DEV Community: Weilu Wang

I Couldn't Grow Up with a Cat, So I Made a Django Package to Fix That

Weilu Wang — Wed, 03 Jun 2026 04:01:27 +0000

I always wanted a cat but never had one as a child. Decades later, I used Doubao API to merge my childhood photos with my cat's images, then built a reusable Django app (django-doubao) that does AI image fusion and image-to-video generation. Here's how it works and why I built it.

The Empty Space

When I was a kid, I longed for a cat. My mother wouldn't allow it. I pressed my face against the neighbour’s window, watching their tabby stretch and roll in the sun – a hollow ache in my chest.

Only after forty years did I finally get my own cat. But then a new sadness crept in: I couldn’t grow up with her.

So I devised a trick.

On Doubao (ByteDance’s AI platform), I took my old, faded childhood photos – gap-toothed, laughing – and merged them with my cat’s plump white face. Suddenly, we were playing together. Then I made videos: the child chasing the cat through a field of flowers, wind lifting our hair, years dissolving.

Each creation sent ripples through my heart – from a quiet centre to the corners of my lips and eyes.

A Programmer’s Solution

I started on Doubao’s official website, but soon I thought: I’m a programmer. Why not build my own app?

I got a Doubao API key, set up a Django project, and wrote a utils.py – not three hundred lines – that became the key to my own toy box.

My ritual:

Pick a beautiful landscape – the old locust tree at my childhood hutong, summer thick with cicadas.
Add my childhood face and my fat white cat (round as a snowball).
Merge them into one image where we play together.
Animate that still picture – the cat rolls on the grass, the child chases, flowers tremble in a breeze I only feel in memory.

I created two Django views:

The first fuses multiple reference images into one AI-generated picture.
The second turns that picture into a short video (5–12 seconds).

When I saw the boy and the cat move across the screen, something long empty inside me was gently filled.

I packed the whole thing into a Python package, django-doubao, and published it on GitHub.

What the Package Does

Multi‑image fusion – combine several reference images + text prompt → one new AI image.
Image‑to‑video generation – turn a single image (generated or uploaded) into a 5–12 sec video. Supports aspect ratios, resolutions, returns video + last frame.
Asynchronous polling – video generation is async; the package polls automatically. No while loops needed.
Task records – via Django models, so you can track or re‑create past generations.

Quick Architecture Overview

django-doubao/
├── doubao/
│   ├── models.py        # ImageGeneration, VideoGeneration
│   ├── views.py         # generate_image_view, generate_video_view
│   ├── urls.py          # routes with namespace
│   ├── utils.py         # API calls: generate_image, submit_video_task, poll_video_result
│   ├── templates/       # standalone Vue3 templates (no base.html required)
│   └── migrations/
├── tests/               # pytest + mock, 15 tests all green
├── pyproject.toml
└── README.md

Key decisions:

No built-in permissions – you wrap views with login_required or user_passes_test yourself. Each project is different.
Self-contained templates – they work even if your project has no base.html. You can override them.
Synchronous polling – good for low‑concurrency personal use; you can swap in Celery later.

Using django-doubao

Install:

pip install django-doubao

Add to INSTALLED_APPS:

INSTALLED_APPS = [
    ...
    'doubao',
]

Set your API key:

DOUBAO_API_KEY = "your-key-here"

Include URLs:

# urls.py
path('doubao/', include('doubao.urls')),

Then visit /doubao/image/ to generate images, and /doubao/video/ to turn them into videos.

Why I’m Sharing This

I could have kept using online tools. But as a programmer, I chose to build a small world with code – one where the child and the cat can finally run, daydream, and watch maple leaves fall together.

If you have a similar wish – to revive an old photograph or animate a memory – I hope django-doubao helps you take fewer detours.

Code is like bricks. The important thing isn’t how complex it is, but what you build with it.

The cat is purring beside me. The video loops on the screen. I turn off the light, leaving only the screen’s glow – and that old hollow feels truly filled.

