Whitepaper · Build or buy
Build it from open source? Count the seams first.
Every part of an enterprise AI stack has an open source equivalent. This paper counts the work between them, shows what Curiosity Studio ships in its place, and why that makes it the place to describe an app and ship it.
What it covers
The parts are free. The seams are not.
A graph database, a search engine, a vector store, connectors, OCR and an agent framework can each be had for free. The cost is the code between them: permissions copied into every index, deletes that reach every copy, one audit trail across all of it. Studio ships those seams closed.
- Part 01
Connectors
Sync jobs per source
- Part 02
Parsing
Files and OCR
- Part 03
Graph
The links between records
- Part 04
Search
Keywords
- Part 05
Vectors
Meaning
- Part 06
AI agents
Tools on top
Chapters
Seven chapters, from the parts to prompt to production
01 · Page 03Chapter 01
Every part exists. The seams do not
What the assembled stack leaves your team to build.
02 · Page 06Chapter 02
Integration carries its permissions
Connectors, extraction and OCR that write one graph.
03 · Page 09Chapter 03
One permission model, on every read
Why access checked in one place beats many copies.
04 · Page 12Chapters 04 · 05
One process, run as one system
One store, one transaction log, sizing, replicas and backups.
05 · Page 18Chapter 06
Describe it, then ship it
Why an app described in a sentence needs a platform under it.
06 · Page 22Chapter 07
When to build it yourself
The trade-offs, and a test on your own data.
Your data. Your infrastructure.