Build a WeChat Mini-Program in 2 Hours with AiOnly + WorkBuddy: A $5K Case Study

Project Overview

The author, a developer, accepted a freelance mini-program order and delivered it quickly using AiOnly (a large-model aggregation service) and WorkBuddy (an AI coding assistant). The project was an event-registration mini-program priced at 5,000 RMB, with core development completed in about two hours. The article walks through the entire process: scoping, tool setup, prompt engineering, iterative generation, verification, and deployment.

Scope Definition

Before coding, the author defined a minimal viable feature set to keep the timeline tight:

Display event details (name, description, time, location, sessions, notices).

Collect registrant name, phone number, and session choice with validation.

Prevent duplicate submissions (same phone per event) and double-clicks.

Save data to cloud; show success only after confirmed write, with clear error messages.

Admin can view, search, and export CSV of registrations; regular users see only their own record.

Deferred features: payments, coupons, distribution, complex membership.

Clear scope is emphasized as the key to controlling development time.

Toolchain: AiOnly + WorkBuddy

AiOnly aggregates models (DeepSeek, GLM, Kimi, MiniMax, etc.) via a compatible API. WorkBuddy is the IDE-like client that uses those models for code generation, modification, and debugging. The author connects them by:

Registering on AiOnly (console provides API base URL, API Key, Model ID).

In WorkBuddy, adding a custom model with those three values.

Verifying connectivity with a test message and a file-creation task.

Api endpoint used: https://llm.aiionly.com/v1/chat/completions (for Token Plan users, use the plan-specific URL). Model ID must match exactly what the platform provides, and the model must support tool calling.

Pricing Reference (as of 2026-10-11)

Lite: 29 RMB/month, 2,175,000 credits

Standard: 79 RMB/month, 6,320,000 credits

Pro: 249 RMB/month, 21,165,000 credits

Max: 499 RMB/month, 44,910,000 credits

Credits expire each cycle; actual models and prices subject to purchase page.

Development Timeline (2 Hours)

First 20 min: Write detailed prompt (see below) and define acceptance criteria.

Next 60 min: Generate pages, implement registration and admin features via WorkBuddy dialogue.

Next 30 min: Verify data integrity, permissions, and mobile operation.

Final 10 min: Prepare handoff notes, to-do list, and release steps.

Prompt Given to WorkBuddy

请帮我制作一个微信原生活动报名小程序。 用户端包含活动首页、报名页和报名结果页。首页展示活动名称、介绍、时间、地点、参加场次和报名须知。 报名表包含姓名、手机号和参加场次。必填项为空时不能提交，手机号需要格式校验。 同一活动中，同一手机号不能重复报名。提交过程中禁用按钮，避免连续点击产生重复记录。 数据需要保存到云端。只有保存成功后才能显示报名成功；保存失败时提供明确提示。 管理员可以查看、搜索报名记录，并导出 CSV。普通用户只能访问自己的报名记录，管理员权限需要在服务端校验。

WorkBuddy can generate a native WeChat mini-program project, optionally wiring cloud database, auth, and file storage.

Iterative Refinement

After the first version, the author used follow-up prompts to adjust UI and harden logic:

Layout tweak: "Move event time and location to top, keep only form and submit button below."

Success page: "Add 'View My Registration' entry and remind user to keep info."

Edge-case audit: "Check duplicate registration, rapid clicks, network errors, save failures. Report each check result and fix issues; don't just reply 'done'."

Manual smoke test followed: normal submit, duplicate block, reopen persistence, admin export completeness, and concurrency simulation.

Testing & Deployment

WorkBuddy's preview is good for UI and basic interaction. Real WeChat runtime (login, etc.) requires publishing a trial or experience version and testing on a phone. Trial version lasts 14 days, max 15 users, cannot convert directly to production. For long-term operation, bind a registered mini-program account, complete ICP filing and WeChat review.

The mini-program only handles business data at runtime; AiOnly model calls happen inside WorkBuddy during development, so no model keys are embedded in the frontend. If AI features (Q&A, content generation) are needed later, design a separate backend with usage controls.

Key Takeaways

Start with a well-scoped small project (event sign-up, store booking, equipment loan, internal request).

Write explicit, testable requirements before invoking AI.

Use model aggregation (AiOnly) to avoid scattered keys and quotas.

Let the AI assistant (WorkBuddy) handle boilerplate and iteration; human verifies logic, data, and UX.

Reserve time for real-device testing and platform review.

The "2-hour build" and "5K fee" are outcomes of disciplined scoping, prompt precision, and rigorous validation — not just raw generation speed.

Illustrative Screenshots

The article includes several WorkBuddy UI screenshots showing model selection, project generation, and final code structure. They illustrate the workflow but are omitted here for brevity.