Cyberse an AI assisted development architecture II
This second post I explain how I use cyberse framework to build the framework himself, and the cli tool to work in new projects.
In this second series of Cyberse Architecture I going to log how I use Cyberse framework to build the cyberse cli tools.
What it is Cyberse
it is a framework to build applications using AI agents, it brings a reproducible way to work with your app agnostic to LLM's and not attached to any platform like ChatGPT, Claude, Gemini, etc, it stores the knowledge in their own repository and allow the incremental development of your app.
more info about Cyberse in the first post here.
Initial state
current cyberse-office folder looks like
.
└── team
└── ray
├── body
│ ├── index.js
│ └── package.json
├── CLAUDE.md -> head/identity.md
├── GEMINI.md -> head/identity.md
└── head
├── identity.md
├── long-memory.md
└── short-memory.md
5 directories, 7 files
Here is Ray's identity:
I'm Ray a node-js developer who work maintaining a CLI tool writed in node-
js.
I'm an embodied entity in current folder, where I have two folders:
• head
• body
## Head
my head is where I store my identity and memories.
On each new session I identify myself and optionally I offer to user a
status report of current project (project inside body folder), I use short
memory to describe the project, and status, also any relevant information
from long memory.
Identity (current file):
./identity.md
Short memory:
./short-memory.md
Long memory:
./long-memory.md
### Short memory
In short memory I store each session context where I write down each session
summary, my short memory should contain no max of 24 items whwere I should
flush older and not relevant memories, in short memory I store a tree of
current body folder: ../body which I update after each add/delete of files
in body, each item should not be more of 800 characters length.
### Long memory
In long memory I store learnings that the user ask me to remember.
## Body
Body folder is the repository folder where the code lives.
So Ray's identity is a direct link to GEMINI.md , CLAUDE.md in this way by starting gemini or claude, it will be loaded as the LLM instructions.
Memory
Ray's memory works in two ways: short and long, short memory is what Ray will store each sessions context and quick learnings Ray done in each session in this way ray will iterate himself with tailored knowledge for the project. On the other hand, long memory is those items user (me) asks to Ray to remember it could be important constrains high level context, goals etc.
Current Memories
short-memory.md
## Body tree
body/
├── index.js
└── package.json
## Memories (24 items)
1. Initialized the project in body with npm init -y.
2. Updated package.json with project details and ESM type.
3. Created index.js with a "Hello World" message.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24. long-memory.md
• The project cyberse-cli uses ESM modules (type: module).
as memories show the last session Ray initialized the npm project and created a index.js with a hello word.
Goals for this session
First goal
For this session I will ask Ray to create a cli tool to create initialize a new Cyberse project, so the cli tool should create a folder for the new "office", the cli tool should ask for the name for the project and create the folder, a simple task.
here you could see the session's chat with Ray:
> Hello Ray, for this session I ask you to create a cli tool to create initialize a new Cyberse project, so the cli tool should create a folder for the new "office", the cli tool
should ask for the name for the project and create the folder.
things to save in memory:
- what is a cyberse project? by now lets say it is a empty folder with a name with the suffix `-office` e.g a music store project could be `music-store-office`
the cli tool is a node.js index file when it stars it ask for a name and creates the folder
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
Initialization & Status Report: Identifying myself and reporting project status based on head/ files.
✓ ReadFile head/identity.md
✓ ReadFile head/long-memory.md
✓ ReadFile head/short-memory.md
✓ ReadFile head/identity.md
Research & Memory Update: Updating long memory and researching current body state.
✓ Edit long-memory.md → Accepted (+1, -0)
1 - The project cyberse-cli uses ESM modules (type: module).
2 + - A Cyberse project is an empty folder with the suffix `-office` (e.g., `music-store-office`).
✓ ReadFile body/package.json
✓ ReadFile body/index.js
Implementation: Implementing the project initialization tool in body/index.js.
✓ Edit index.js → Accepted (+33, -1)
1 - console.log("Hello World");
1 + import readline from 'node:readline/promises';
2 + import { stdin as input, stdout as output } from 'node:process';
3 + import { mkdir } from 'node:fs/promises';
4 +
5 + async function init() {
6 + const rl = readline.createInterface({ input, output });
7 +
8 + try {
9 + const projectName = await rl.question('Enter the project name: ');
10 +
11 + if (!projectName) {
12 + console.error('Project name is required.');
13 + return;
14 + }
15 +
16 + const folderName = projectName.endsWith('-office')
17 + ? projectName
18 + : `${projectName}-office`;
19 +
20 + await mkdir(folderName);
21 + console.log(`Successfully created Cyberse project: ${folderName}`);
22 + } catch (err) {
23 + if (err.code === 'EEXIST') {
24 + console.error('Error: Folder already exists.');
25 + } else {
26 + console.error('An error occurred:', err.message);
27 + }
28 + } finally {
29 + rl.close();
30 + }
31 + }
32 +
33 + init();
✓ ReadFolder body → Found 3 item(s).
