Flazy API

The first progressive self-coding API server. A truly flashy way to create a backend API system, though it feels a bit like lazy coding

What am I doing?

I've seen so many AI-based software builders. They all start the same way: by creating a specification, even a PRD (Product Requirements Document), and then implementing the actual code. It takes a really long time to build, run, and test. Why not just create the blueprint, run the boilerplate, then implement it and test the features one by one while running the program? Is that technically possible?

Flazy is a framework for that dream. It prepares everything from specification/blueprint, code-generation, database, and API router, to self-coding, self-fixing, and testing - all progressively. At first, it only creates the blueprint, and the actual code is generated while the program is running.

Install

npm install -g @naiv/flazy-api

flazy-api must be installed globally since user project depends on Flazy node-modules

Usage

Before running the cli, you must expose your Open Router API Key on your terminal environment

export SK=sk-or-v1-42e4ca6....

Verify

echo $SK
sk-or-v1-42e4ca6....  # output

1. Prepare Model Data & API Specification

$ mkdir my-api        # create new project directory
$ cd my-api           # enter directory
$ flazy-api --chat    # begin requirements generation

Output

? Explain briefly your API program ‣ buatkan program backend untuk pos kasir
✔ Generating data & API specification

After completing this step, you will have a file called api-specification.naiv inside your my-api folder. This file contains the DSL for your data model and API specification for your backend API.

2. Create Empty Database

In this example I use NAIV Zero Services to create an empty database, you can use your own database read Custom DB Provider section how to do it.

• Open https://naiv.dev/zero, login/register your account
• Go to Database tab
• Add new database
• Copy dbkey 64-character string long, something like this: e4e4eed9a....

3. Run Program

$ flazy-api --run --dbkey <your-dbkey>

Output

DB Key Accepted!
db initialized
⚠️ implementation for api 'get /api1' not found, skip.
⚠️ implementation for api 'post /api2' not found, skip.
⚠️ implementation for api 'post /api3' not found, skip.
...
completed.

⚡️[server]: Server is running at http://localhost:9000

Congrats your API progam is now running!

For Real?

Just kidding... kind of! your API project is running, but the implementation is still empty. Your API will implement itself the first time you call it. You might notice a slightly longer response time on the first call, because it’s running a LLM-conversation to generate the API functions.

Full CLI Available Command Example

Build NAIV Specification

flazy-api --chat

Run API

# run api on current folder port 9000
flazy-api --dbkey a83cd12.. --run .

# run api on current folder port 9000
flazy-api --dbkey a83cd12.. --run

# run api on folder my-api port 9000
flazy-api --dbkey a83cd12.. --run my-api

# run api on current folder port 8080
flazy-api --dbkey a83cd12.. --run --port 8080

# run api on current folder port 8080 with OpenRouter model z-ai/glm-5
flazy-api --dbkey a83cd12.. --run --port 8080 --model z-ai/glm-5

# run api on current folder port 9000 with 
# dbkey provided by API 'http://localhost:3000'
flazy-api --dburl 'http://localhost:3000' --dbkey a83cd12.. --run

Custom DB Provider

Look at this code from exec.ts on this project, this is how Flazy fetch translate database key into database credentials:

axios.get(dburl ?? 'https://api-x.naiv.dev/db-credential', { headers: { uniqid: database_key } })

You must provide an API that handle GET /db-credential with headers uniqid: string required. Expected response type:

export interface DBCredential {
  db_type: string
  db_host: string
  db_name: string
  db_username: string
  db_password: string
  db_port: number
}

Limitation

• LLM provider on this prototype right now only support openrouter.ai, will add more other LLM provider soon. LLM model can be change via --model <model-slug> argument.
• Users project really depends on flazy-api dependency, it makes your project almost not supporting any third-party library except flazy-api dependency (bcrypt, jsonwebtoken, etc).
• Self-healing feature is not implemented yet, in theory this program should be able to fix itself when a 5xx error (server side error) occured, I still think how to handle it efficiently in real solution.
• Unsupported direct npx command, since npx binary program doesnt have its own node-modules exposed.
• There are still many limitations, as this project is still new. We really appreciate any helpful feedback!