CASE STUDY · AI IMAGE ANALYSIS
The client already had an oral-image analysis capability. We wrapped it in a WeChat mini program: users shoot or pick a photo, a business service calls the analysis API and writes back findings, and check records plus an admin console turn a single API into a product people can actually use.
WHY IT WAS NEEDED
The brief was precise: let a user submit one oral photo and receive an analysis. The product did not need a full clinical suite. It did need a path that stably connects the image, the request, and the result.
User
No entry
The call has nowhere to land
Product entry
Without a mini program, submitting photos, waiting, and reading results depended on manual handoff. The capability could not be used—or operated—as a product.
Accuracy in the API does not matter if there is no entry.
Uploading / retry
Upload
Camera, album, preview, uploading, and failure all need visible feedback. Without it, people cannot tell whether the image reached analysis.
An unclear upload stops the flow at step one.
Advance the same check, don’t just spin
Analysis state
Submitting, analyzing, done, and failed belong on the same check. The client cannot pretend a tap returns a finding.
An incomplete state machine cannot bind a result to this submission.
This result
Gone after reading
Retention
Without check records, users cannot reopen history. Without an admin console, operations cannot query by user, time, or result.
A one-off response is not manageable check data.
Behind a short flow, the work is connecting capture, the analysis request, result handling, and durable check records.
HOW WE RETHINK IT
Calling an analysis API became a check a user can finish: the mini program captures and presents, the business service uploads, calls, and writes back, the analysis service stays independent, and admin only manages the data this product actually creates.
API only
A complete mini program
API only
No capture, wait, or review
A full check
Users do not need to understand the analysis service. They need to photograph and submit.
WHAT WE BUILT
Capabilities follow the real check path. If the analysis API changes later, the user path does not have to be rebuilt.
Preview & send
Bound to this check
IF THIS SOUNDS FAMILIAR
Image analysis, detection, or recognition that needs upload, invoke, display, and retention can follow this path.
HOW IT WORKS
The path is one check: open the mini program, start, capture or upload, submit, wait for analysis, read the result, and keep the record.
Users do not need to understand how analysis works. They photograph and submit.
HOW IT IS BUILT
The engineering focus is embedding analysis in the user flow: the photo reaches the service, the finding maps to this check, async states are visible, and history stays linked.
Architecture
The WeChat mini program handles capture, upload, waiting, and display. The business service processes the image, calls analysis, shapes the result, and writes the check. The analysis service only analyzes. Admin reviews users, photos, findings, and records.
WeChat mini program · Camera / album · Preview and submit · Status and result
Image handling · Analysis API call · Result shaping · Check persistence
Receive image · Analyze · Return findings · Decoupled from UI
Users · Photo review · Findings and checks · Query and export
Keywords
WHAT SHIPPED
Development is complete and the product is live. This lists the delivered scope. Call volumes and clinical metrics are not disclosed.
WHERE ELSE IT FITS
Any “image in, finding out” capability can reuse mini program, business service, and admin.
Intraoral photos, region findings, check history
Photo upload, itemized results, history comparison
On-site capture, async results, admin image review
Material upload, archived findings, export by user
Wrap a live analysis API in a mini program, app, or web entry
NEXT STEP
The first step is often not retraining a model. It is connecting upload, the API call, result display, and retention inside a product people can use.
Related cases
Other projects delivered by Wavesteam — AI, IoT, platform builds and enterprise software.
A branded retail mini program built around growth-led user journeys, with visualized outcomes for marketing and operations teams.
Read the case study
Systematized resume screening, talent search and pipeline analytics — built to cut manual triage and handover cost for in-house HR teams.
Read the case study
A donation and student-aid platform for non-profit organizations — platform architecture and user flows tailored to charity operations.
Read the case studyThese projects usually involve business systems, device integration, AI workflows or multi-role back-offices. We assess feasibility against real delivery constraints and give recommendations close to the implementation stage.
