🖥️ Creating your own HomeLAB for trading

mazibee wrote: Thu Apr 16, 2026 5:10 pm I removed the old gemma3 models (that were already installed) and downloaded gemma4:e2b on my AIO locally.

It takes some time for HomeLAB to load once I start and after that I can not see any difference.

Unfortunately when I bought Dell AIO , my intention was space saving, i didnt knew at that time I will need a lot of VRam in future

You can also install LM Studio on your machine and use its endpoint http://127.0.0.1:1234 instead of Ollama and use gemma-4-e2b-it-uncensored 3.19 GB for example.

I switched to that Cloud Edition myself

HomeLAB v4.0 – Cloud Edition for windows11, Intel, Nvidia and 22" screen.

Log order reversed, meaning the newest log is now at the top, because on a small screen (22"), you don't need to scroll as much.

I'll see how this "new feature" works, I just did it. It needs auto-analyze to be on.

Walkthrough: Signal Alert Indicator
I have added a persistent Signal Alert indicator to the header of the HomeLAB dashboard. This element provides an immediate visual confirmation of the AI's latest trading recommendation.

Key Features
Header Integration: The indicator is prominently placed in the top header, making it visible at all times.
Dynamic State Management:
NO SIGNAL: The initial state before any analysis is performed.
SIGNAL: BUY: Displays in Green when a Buy action is identified.
SIGNAL: SELL: Displays in Red when a Sell action is identified.
STAY_CASH: Displays in Gray when the AI explicitly advises staying out of the market.
Automated Updates: The indicator updates automatically after every manual analysis or auto-analysis cycle.
Implementation Details
GUI Header Update
Created a new label self.signal_alert_label and adjusted the grid layout to accommodate it without causing UI jitter.

Intelligence Logic
Implemented _update_signal_alert which uses regex pattern matching to extract the exact recommendation from the AI's complex reasoning text.

TIP

This tracker works even during "THE_VOID" bypass cycles, ensuring you always know the current system status even when the AI skips a cloud call to save tokens.

Pelle, who wrote that he doesn't know anything about coding, really shines in this topic

Vetis wrote: Sat Apr 18, 2026 4:51 am Pelle, who wrote that he doesn't know anything about coding, really shines in this topic

As AI is getting better you can write code and do all sorts of things that you would otherwise be just scratching your head. The key to using AI is ask a better question you will get a better answer. Here a video that explains why Deepseek is so good.

First off all, thanks to @Xard for sharing his HomeLAB_v4.0 and to @Pelle for sharing an updated version of his HomeLAB.

The XU EFFECT and the Hebbian concepts have genuinely changed how I look at market structure.

I managed to get the HomeLAB running and will try on BTCUSD during the weekend. I have tried to add below:

Added Shortcut Keys (F1-F4) for the Log Filter .
Added Shortcut Keys (F5-F8) for the Strategy, Analyst, Local, Board.
I’ve learned a lot from this thread about AI, Python and MT5

I’m posting because I think I may be misunderstanding something basic, and I’d really appreciate a bit of clarification.

I’ve been studying the AI.md using different LLMs and working through the 24 rules , mainly testing on XAUUSD.

My goal has been to understand what the homelab is displaying in the logs when Analyzer is ON.

As I am mentioend in my earlier posts that most of the time, the Hebbian value seems to sit around 0.5000 (roughly 0.4995–0.5005). This looks like the market’s normal state.

At Hebbian = 0.50000 (which is 95%+ of market time):
RULE #0 makes BUY impossible
RULE #4/#5 makes SELL impossible
RULE #1 never triggers
RULE #7 never triggers
RULE #16 never aligns
RULE #17 never reaches threshold

BUT THE AI WILL NEVER SAY "BUY" OR "SELL", IT WILL ONLY SAY "WAIT" - STAY IN CASH.

Since Hebbian sits near 0.5000 most of the time, the HomeLAB seems to spend most of its time telling not to trade.

Is the framework intentionally designed to act only during extreme conditions, when Hebbian moves into imbalance or collapse zones , rather than during normal equilibrium?

