Browse categories
Explore
Fiverr Pro
English
$
USD
I will test your ai chatbot, rag app or ai agent before launch
About this Gig
I'll test your AI app for unsupported answers, context loss, prompt injection and incorrect tool calls. You'll receive documented results, reproducible issues and priorities for fixing them. Premium includes automated regression evaluations.
Testing application:
Software
Development technology:
.NET
•
Go
•
Python
•
SQL
•
TypeScript
Device:
PC
•
Linux
•
Android mobile phone
My Portfolio
FAQ
What counts as a test scenario?
One conversation of up to five user turns with a defined expected behaviour. Basic covers one use case; Standard and Premium cover up to three. Each package covers one application, environment and model configuration.
Do you need access to my source code?
Basic and Standard usually need only a staging URL or documented API and a test account. Premium also needs a repeatable API or callable interface. I can deliver the evaluation code in a separate repository.
Is testing manual or automated?
Basic and Standard are hands-on testing. Premium adds automated evaluations for 20 agreed scenarios, combining deterministic checks with an LLM scoring rubric. It includes setup instructions and one GitHub Actions workflow.
Are API costs included?
No. You provide funded access to the application and approve any external model used for evaluation. We'll agree a spending cap before running tests.
Does this include security testing or fixing the app?
It includes agreed prompt-injection and AI behaviour checks. It does not include infrastructure pentesting or code fixes. Standard and Premium cover at most one tool integration; more complex agent workflows need a custom quote.
What if you don't find any serious bugs?
You'll still receive the scenario log, observed results and testing limitations. A test pass cannot guarantee that an AI system will never fail. API-only orders use request/response logs in place of screenshots.
What do revisions and retesting cover?
Revisions correct or clarify the agreed deliverable. Standard and Premium also include one retest of up to five reported issues, requested within seven days of delivery. New features, model configurations and additional scenarios require a separate order.
