The task
People often know what they need but struggle to explain it to an AI assistant. Promptchi focuses on making the instruction clearer while preserving the original intent, language and supplied facts.
My role
The project covers prompt design, the Python API, the browser extension and evaluation scenarios. The product is centered on improving text that the user can inspect and reuse.
How it works
A user submits a draft or selected text. The service improves the prompt, can ask clarification questions and applies the answers to the next version. The result can be copied into the intended AI tool.
Technical decisions
The backend exposes FastAPI endpoints for improvement and refinement. The Manifest V3 extension provides the browser interface. Provider adapters and prompt presets keep the model integration separate from interaction logic.
Reliability and boundaries
Evaluation scenarios cover language, preservation of user information and unsupported additions. Quotas are checked on the server. The tool edits the prompt; it does not perform external actions on the user’s behalf.
What was delivered
The implemented product prototype includes the improvement and refinement API, a browser extension and dedicated evaluation tooling. Further development focuses on checking complete user journeys and refining the product through practical feedback.