If that’s the idea, it actually makes a lot of sense. Most traders probably lose by trading too often, so a system that waits for true exhaustion or accumulation would be very logical.

But if I’m misunderstanding and trades should appear around 0.5000 conditions, then I know I’m missing something important.

What I’m trying to understand:

a) When Hebbian ≈ 0.5000 — should I be trading or simply waiting?
b) During long Cash/Wait periods, what should my focus be? Monitoring divergence, waiting for collapse, something else?
c) Is there a rule or concept I should revisit that ties normal conditions together?
d) Do you personally treat normal vs extreme conditions differently, or does the same framework handle both?
e) AI.md rules need to be updated for the taking into account HEBBIAN staing around 0.5000?

I am just trying to learn how to interpret HomeLAB logs/ Output correctly. Any guidance would be greatly appreciated.

Thanks again for everything you have already shared and contributed to the community.

Best
Mazi

XARD: Hi Mazi, I like what you have done with the HomeLAB -- Like you said you have learned a lot and that is great -- let me explain myself in the next post.

The Evolution: From HomeLAB v4 to the v5 Sovereign Stack
Best,
Xard777

HomeLAB v4 was never about the profit on the screen; it was about solving the fundamental "Mayday" issues of building a self-sufficient trading ecosystem. It was the experimental phase—the bridge between the old "Laboratory" and the new "Sovereign Sanctuary."
The v4 Breakthroughs: Building the Foundation

The EFFECT Indicator: A mechanical masterpiece. By moving calculations and data-clipping to the indicator level before sending JSON payloads to Common/Files, we eliminated the lag and high-latency "noise" of the legacy setup.

The Cloud Shift: Moving 100% to the Cloud was our first move toward decentralization. It killed the local hardware bottlenecks and gave the system "Infinite Scale." and a level playing field for every user that used it.

The MT5 Bridge: We gave the AI "Eyes" and, in return, gave the human trader "Enhanced Analysis." This created a partnership in sync—a fusion of machine speed and human market awareness.

SQLite to QuestDB: Experimenting with local DBs paved the way for QuestDB, turning raw data into actionable intelligence.

Introducing HomeLAB v5: The Docker-Stack Sovereign

The lessons of v4 have culminated in v5, a system designed for resilience, modularity, and total market dominance.

The Containerized Fortress: HomeLAB v5 now runs on a Docker Stack with nine independent containers. This isn't just a setup; it’s a "Sovereign Dept." system. If one department goes down, the rest of the stack continues to trade. No single point of failure.

QuestDB + DS4 Logic: We’ve combined massive data input with DS4 reasoning. The AI wont just see the data; it will understand the intent. The analysis will evolve from "Signals" to "Strategies."

Synchronized Deployment: The new v5 Prompts enable a dual-threat approach. Whether you are hunting for M6 Snipers or managing Intra-day Sync trades, the system will adapt its logic to the timeframe in real-time and provide BUY/SELL and Stay Cash ANALYSIS..

The Journey & The Community

v4 was a learning and growing experience. It was about refining the "Handshake" between the code and the trader. To everyone who provided feedback and contributed along the way: The progress has been awesome.

