... into a messy pipeline issue recently where an LLM was supposed to clean up incomplete user inputs before a database lookup. The problem was that downstream functions started treating the model's cleaned-up version as the actual ground truth instead of just a search hint. It ended up matching fake data with real database entries because it was basically judging its own guesses instead of what the user originally gave us. Now I keep raw user data completely locked ...

... history couldn t always be one tap away. So I built Auto-Mobile . What it does: Maintenance log for every vehicle you own Part numbers Fuel and EV expense tracking, with receipt scanning so you skip the typing VIN lookup to fill in vehicle details quickly Warning light library, so you can look up what a dashboard symbol means Document vault for insurance, registration, AAA and similar cards, with images and PDFs, encrypted and lockable Backup, so a lost phone ... ... collision, your entire vehicle history is right there at your fingertips where you need it. Privacy Your records are stored encrypted on your phone first, not in someone else s cloud. Only two features leave your phone, the VIN lookup

... answer in your material and uses it. This works with long, multi-page documents. Lets you to add or update content without rewriting the agent and you can switch it off for agents that do not need lookups. A short pause can happen while it looks something up. The answer is worth the wait. What is the one document you would upload first? Comment from Chandrashekran Y(@chandrashekran_y): Would love for everyone to try out the Knowledge base feature ...

... exactly the pattern. you set SPF, DKIM and DMARC once at launch, then never look again. Until the welcome email lands in spam and nobody knows why. The 3 things that usually break later: SPF goes over 10 DNS lookups after you add one more tool the tool that actually sends your emails has no DKIM DMARC stays at p=none forever, or doesn't exist at all That's why we built mailX. Type your domain ...

... clever. I came up with it after my own dashboard showed $0 on 176 requests that weren't free. An integration sent its requests with the provider set to "gemini". My price table filed those models under "google". The lookup found nothing, so the cost was stored as $0 and the status as Success. Nothing logged an error, because as far as the code knew, nothing had gone wrong. Cost is calculated once, when a request lands, so adding ...

... language switcher and the internal links on every localised page were supposed to keep visitors in their language. For months they did not. Every locale page quietly linked back to the English homepage. The cause was one line. A lookup used slashless keys like /for/bakeries while the site's own canonical path helper appended a slash. Every lookup

... switch that said no was not being obeyed. The Overview builds its link panel three different ways depending on how wide the window is. The one the default window size lands in skipped the guard, so with the outbound lookup switched off, that panel still reached outside your network. The two arrangements that were correct were the rare ones which is exactly why it survived this long. Plus: three things were showing English to people who do not read ...

... card says exactly that. Two warnings before they bite: runs out within three weeks, or not valid yet (which is usually the device's clock, not the certificate). On demand only, behind the same privacy switch as the service lookup. Netfox never starts a conversation unasked, and the card stays hidden in Demo Mode. The service lookup

... Both are still requests. The one enforced version is a JSON schema with a reasoning field placed ahead of the answer field, because structured outputs keep schema order. Our take: Demand the answer first when the answer is a lookup, a path, a command or a status. Let the reasoning stay on the page, briefly, when the answer is a judgment and the model has no hidden channel. Put the rule where it will still be in front ...

... looks more like: long-context-first for anything under ~100k tokens of source material, agentic multi-hop retrieval for anything bigger or dynamic, and RAG-as-embeddings-search only as a last resort for pure "needle in haystack" lookup. What's actually in your retrieval stack right now - and did you rip out RAG, or double down ...

... router with no WAN, a home router that lost its upstream link, all read as online while every real request quietly times out. Ended up not trusting the OS flag at all. The check now does a real DNS lookup against a known host with a short timeout before marking anything online, and only flips the app into offline mode once that actually fails. Adds a few hundred milliseconds to every check, worth it for not showing a broken screen ...

... dashboards? This is the operational layer of AI development, and it is easy to underestimate. We built ChinaRouter to make this layer easier to manage. ChinaRouter is a unified AI API gateway and admin dashboard for model access, usage lookup, documentation, and dashboard workflows. Website: https://chinarouter.net/ For people building AI products: what part of AI API management becomes painful first ...

... small teams, the best tools are usually the ones that reduce operational friction. ChinaRouter is our attempt to create a cleaner AI API workflow for developers, makers, and teams. It brings model marketplace access, API gateway workflow, usage lookup, documentation, and dashboard management into one place. Website: https://chinarouter.net/ If you are an indie maker using AI APIs, what tools do you wish existed ...

... your team? Four months of building forced us to separate four different things: a source commit proves code exists, a passing test proves a defined behavior, a notarization receipt proves an artifact cleared a gate, and a live-store lookup proves distribution. None of those proves adoption, revenue, or a customer outcome. We ve started treating completion as a proof ladder instead of one status: implemented built installed live customer-ready. It has made product reviews less dramatic and much ...

