settings:<p>{ &quot;mcpServers&quot;: { &quot;kalendis&quot;: { &quot;command&quot;: &quot;npx&quot;, &quot;args&quot;: [&quot;-y&quot;, &quot;@kalendis&#x2F;mcp&quot;] } } }<p>How to try it:<p>Create a free account → get an API key. (<a href="https:&#x2F;&#x2F;kalendis.dev" rel="nofollow">https:&#x2F;&#x2F;kalendis.dev</a>)<p>Hit an endpoint:<p>curl -H &quot;x-api-key: $KALENDIS_API_KEY&quot; \ &quot;https:&#x2F;&#x2F;api.kalendis.dev&#x2F;v1&#x2F;availability&#x2F;getAvailability?userId=&lt;user-id&gt;&amp;start=2025-10-07T00:00:00Z&amp;end=2025-10-14T00:00:00Z&amp;includeExceptions=true&quot;<p>What feedback would be most useful:<p>Gaps in the endpoint surface (what&#x27;s missing for your use case?).<p>Features that would benefit this service integration in your app.<p>MCP generator output—anything you&#x27;d want it to emit differently?<p>Happy to answer questions and post example snippets in the thread. Thanks for taking a look!<p>— Dave (Kalendis)

[Other] Launch HN: LlamaFarm (YC W22) – Open-source framework for distributed AI Hi HN! We&#x27;re Rob, Matt, and Rachel from LlamaFarm (<a href="https:&#x2F;&#x2F;llamafarm.dev">https:&#x2F;&#x2F;llamafarm.dev</a>). We&#x27;re building an open-source AI framework based on a simple belief: the future isn&#x27;t one massive model in the cloud—it&#x27;s specialized models running everywhere, continuously fine-tuned from real usage.<p>The problem: We were building AI tools and kept falling into the same trap. AI demos die before production. We built a bunch of AI demos but they were impossible to get to production. It would work perfectly on our laptop, but when we deployed it, something broke, and RAG would degrade. If we were running our own model, it would quickly become out of date. The proof-of-concept that impressed the team couldn&#x27;t handle real-world data.<p>Our solution: declarative AI-as-code. One YAML defines models, policies, data, evals, and deploy. Instead of one brittle giant, we orchestrate a Mixture of Experts—many small, specialized models you continuously fine-tune from real usage. With RAG for source-grounded answers, systems get cheaper, faster, and auditable.<p>There’s a short demo here: <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=W7MHGyN0MdQ" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=W7MHGyN0MdQ</a> and a more in-depth one at <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=HNnZ4iaOSJ4" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=HNnZ4iaOSJ4</a>.<p>Ultimately, we want to deliver a single, signed bundle—models + retrieval + database + API + tests—that runs anywhere: cloud, edge, or air-gapped. No glue scripts. No surprise egress bills. Your data stays in your runtime.<p>We believe that the AI industry is evolving like computing did. Just as we went from mainframes to distributed systems and monolithic apps to microservices, AI is following the same path: models are getting smaller and better. Mixture of Experts is here to stay. Qwen3 is sick. Llama 3.2 runs on phones. Phi-3 fits on edge devices. Domain models beat GPT-5 on specific tasks.<p>RAG brings specialized data to your model: You don&#x27;t need a 1T parameter model that &quot;knows everything.&quot; You need a smart model that can read <i>your</i> data. Fine-tuning is democratizing: what cost $100k last year now costs $500. Every company will have custom models.<p>Data gravity is real: Your data wants to stay where it is: on-prem, in your AWS account, on employee laptops.<p>Bottom line: LlamaFarm turns AI from experiments into repeatable, secure releases, so teams can ship fast.<p>What we have working today: Full RAG pipeline: 15+ document formats, programmatic extraction (no LLM calls needed), vector-database embedding, universal model layer that runs the same code for 25+ providers, automatic failover, cost-based routing; Truly portable: Identical behavior from laptop → datacenter → cloud; Real deployment: Docker Compose works now with Kubernetes basics and cloud templates on the way.<p>Check out our readme&#x2F;quickstart for easy install instructions: <a href="https:&#x2F;&#x2F;github.com&#x2F;llama-farm&#x2F;llamafarm?tab=readme-ov-file#-quickstart-tldr" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;llama-farm&#x2F;llamafarm?tab=readme-ov-file#-...</a><p>Or just grab a binary for your platform directly from the latest release: <a href="https:&#x2F;&#x2F;github.com&#x2F;llama-farm&#x2F;llamafarm&#x2F;releases&#x2F;latest" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;llama-farm&#x2F;llamafarm&#x2F;releases&#x2F;latest</a><p>The vision is to be able to run, update, and continuously fine-tune dozens of models across environments with built-in RAG and evaluations, all wrapped in a self-healing runtime. We have an MVP of that today (with a lot more to do!).<p>We’d love to hear your feedback! Think we’re way off? Spot on? Want us to build something for your specific use case? We’re here for all your comments!