HomeLAB v5 is in the making.... Once complete I will share the file (inc source)
xard@xard-System-Product-Name:/mnt/storage/VAULT-AI/HomeLAB_v5$ tree
.
├── backup
│   ├── daily
│   │   ├── backup_results_20260408.json
│   │   ├── methodology_wire_20260408_212441.tar.gz
│   │   ├── methodology_wire_20260408_213306.tar.gz
│   │   ├── postgres_20260408_212439.sql
│   │   └── postgres_20260408_213304.sql
│   ├── monthly
│   └── weekly
├── config
│   ├── backup-schedule.json
│   ├── monitor.json
│   ├── nginx.conf
│   ├── postgres-init.sql
│   └── redis.conf
├── data
│   ├── ai_models
│   ├── backups
│   ├── market_data
│   └── methodology_wire
│   ├── archive
│   ├── processed
│   └── raw
├── depts
│   ├── Analyst
│   │   ├── config
│   │   ├── outputs
│   │   ├── queries
│   │   └── scripts
│   ├── Board
│   │   ├── config
│   │   ├── outputs
│   │   ├── queries
│   │   └── scripts
│   ├── Local
│   │   ├── config
│   │   ├── outputs
│   │   ├── queries
│   │   └── scripts
│   └── Strategy
│   ├── config
│   ├── outputs
│   ├── queries
│   └── scripts
├── display
│   ├── assets
│   ├── config
│   ├── create_desktop_shortcut.sh
│   ├── homelab_background.old.py
│   ├── homelab_background.py
│   ├── homelab_v5.old.py
│   ├── homelab_v5.py
│   ├── src
│   ├── start_background.sh
│   └── themes
├── docker
│   ├── analysis-department
│   │   ├── analysis-route.conf
│   │   ├── analysis_service.py
│   │   ├── complete-nginx.conf
│   │   ├── current-nginx.conf
│   │   ├── Dockerfile
│   │   ├── new-nginx.conf
│   │   ├── requirements.txt
│   │   └── simple-nginx.conf
│   ├── backup
│   │   ├── backup_service.py
│   │   ├── Dockerfile
│   │   └── requirements.txt
│   ├── backup-fix.py
│   ├── backup-service
│   │   ├── backup_service.py
│   │   └── Dockerfile
│   ├── build_and_test.sh
│   ├── complete_restart.sh
│   ├── data-department
│   │   ├── app.py
│   │   ├── Dockerfile
│   │   └── requirements.txt
│   ├── docker-compose-postgres18.yml
│   ├── docker-compose.yml
│   ├── docker-compose.yml.backup
│   ├── docker-compose.yml.backup.1775686109
│   ├── docker-compose.yml.backup.1775686176
│   ├── docker-compose.yml.backup.1775686548
│   ├── docker-compose.yml.backup.1775687377
│   ├── execution-department
│   │   ├── Dockerfile
│   │   ├── execution.py
│   │   └── requirements.txt
│   ├── fix_docker_compose.sh
│   ├── fix_monitor_socket.sh
│   ├── init.sql
│   ├── monitor
│   │   ├── Dockerfile
│   │   ├── monitor.py
│   │   └── requirements.txt
│   ├── monitor-fix.py
│   ├── monitor_patch.yml
│   ├── monitor.py
│   ├── nginx
│   │   ├── html
│   │   ├── nginx.conf
│   │   └── nginx.conf.incomplete
│   ├── nginx-config
│   │   ├── nginx.conf
│   │   └── nginx.conf.backup
│   ├── postgres
│   ├── redis
│   ├── requirements-fix.txt
│   ├── strategy
│   │   ├── Dockerfile
│   │   ├── requirements.txt
│   │   └── strategy_service.py
│   ├── test_all_services.sh
│   ├── test_fixed.sh
│   └── verify_setup.sh
├── logs
│   ├── depts
│   ├── display
│   ├── docker
│   ├── nginx
│   │   ├── access.log
│   │   └── error.log
│   └── system
└── recovery
├── scripts
└── verify

62 directories, 67 files
xard@xard-System-Product-Name:/mnt/storage/VAULT-AI/HomeLAB_v5$

However before HomeLAB v5 rolls out a v4.5 will be created to follow 3 Markets of users own choosing (default being Bitcoin, Gold & Dow).

xard777 wrote: Sun Apr 19, 2026 8:36 am The Evolution: From HomeLAB v4 to the v5 Sovereign Stack
Best,
Xard777

HomeLAB v4 was never about the profit on the screen; it was about solving the fundamental "Mayday" issues of building a self-sufficient trading ecosystem. It was the experimental phase—the bridge between the old "Laboratory" and the new "Sovereign Sanctuary."
The v4 Breakthroughs: Building the Foundation

62 directories, 67 files
xard@xard-System-Product-Name:/mnt/storage/VAULT-AI/HomeLAB_v5$

However before HomeLAB v5 rolls out a v4.5 will be created to follow 3 Markets of users own choosing (default being Bitcoin, Gold & Dow).

@Xard, Thank you very much for sharing this update and for continuing to guide the community.