... overhead: 1. In-Memory Redis Caching: We cached high-frequency read-only data with a 5-minute TTL (Time To Live). This dropped direct database hits by over 60%. 2. Query Optimization: Replaced heavy nested SQL queries with indexed lookups, reducing response time from ~320ms to under 45ms. 3. Asynchronous Task Queue: Moved non-essential tasks (like sending welcome emails or triggering notifications) to background workers using Celery. This simple optimization not only handled 5x higher concurrent users during stress ...

... Over the last few weeks I noticed that most email tools were scattered across different websites. You'd use one site for SPF, another for DKIM, another for MX lookup, and another for temporary email. So I decided to build @FreeMailTools a growing collection of free email utilities in one place. Current features include: Email Validator Temporary Email MX Lookup

... built a fun spa called www.1000songs.co.uk it works on lookups from iTunes. One feature is a name that tune feature I have fairly niche tastes so I also built a "create collection" import where you can paste in in the format year, artist, song and play the games! Example format below 1964, The Beatles, I Want to Hold Your Hand 1971, Led Zeppelin, Stairway to Heaven 1977, Bee Gees, Stayin' Alive 1983, Michael Jackson, Billie Jean 1984, Prince, Purple Rain ...

... taxes tracked by hand, receipts in shoeboxes, brokers who ghost on payment, quarterly filings built from sticky notes. So I've been building the back office: automatic per-state GPS mileage for IFTA, receipt + expense tracking, invoicing, FMCSA broker lookups before you book a load, route weather with air quality, and a live driver chat. It's free right now while in beta, and every feature so far came from a real driver complaint. Happy to swap notes with anyone ...

... driving: IFTA fuel taxes tracked by hand, receipts in shoeboxes, brokers who ghost on payment, quarterly filings built from sticky notes. So I've been building automatic per-state GPS mileage for IFTA, receipt + expense tracking, invoicing, FMCSA broker lookups before you book a load, route weather with air quality, and a live driver chat. It's free right now while in beta, and every feature so far came from a real driver complaint. Happy to swap notes with anyone ...

... Relay a runtime security proxy that sits between the agent's decision and the actual tool execution. How it works: Deterministic Policy Gate: Instead of an LLM guessing if an action is safe, Relay does an O(1) hash lookup on the tool name. (e.g., github_delete_repo = blocked, github_read = allowed). Zero latency, zero hallucinations. Networkless Sandboxing: For coding agents, you can't regex-check every bash command. Relay routes shell tool calls into an ephemeral Docker container with ... ... their AppSec/CISO review. It's free to test right now. Would love for you to tear it apart and give me feedback on the architecture. Live app & docs: https://relay-security-lemon.ver... What do you guys think about the "deterministic vs

... learned building this: Intent detection is the core problem. The AI needs to understand not just what the user said, but what they actually need done. "Check if John is on a paid plan" should trigger a database lookup, not a paragraph explaining what a paid plan is. Isolation matters more than you think. Each action needs to be stateless and sandboxed. When the AI is calling functions autonomously, you need guaranteed blast radius one bad call shouldn't affect ... ... debug. Every action call needs full params + response logged so you can understand exactly what the AI did and why. Curious how others are thinking about this: How are you deciding which actions to expose to your AI vs

... Product Hunt! I'm the solo maker behind PeekMyIP a free privacy & network toolkit. Instant IP lookup, DNS leak test, proxy check, browser fingerprint and more. No sign-up, no ads. Launching this Tuesday May 12th would love your support! Built it because I was tired of IP tools cluttered with ads that required accounts just to see basic info. Happy to connect with other makers here! Comment from Rian Robertson(@rianbrob): Love this...a clean, ad-free toolkit ...

... learned building this: Intent detection is the core problem. The AI needs to understand not just what the user said, but what they actually need done. "Check if John is on a paid plan" should trigger a database lookup, not a paragraph explaining what a paid plan is. Isolation matters more than you think. Each action needs to be stateless and sandboxed. When the AI is calling functions autonomously, you need guaranteed blast radius one bad call shouldn't affect ... ... debug. Every action call needs full params + response logged so you can understand exactly what the AI did and why. Curious how others are thinking about this: How are you deciding which actions to expose to your AI vs

... break your flow. Quicktionary is a lightweight floating dictionary for Android that lets you check meanings instantly without leaving the app you re using. What makes it useful: - Floating overlay works on top of any app - Quick word lookup in seconds - Minimal, distraction-free UI - Designed for reading, browsing, and learning on the go - Quiz - much more.. Whether you're reading articles, chatting, or scrolling, Quicktionary keeps definitions just a tap away. Built for speed and simplicity no unnecessary ...

... that were killing our sender reputation. Most email verifiers do a simple syntax check and call it done. We weren't satisfied with that, so we built a 6-layer verification engine that checks: Syntax validation Domain & DNS lookup MX record check SMTP handshake Disposable/temporary email detection Role-based address flagging The result? 99.8% accuracy, and over 50,000 users have run their lists through it. We offer: - Real-time API for developers (integrate at signup forms) - Bulk list ...