RAG over PDF manuals
Manual pages are chunked, embedded and searched with pgvector so the model gets the most relevant procedures.
AI OPERATIONS ASSISTANT
FieldOps AI — every tech, instantly expert
Your most experienced technician's knowledge shouldn't live only in their head or buried in PDFs. This assistant turns manuals, customer records, and service history into instant, sourced answers anyone on the team can get.
Live demo — try it right here
Real OpenAI + Supabase, not a recording.
The embedded demo is taking longer than expected
If the demo is blocked by your browser, open it directly in a new tab.
Open demo in a new tab ↗If the demo is blocked by your browser, open it directly in a new tab.Open demo in a new tab →
How it’s built
One answer, two kinds of business knowledge.
The agent retrieves only what the question needs, then combines structured CRM records with the relevant manual pages into one useful response.
CRM tools via function calling
The agent looks up customers, installed equipment and service history through server-side tools instead of dumping the database into a prompt.
Page-level citations
Every manual-backed recommendation links back to a concrete document and page, making the answer easy to verify.
Live call suggestions
A rolling transcript triggers focused checks during a service call, without sending a request for every spoken word.
Question→CRM + Manuals→Sourced answer→Live call assist
Technology
A focused, server-side AI stack.
Next.js and TypeScript run the interface and server routes. Supabase provides Postgres, pgvector and Auth. OpenAI handles embeddings, tool calling and answers; Vercel hosts the application.
Next.jsTypeScriptSupabasePostgrespgvectorAuthOpenAIVercel
Build something useful
Have operational knowledge scattered across systems and documents?
I can turn it into a focused assistant your team can use in the office or in the field.
NEXT.JS · TYPESCRIPT · SUPABASE · POSTGRES · PGVECTOR · OPENAI · VERCEL