Your explanation really helped me put things into perspective. I realise now that I was probably trying to understand HomeLAB v4 too literally, while it was actually part of a much bigger journey ahead and learning process.

I just want to say thank you personally. As someone with no prior coding or programming experience, everything I have learned so far has been possible only because of what you have selflessly shared with the community. It genuinely means a lot to me.

Very grateful to be learning along the way and anxiously looking forward to the HomeLAB v4.5 / HomeLAB v5 journey ahead.

Stay Blessed always.
Mazi

Looking forward to the next version when it is ready. Meanwhile when I installed V4 Soveriegn Edition there were many errors in the CLI, I don't know why that was the case but I have been working through them and all seems to be well now if anyone wants it.

TIP:If Antigravity says you have to wait to use it again just switch it to another Google account in the top right and away you go again.

Switching to cloud based Deepseek is brilliant. The cost of running Deepseek is tiny, I commented previously about costs to run AI locally, you have upfront hardware cost for a high spec GPU and ongoing electricity run it. Been using Deepseek on and off for a couple days now and I am sure for most it would cost more to power a GPU running local than what Deepseek charges.
XARD: WOW! $20... Is that a five year plan lol

Updated CodeSage v1.2.0
Feel free to post improvements.
Enjoy!
Best,
Xard777


It looks like you’ve reached a solid milestone with CodeSage v1.2.0. At 378 lines, the script is lean but pack-heavy with "under-the-hood" improvements.

Based on the code you shared, here is a breakdown of why this specific version is so much more robust than its predecessors and what that line count represents:
1. Robust Engine Room (The Logic)

While the UI remains familiar and compact, the logic has been "hardened" to handle real-world failures:

Path Resilience: The dynamic path configuration now correctly identifies if it's running from a /src subfolder or the project root.

Safe Cleanup: Unlike previous versions that might have accidentally archived the script itself, the StorageManager.cleanup() now specifically ignores the running file.

Defensive API Client: The DeepSeekAPI class now includes resp.raise_for_status() and explicit handling for requests.exceptions.Timeout, ensuring the app doesn't crash if the server is down or the network hiccups.

2. Thread-Safe Architecture

One of the main reasons for the line increase is the addition of thread safety:

RLock & Lock: The use of threading.RLock() for the history and threading.Lock() for the API key ensures that background processes don't collide with UI updates.

Anti-Spam UI Locking: The _is_processing flag prevents multiple threads from spawning if a user clicks "Send" repeatedly, which protects both your local resources and your API credits.

3. Cross-Platform "Blackboard" Management

The open_bb function now intelligently detects the operating system (Linux, macOS, or Windows) to use the correct system command (xdg-open, open, or os.startfile) for opening your Markdown files.
4. Performance & Memory Management

History Trimming: The script now automatically trims the conversation history to the last 20 pairs (40 messages). This prevents the "token bloat" that can slow down responses and increase costs over long sessions.

Hot-Reloading: The .env file is re-read on every query, allowing you to swap API keys or settings without restarting the application.

UI Summary at a Glance
Feature Implementation
Status Bar Centered using place(relx=0.5) to align with the OS window title.
Shortcuts Added Ctrl+N (New), Ctrl+L (Load), and Ctrl+S (Save).
Project Persistence .sage files now include the full log, history, and blackboard state.

You've essentially moved from a "script" to a "Platform". Since you mentioned this is for the community to make their mark, this 378-line foundation is exactly what they need—stable enough to use daily, but readable enough to modify.

I have simply zipped up the CodeSage folder --
check the readme file inside on getting started or ask Antigravity
When running a project you can click on Blackboard to see what the AI is thinking and you can add/delete as you see fit then save the Project and recall it whenever suits you.

To run the application:
bash

# Activate virtual environment
source venv/bin/activate

# Install/update requirements if needed
pip install -r requirements.txt

# Run CodeSage
python src/codesage_v1.2.0.py or startLAB.sh

Some will add their mark by updating CodeSage, others will leave their mark from using CodeSage -- this is what is great about these projects in 2026 -- the next iteration of CodeSage will no doubt have some form of OpenClaw running alongside DS4