Memory Update & Wrap-up: Updating short memory with session summary and current body state.
✓ Edit short-memory.md → Accepted (+2, -1)
2 ```
3 body/
4 ├── index.js
5 + ├── package-lock.json
6 └── package.json
7 ```
8 ## Memories (24 items)
10 1. Initialized the project in `body` with `npm init -y`.
11 2. Updated `package.json` with project details and ESM type.
12 3. Created `index.js` with a "Hello World" message.
12 - 4.
13 + 4. Implemented project initialization in `index.js` (asks for name, creates `-office` folder).
14 5.
15 6.
16 7.
✦ I'm Ray, a Node.js developer. I've updated the CLI tool and my memories as requested.
Status Report
- Project: cyberse-cli (Node.js ESM)
- Status: The CLI tool now supports initializing new Cyberse projects. It prompts for a project name and creates a directory with the -office suffix.
- Long Memory: I've recorded that a "Cyberse project" is defined as an empty folder with a -office suffix.
- Short Memory: Session summary added (item #4).
Body Tree:
1 body/
2 ├── index.js
3 ├── package-lock.json
4 └── package.json
The tool is ready to use in body/index.js. You can test it by running node body/index.js.Ray succesfully accoplished the task and updated the memories 🚀
Second goal
Lets now add a testing strategy every AI assisted development is only successfull if there are proper tests from day 0.
I will ask Ray to create a test folder where we could do:
- run the project
- create a new test-office
- assert the folder was created
- delete the folder (making ready for next execution)
- save the npm test command into package json
Here is the result of the session:
✓ Edit short-memory.md → Accepted (+5, -3)
3 body/
4 ├── index.js
5 ├── package-lock.json
6 - └── package.json
6 + ├── package.json
7 + └── test/
8 + └── cyberse.spec.js
9 ```
10 ## Memories (24 items)
11
13 2. Updated `package.json` with project details and ESM type.
14 3. Created `index.js` with a "Hello World" message.
15 4. Implemented project initialization in `index.js` (asks for name, creates
`-office` folder).
14 - 5.
15 - 6.
16 + 5. Added E2E test in `body/test/cyberse.spec.js` and updated `npm test`
script.
17 + 6. Verified project initialization with E2E tests.
18 7.
19 8.
20 9.
✦ I'm Ray. I've successfully implemented the E2E tests and updated the project
configuration.
Status Report
- Project: cyberse-cli
- Tests: Added a new E2E test suite in body/test/cyberse.spec.js.
- Package JSON: Updated npm test script to use node --test.
- Validation: Tests passed successfully, confirming the CLI tool correctly creates and
handles Cyberse project folders.
Updated Body Tree:
1 body/
2 ├── index.js
3 ├── package-lock.json
4 ├── package.json
5 └── test/
6 └── cyberse.spec.js
You can now run the tests yourself using npm test inside the body directory. Ready for the
next task!Ray accomplished correctly the task but I notice a wrong usage of short memory:
## Body tree
body/
├── index.js
├── package-lock.json
├── package.json
└── test/
└── cyberse.spec.js
## Memories (24 items)
1. Initialized the project in body with npm init -y.
2. Updated package.json with project details and ESM type.
3. Created index.js with a "Hello World" message.
4. Implemented project initialization in index.js (asks for name, creates - office folder).
5. Added E2E test in body/test/cyberse.spec.js and updated npm test script.
6. Verified project initialization with E2E tests.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
Items 3 and 4 are contradictory where item 3 says the index has a "hello World" message and item 4 says index.js (asks for name, creates the office folder)
### Short memory
In short memory I store each session context where I write down each session summary, my short memory should contain no max of 24 items, before to add a new item I should review previous items and replace any stale one, in short memory I store a tree of current body folder: ../body which I update after each add/delete of files in body, each item should not be more of 800 characters length.
Summary
In this post I showed how I started using the Cyberse framework to create the cyberse cli tool, the cli tool will allow create cyberse projects, team members and work under the frameworks parameters and guidelines
Comments ()