What do you think? Have you ever used AI to fill a childhood gap? Let me know in the comments.

RAG in production: three levels from search to conversation

Weilu Wang — Fri, 29 May 2026 03:23:03 +0000

DEVELOPER. So, you want to build a website that thinks?

USER. I want it to answer questions. Not just search.

DEVELOPER. Then let’s walk the path. First stop: the quiet room.

(He gestures slowly, almost lazily.)

DEVELOPER. Level One. RAG as a librarian. You chop up your help pages – articles, product specs, manuals – into chunks. Run each chunk through an embedding model. Vectors. No AI generation. No talking back. Just a silent index.

USER. Like a smarter search bar?

DEVELOPER. Exactly. You search for “lightweight running shoes.” It finds “super-light marathon trainers.” Same meaning, different words. But the user still gets a list of snippets. Scroll. Click. Read three paragraphs. Patch the answer together. It’s… polite. But dull.

USER. So where’s that useful?

DEVELOPER. “You might also like” recommendations. Back-office content scanning. Batch processing where no human waits for an answer. Not for real conversations.

(His tone quickens.)

DEVELOPER. Level Two. Now we add the voice. The AI.

(He snaps his fingers.)

DEVELOPER. User asks: “How do I reset my password?” The vector library hunts for relevant chunks. Then we stuff the question plus those chunks into a large language model – DeepSeek V4-Flash, GPT-4o-mini, whatever. And the model writes an answer. Complete. Natural. No fluff.

USER. So they don’t have to read ten pages.

DEVELOPER. They get one answer. And they can follow up: “What about the admin reset?” The model remembers the conversation. That’s the jump. From search to answer.

USER. How much does that cost?

DEVELOPER. Pennies. DeepSeek V4-Flash costs ~0.87 per million output tokens. A typical Q&A uses maybe 2000 tokens. That’s 0.0017 per query. For a thousand queries a day?
1.70.Runamediume‑commercesiteforamonth–maybe10.

USER. That’s nothing.

DEVELOPER. Exactly. That’s why Level Two is the sweet spot today. Perfect for internal knowledge bases, product help centers, e‑commerce Q&A, legal summaries, news article deep dives. Anywhere you want a guide, not a search box.

(He raises a finger.)

DEVELOPER. Level Three. CaaS. Conversation as a Service.

(His pace accelerates, words running together.)

DEVELOPER. Now you don’t just index documents. You index APIs. Function specs, endpoint descriptions, parameter lists, call examples. The user says: “Pause my subscription for next week.”

The vector store retrieves the relevant API: PUT /subscriptions/{id}/pause with start_date and duration. The model understands intent and outputs a structured call – JSON, a curl command, whatever your system expects. Then your backend either suggests the action (“Click to confirm”) or executes it directly, depending on safety.

USER. That’s dangerous.

DEVELOPER. It’s powerful. You design a confirmation step when needed. But the leap is clear: the user talks, the website acts. No menus. No forms. For SaaS, CRM, IoT dashboards – any place where talking is faster than clicking.

USER. Where do I start?

(He slows down, leans back.)

DEVELOPER. Start simple. Take your existing help docs. Your FAQ. Your user manual. Throw them into a vector database (Milvus Lite, Qdrant, Chroma). Write a tiny endpoint that: (1) embeds the user’s question, (2) retrieves top‑k chunks, (3) calls an LLM, (4) returns the answer. Add a chat bubble on every page. Two weeks. That’s your first step.

USER. And then?

DEVELOPER. Then you evolve. Add more documents. Tune chunk size. Switch from “retrieve then generate” to a proper RAG pipeline with reranking. And only after that – when you have real APIs – move to Level Three. From “tell me how” to “do it for me.”

(He smiles, pauses.)

DEVELOPER. One sentence: Don’t automate everything at once. Just put your existing help into RAG, add a layer of AI, and you get an assistant that answers “how do I use this feature?” instantly. Happier users. Fewer support tickets. Later, plug in the real APIs. That’s the path.

(Exit.)