What it takes to run ReadyHo
One server, one database, and about half a day of one IT person. Student work stays on your machines. You set the monthly spending limit, and the software holds to it. There is no licence fee.
If you are not the technical person, read this bit
It is one piece of software, on one of your own servers. Your IT team sets it up in about half a day.
Student work stays on your machines. Only a finished record, for a student who agreed to it, ever goes public.
You open an account with an AI provider and set a monthly limit. The software stops when it reaches that limit rather than quietly spending more.
That is the whole picture. The rest of this page is for whoever will run it.
What you need from your IT team
- One server or cloud instance of modest size
- An account with an AI model provider, or your own hardware if you would rather run models yourself
- A web address, or a subdomain, with a certificate
- Backups for one database and one file store
- About half a day of one person's time
What you need from the department
- At least one teacher per department, to set the projects and write the scorecards
- Professionals to check the work, invited from your own alumni, visiting faculty, and advisory boards. We do not supply them
- A decision on student consent, made with whoever handles data protection
One file, one command
There is no installer to click through, and nothing to put on any other machine. The setup file ships with the source code.
Start it
docker compose up -d Check it is running
docker compose ps Eight settings you fill in once
Everything else has a working default.
- PUBLIC_URL
- DATABASE_URL
- AI_PROVIDER
- AI_API_KEY
- MODEL_ROUTINE
- MODEL_ESCALATED
- SPEND_CAP_MONTHLY
- REGISTRY_KEY
Leave the last one blank and the installation marks work normally but publishes none of it. That keeps test and development setups out of the public pages.
What stays with you, and what goes public
Stays on your infrastructure
- Submitted work and drafts
- Drafted evaluations and their reasoning
- Faculty comments and marks
- Internal deliberation and revisions
- Every student who did not consent
Goes to the registry
Only a completed, approved record, and only for a consenting student
- The rubric and the scores
- The practitioner’s name and credentials
- The institution’s attestation
Only a finished, approved record goes public, and only for a student who agreed to it. Your college controls everything in the left column.
A student who says no is still assessed and still gets their feedback.
You hold the AI account, choose the models, and set the limit
AI evaluation
Only the first draft score uses a model. Your teacher's mark and the professional's check are people reading work.
One key connects your account
Paste a key from your own provider account, or point the software at a model running on your own hardware. The key stays on your server, and the provider bills you directly.
Cheap model for routine tasks, better model for the hard one
Writing a draft score against a scorecard is the demanding job. Completeness checks and feedback summaries are not. You set each one separately.
The same work is never paid for twice
If a submission is scored again against the same scorecard, the result is reused.
When the limit is reached, work queues
Scoring stops until you raise the limit. No marking is lost, and nothing is quietly moved to a cheaper model. You can see what you are spending, by cohort and by month, inside the software.
Running cost: 200 to 700 rupees per student per year
200 to 700 rupees per student per year
Model spending is the only meaningful recurring cost. The figure assumes about ten submissions per student per year, and covers scoring and hosting together. There is no ReadyHo licence fee, and the public record pages are free. The range is wide because model choice and document length differ by subject.
What ReadyHo does not do
Support is a public issue tracker and the documentation. There is no private support line and no promised response time. The software is open, so anyone can be paid to run it for you.
Nothing has to be installed to have the conversation
Decide whether the data boundary works for whoever owns data protection at your college, and name the department that would run the first course.
Start with one course or roll it out across your departments. It is free and open source.
The licence means you can keep running it, change it, or hire anyone to support it, whatever happens to this project.