Scriba — AI Writing Assistant (API + Cloud + AI)

• 

  Scriba — AI Writing Assistant (API + Cloud + AI)

Inspiration

Inspiration

Most AI writing tools are either locked to one vendor or silent about where the text comes from. We wanted a writing partner that is (a) vendor-neutral, (b) runs entirely in the cloud with no setup, and (c) can ground its output in real, current facts instead of guessing. DevNetwork's "API + Cloud + AI" theme was the perfect frame: Scriba is literally an API-driven, cloud-hosted, AI-powered app.

What it does

Scriba guides you through the writing lifecycle with three modes:

• Generate — co-write a complete piece from a one-line idea.
• Optimize — rewrite messy text into clean, professional copy.
• Summarize — distill long text into key points.

In Generate mode you can enable live web grounding: Scriba queries the SerpApi search API and cites real sources inline, so drafts are fact-checked, not fabricated.

How we built it

• Frontend: Next.js (Pages Router) + React.
• AI layer (the API): a small serverless route (/api/generate) calls any OpenAI-compatible chat-completions endpoint — here the agnes relay — with an optional Gemini fallback. Vendor-neutral by design.
• Cloud (the Cloud): deployed as a serverless app on Vercel; zero local setup.
• Live search (the AI + API): /api/search calls SerpApi when SERPAPI_API_KEY is set, returning the top 5 results that are injected as grounding context.
• Memory: /api/generate also stores and retrieves context via HydraDB, so the assistant personalizes from past interactions.

Everything is MIT-licensed and open source.

Challenges we ran into

Keeping the AI layer truly vendor-neutral while still being trivial to deploy. We solved it with an env-driven provider swap (agnes / OpenAI / Gemini) and a graceful SerpApi fallback that degrades to pure-LLM generation when no search key is present.

Accomplishments that we're proud of

A clean, dependency-light writing assistant that anyone can fork and run with their own API key in minutes — and a web-grounding feature that makes AI writing citeable.

What we learned

How little code it takes to wire multiple APIs (LLM + search + memory) into one coherent cloud experience, and how sponsor APIs like SerpApi turn a generic app into a targeted, prize-eligible submission.

What's next

• Multi-document grounding and source export.
• More modes (translate, rewrite-for-tone).
• A public API so other tools can embed Scriba's writing pipeline.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Scriba — AI Writing Assistant (API + Cloud + AI)

Built With

• cloud
• css
• gemini
• google
• html
• javascript
• next.js
• node.js
• openai
• react
• vercel