Zhoukai960 4891·BabyLovable: an open-source coding agent with AI SDK, Workflow and SandboxHi Vercel community! I’m building BabyLovable, an open-source, multi-user coding agent that turns a description into a Next.js app in an isolated workspace, with a live preview and iterative edits through chat. GitHub: github.com/1nFrastr/baby-lovable Demo:...
Santosh Sibyala·StreamTrust: open-source CI checks for AI SDK chat persistence, Stop, and retryHi everyone—I built StreamTrust, an MIT-licensed CI verifier for streaming AI apps. Version 0.3.0 adds a Vercel AI SDK preset checking saved answers, Stop propagation, regeneration history, and retry after a provider disconnect using a real browser and a controlled model...
Abhivarde·last commentTryeve // an agent runtime for Evedescribe an agent. get a working one. features build & test ↳ describe an agent in plain english ↳ generates real, working eve files ↳ tested against a live eve runtime before you see it ↳ generation and testing run as one durable step, survives crashes ↳ if a build fails, the...12 comments from Abhivarde and BestCodes
Iobert·Extending eve's Twilio channel for multi-channel memory: what I learnedDisclosure: I work at Twilio. I've been building with eve and I like how it's structured. The built-in Twilio adapter is great for a quick SMS agent. At the version I tested (0.74.0), though, it doesn't forward inbound media to the model. It also treats SMS and WhatsApp from the...
Sneroboto·last commentPlayed with Vercel Lab's json-render with Jev to read Posthog dataHi everyone I had so much fun playing around with a chat that answers PostHog questions without opening PostHog. It uses Vercel Lab's json-render to pull analytics dynamically and displays in a way that's the least confusing to read. I am using Jev to get fast deterministic...Jonoroboto Sick thumbnail - love it. Can't wait to build a full blown accounting tool with Jevtope love to see this, welldone
Olivia Carter·Built a Cost Estimation Tool and Exploring an AI-Assisted VersionHi everyone, I’ve been working on a web-based cost estimation tool and wanted to share the current version while I explore where I can take it next. The current calculator guides users through a few simple inputs and generates an estimate based on things like the type and amount...
weekeight0418·I built Classp — real-time multilingual lecture notes with Next.js 16 and AI GatewayHey everyone — I’ve been building Classp, a lecture companion for university students who find it difficult to listen, understand, and take complete notes at the same time, especially when the lecture is not in their strongest language. The workflow starts when class begins:...
Wonbinbag31·last commentI built 3 mini apps as a solo developerHi! I’m a solo developer from Korea 🇰🇷 I recently launched three small apps: ✍️ WorkDraft — AI document creation 🎨 Nakseomong — expense tracking + doodling ⚡ Choseong Speed — Korean word reaction game They’re currently available as Toss Mini Apps in Korea. I’m starting from...Amy Egan so cute 😍
Anayat·Agentcn | Installable AI AgentsHey everyone 👋, I just launched AgentCN An open-source library of installable, customizable AI agents for modern applications. One-command installation Own the agent code Fully customizable Would love for you to check it out and share any feedback! Site: agentcn.dev Github:...
Zachi·ShotLuma - Free & Open Source AppStore Screenshot GeneratorI’ve been building Shotluma, an open-source editor for creating iOS App Store screenshots with the Power of Vercels AI-SDK. The main idea is pretty simple: I didn’t want the AI to generate a flat image that looks nice but becomes annoying the second you want to change something....
Olivier 4662·We built Terse AI - A way to build automations with your coding agent efficientlyWe built Terse, a workflow platform that reduces the cost and time to build workplace automations. We help coding agents build automations more efficiently by: Deterministically generating a typed SDK of your workspace. It gives the AI the building blocks it needs to efficiently...
Abhivarde·last commentNothify // Autonomous 404 page generation for Next.jsNothify is an AI-powered agent that analyzes a Next.js repository and generates a framework-correct 404 page, either as downloadable code or as a clean, review-ready GitHub pull request. What Nothify does: Analyzes a Next.js repository (App Router or Pages Router) Detects...+226 comments from Abhivarde, Maya Avendaño, Amy Egan and 1 other
Syedwali9286 1132·provider-guard: catches AI Gateway responses that bill tokens but return nothingAI Gateway falls back to another provider only when a request fails outright. But a provider can return a 200, bill you for output tokens, and deliver nothing usable, no text, no tool call. Since the request technically succeeded, nothing falls back, and you're charged for an...
Zhoukai960 4891·Coordinating durable coding agents with WorkflowAgent and ephemeral sandboxesHi! I’m exploring a durable coding-agent architecture in BabyLovable and would love to compare approaches with others using AI SDK and Workflow. The builder uses AI SDK v7 WorkflowAgent and Workflow DevKit for long-running agent turns. Each session has a remote Vercel Sandbox...
Filippo·OCR Model AvailabilityI noticed AI Gateway does not provide access to SOTA OCR models, or any OCR models for that matter. Would love to see all OCR models there. We have tons of volume and perfect use case for OCR models, and I'm sure more people do too. Pretty please.
