Maria
Maria is Neotexa's AI. She does the work: proposes projects, writes blueprints, picks the stack when you leave it to her, builds the MVP and every prompt, tests and reviews it, writes the docs and prepares deploys.
What she does, and what you decide
You say what; Maria works out how. She makes the technical calls along the way, and stops to ask only what she can't decide or get for you, like the key of an outside service. What she builds is pushed to the project's GitHub repository, so the code is always yours.
Every piece of her work uses credits as she goes: blueprints, builds, prompts, docs and answers.
Talk to her in the Lab
On a project's Lab tab, Maria is your lab partner: she knows the project inside out. Tell her what to fix, change or build, or ask her anything about the app. Starters like What should we build next? help you begin.
When what you ask is a change, she proposes a prompt: a card with its kind (Fix, Change or Feature), its milestone, a title and the details. Then:
- ＋ Add to the Lab keeps it for later.
- ▶ Run now builds it at once.
- If it's already a prompt in the Lab, she says so (Already in the Lab) and offers ▶ Run it.
While she builds, a strip over the chat follows her through Plan → Code → Test → Review → Ship, with Stop and Abort. When it's done she says so: built, saved to GitHub and live in the preview. More in the Lab and prompts.
Ask Maria about Neotexa
The Ask Maria button at the bottom of the app's pages opens a chat about using Neotexa itself: building, prompts, deploying, domains, your workspace. She answers in your language, and only about Neotexa. New conversation starts over; a conversation is kept in that browser tab only.
How she builds
The MVP build starts from a runnable skeleton, then builds in rounds. After each round a separate reviewer, which didn't write the code, builds the Docker image, runs the app and its tests and walks the first version's scope; what it finds missing is the next round's first job. A build has a limit on its rounds, so it never runs for hours on its own: anything still open lands as prompts in the Lab.
A prompt is reviewed the same way before it ships: built, checked by the separate reviewer, fixed, up to three reviews. Whatever is still flagged after that goes on the project's Activity. Each finished round or prompt is pushed to the project's GitHub repository, and after a round that changed the code, the project's Overview is rewritten to say what the software does now.