[Testing] Show HN: FizzBee – Formal Model based autonomous testing GitHub: <a href="https:&#x2F;&#x2F;github.com&#x2F;fizzbee-io&#x2F;fizzbee-mbt-examples" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;fizzbee-io&#x2F;fizzbee-mbt-examples</a> Quick Start: <a href="https:&#x2F;&#x2F;fizzbee.io&#x2F;testing&#x2F;tutorials&#x2F;quick-start&#x2F;" rel="nofollow">https:&#x2F;&#x2F;fizzbee.io&#x2F;testing&#x2F;tutorials&#x2F;quick-start&#x2F;</a><p>Most developers agree testing is important. At the same time, most developers don’t enjoy writing tests. With AI generating code faster than ever, testing is becoming even more crucial. But even AI-generated tests need review and maintenance, which makes them another burden.<p>I&#x27;m introducing another form of autonomous testing - &quot;model-based testing&quot;. Instead of writing test cases, you describe expected behavior in a Python-like specification language.<p>The FizzBee model can be: - Verified exhaustively for design bugs (like formal methods). - Mapped to your actual system, automatically generating the tests.<p>This gives you:<p>- No hand-crafted test cases - Automatic testing of concurrent as well as sequential behavior - No cascading test rewrites when behavior changes - No cluttering the SUT with tracing code<p>With FizzBee, you get both design validation (like in formal methods) and automatic test generation, saving time and effort.<p>Currently, only Go is supported. Java and Rust are next and would love to hear which language you’d want supported next.<p>I’d love your feedback!

[API/SDK] Show HN: Mix – Open-source multimodal agents SDK Why we built it: • Claude Code: great for coding, but no video&#x2F;audio support, localhost only • OpenAI SDK: single-model, no native multimedia tools • Both: no integrated DevTools for debugging agent reasoning<p>So, we built Mix as an alternative for multimodal applications. • Native video&#x2F;audio&#x2F;PDF analysis tools (via Gemini for vision, Claude for reasoning) • Multi-model routing instead of single-provider lock-in • One-command Supabase setup for cloud deployment (vs localhost-only) • HTTP architecture that enables visual DevTools alongside agent workflows • Go backend: 50-80% lower memory footprint than Node.js—efficient for concurrent agent sessions. Python and typescript clients are available<p>Example use cases in the demo video: portfolio analyzer that reads Excel and generates charts, YouTube search agent that finds and edits video clips.<p>GitHub: <a href="https:&#x2F;&#x2F;github.com&#x2F;recreate-run&#x2F;mix" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;recreate-run&#x2F;mix</a> Demo video: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;IwgKt68wQSc" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;IwgKt68wQSc</a><p>Would appreciate feedback, especially from folks building multimodal agents.

[Other] Stress test for parallel disk i/o using git and pnpm

[Other] A lightweight, powerful framework for multi-agent workflows

[DevOps] Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops (macOS, Linux, Windows).

[Other] Show HN: 1-Bit Pixel Art Font Editor I like making small fonts for Arduino and ESP32 projects with those little SSD1306 OLED screens. So I made this browser-based tool to help design pixel art fonts for use in those embedded projects.<p>It has a preview area that updates when you save your character. It can support fixed-width and variable-width fonts. And when you&#x27;re done you can export the font in a few different formats.<p>It comes with a bunch of example fonts built-in. Also it&#x27;s fully client-side code so you can save the webpage to use it offline.<p>If you want some inspiration for fonts, I made a bunch of really small fonts a few years ago: <a href="https:&#x2F;&#x2F;www.moonbench.xyz&#x2F;projects&#x2F;tiny-pixel-art-fonts&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.moonbench.xyz&#x2F;projects&#x2F;tiny-pixel-art-fonts&#x2F;</a>

[CLI Tool] Show HN: NewsGoat – A terminal-based RSS reader written in Go

[Other] A chat-centric workspace for builders and modern teams Orchestra makes workflow simple for teams who build products and rely on communication. With messaging at its core, and a task tracker and AI agents natively built in, work contexts are aligned and teams are connected.