Scrawny Minks5e 5702·last commentAI Gateway Always Returns 403 Despite Having More Than 6,000 Credits RemainingHi Vercel team, I’m experiencing an issue with Vercel AI Gateway. My Vercel account still has more than 6,000 credits remaining, and I have barely used them. However, when I try to use AI Gateway, the request returns: HTTP 403 Forbidden I would like to understand why AI Gateway...Skindowshao 3161 very well . me too.
roedyrustam·Showcase: SinapsAI – Zero-Dependency In-Process Circuit Breaker & Failover Adapter for Vercel AI SDKHi everyone! 👋 I've been building LLM-powered applications using Next.js 15 and the Vercel AI SDK (ai), and love the clean developer experience of streamText and React hooks. However, in production we constantly ran into two recurring bottlenecks: Provider Outages & 429 Rate...
ousama·AI Gateway drops the resolved Jev model versionFiled github.com/vercel/ai/issues/21213 against the AI SDK repo, where it was reproduced and then labelled out of scope, since the fix sits in the Gateway service rather than the package. Posting here so it reaches the right people. Through AI Gateway, typesafe-ai/jev is the only...
Frdwings·last commentAI Gateway returns HTTP 403 pre-inference for GPT-5.4 and GPT-5.6 Terra on Hobby teamHi Vercel Support, I'm experiencing consistent HTTP 403 responses from AI Gateway before inference starts when using higher-capability OpenAI models. The same project, AI Gateway API key and runtime configuration successfully execute openai/gpt-5.4-mini, including normal token...Frdwings Successful control: gen_01M2CQE67VEZVYMBTG88XF787H GPT-5.4-mini → OpenAI → HTTP 200 → inference completed Failed current control: gen_01M2EE8Y8GE0B21BARH3R3XW7T GPT-5.4-mini → Azure → HTTP 403 → pre-inference Gateway API Key ID: 2anmn74Bq2bVwWbkAWWEP09dres3z3ha85VOzqvUT9jJM2M0
Billy Kuykadia·ChatGPT Work MCP plugin returns 404/403 for specific team projectHello, We are using the ChatGPT Work Vercel plugin connected via OAuth to our team. The integration functions properly for other projects in the team, but fails specifically for one project: Team ID: team_iJRDogPoJ9r15Dhfae2ZnkBD Project ID: prj_PoJJrUW50TNwBlnl9lbYLqAJnsev...
Tadeu Tupinambá·last commentVeo 3.1 result exceeds ~16 MiB (undocumented inlineTooLarge)Hi Vercel team, We run video generation through AI Gateway's asynchronous job API (experimental_startVideo + experimental_getVideoStatus). For google/veo-3.1-generate-001 at 4K, the gateway reports a job as "completed" but returns an empty videos array whenever the rendered file...Tadeu Tupinambá OBSERVATIONS The largest inline payload ever returned to us is 16,123 KB. Re-running the dropped prompts dropped them again, and the delivered ones came through again, which points to a fixed size threshold rather than a transient fault. On dropped jobs expired is already true at completion time, while expiredAt is completion + 24 h, the same window a hosted result gets. The result appears to be discarded rather than stored. The status is "completed" with empty warnings. There is no error, no 413, and the webhook payload carries no bytes and sends the receiver back to the same status call. The
Kensnzk·AI Gateway streaming requests terminate at ~785–790s despite a 1800s Function durationAI Gateway terminates a streaming request after ~785–790 seconds with gateway_stream_timeout, although the calling Workflow step has Extended Max Duration enabled and is deployed with maxDuration: 1800. The provider started streaming after 6 seconds and continued returning...
Nick R.·Add photo support for ChatGPT real-time in AI GatewayRealtime sessions through the AI Gateway can't take an image. The normalized protocol only has text, audio, and function-call-output conversation items, so there's no way to hand a photo to openai/gpt-realtime-2.1 even though the underlying Realtime API accepts image input. The...
Khadjimurad·AI Gateway transcription drops Gemini speaker diarization (word_info / spk_*)Product: Vercel AI Gateway Feature: Speech-to-text / transcription Model: google/gemini-3.5-transcribe Endpoint: POST ai-gateway.vercel.sh/v4/ai/transcription-model Auth: AI Gateway API key (system credentials, not BYOK) Example generationId: gen_01M11K75WKZPWRXY37603K104G What I...
Udayganguru2000 1232·last commentIssue: Parameters output mismatchI configured a tool with zod version 4 schema and wrote an execute function as follows { runAgentTask: tool({ description: "Runs an agent with a task.", parameters: z.object({ task: z.string().describe("Complete task description.") }), execute: async (input) => {...Udayganguru2000 1232 May be I have to switch to lang graph, this is not working for some reason. advise me it it's a better idea.
Justin Leonard·Pro Credits and AI Gateway BillingHi Vercel team, I wanted to give some feedback on the way Pro account credits work, particularly with AI Gateway. I pay $20 per month for Pro, and in many months I only use a fraction of the included credit. The remainder simply expires. At the same time, those credits cannot be...
+2
