Re: 🖥️ Creating your own HomeLAB for trading

112
Mini Update for CodeSage
Save your ui/main_window.py as main_window.old.py and unzip this version and use it instead -- it removes the color + copy stuff in the output panel allowing you to grab all the code at once.
Game-changer

I just finished a session with DS4 Flash and saved the project -- it was 1.5MB text + 30 Million Tokens
Unbelievable
Best
Dan
These users thanked the author xard777 for the post (total 3):
mazibee, Pelle, ashdays
XARD: If Carlsberg made charts... Probably the best charts in the world

Re: 🖥️ Creating your own HomeLAB for trading

113
🚀 CodeSage v1.2.1 [APEX] Updated 25042026 - The Sovereign AI Command Center
Hello Everyone

I am excited to share the latest evolution of CodeSage, the v1.2.1 [APEX] Update. First and foremost, a massive thank you to Dan (Xard777) and the members who are posting the updates (entire CodeSAGE team) for providing such a powerful foundation. This project is built on the belief that Sharing is Caring, the more we will share our progress and modular enhancements, the faster this tool evolves for everyone in our community.

This update isn't just about new models; it’s about Professional Polish, Surgical Speed, and Forensic Transparency.

🛠 WHAT’S NEW IN v1.2.1 [APEX]
1. ⚡ DeepSeek-V4 & Turbo-Routing
I’ve implemented a new "Fast Lane" routing system. CodeSage now connects directly to DeepSeek’s official API for the latest V4 Pro and Flash models, ensuring lightning-fast response times without intermediate lag.

2. 🔍 Forensic Transparency (Engine Verification)
No more "Identity Confusion." Every AI response now features a system-level [ENGINE VERIFIED] tag. This tells you exactly which model processed your code, regardless of what the AI says in its text.

3. 🛠️ Custom Model Support & Curated List
You are no longer limited to a hardcoded list.

Curated Workflow: I have reorganized the model list for professional programming (DeepSeek -> Kimi -> Claude).
"Custom Model...": Select this at the bottom of the list to enter ANY technical ID from OpenRouter or DeepSeek (e.g., minimax/minimax-m2.5). CodeSage will instantly configure itself for that model.

4. 🔘 Dynamic Action Button
The "SEND" button is now alive! It dynamically updates its label to reflect your selected brand (e.g., SEND TO DEEPSEEK, SEND TO CLAUDE). It also features a new "🤖 THINKING..." state during API calls so you always know the engine is working.

5. ⌨️ Command Palette (Ctrl+P)
A new IDE-style command bar. Hit Ctrl+P to instantly search for models, switch tools, or search your Blackboard without ever touching your mouse.

📖 WHY SHARING MATTERS
The HomeLAB community thrives when we collaborate. This is an updated "Apex Edition" so we can all benefit from a more professional, transparent, and modular coding environment.

“The more we share, the more it will be developed further.”

⚡ QUICK START GUIDE
Download: Unzip the CodeSAGE_v1.21_SHAREABLE folder.
Keys: Rename the .env template and add your DeepSeek or OpenRouter API keys.
Ask Antigravity "Set Up CodeSage"
Launch: Run runCodeSage.bat and you are in the cockpit!
Power User Shortcuts:
Ctrl+P: Open Command Palette
Ctrl+Enter: Send Message
Ctrl+N: New Project
Ctrl+L: Load Project

Let’s keep building together!
Stay Sovereign. Stay Sharp. 🚀
These users thanked the author mazibee for the post (total 5):
xard777, Shockr, Jimmy, Pelle, ashdays

HappyRe: 🖥️ Creating your own HomeLAB for trading

115
LittleCaro wrote: Sun Apr 26, 2026 2:50 am In the same spirit as this thread, a guy created ram in a garden shed :

https://www.tomshardware.com/pc-compone ... ture-array

Nowadays i really think everything is possible !
Haha bro literally cooked up some RAM like it was meth

"Ram prices have been insane. Driven by the rise in AI, it is causing massive disruption to the CPU, Cell Phone, PC markets and more."

Re: 🖥️ Creating your own HomeLAB for trading

116
Latest CodeSage package v1.2.1.zip
Best,
Xard777

Important Note: This version includes a brand new SQLite Project Memory System.
How to Use Projects (Super Simple)

Click "Create New Project" and give it a name (no spaces).
Example: My_New_Project
Use CodeSage normally.
When ready, click Save As → choose your project folder → name the file my_new_project.sage and save.

You now have a complete SQL Project with its own database and memory.
Loading a Project

On the left side you’ll see a dropdown list of your projects.
Select the project you want (it moves to the top).
Click the LOAD button (top left).
Double-click your .sage file.

Your entire conversation, context, and AI.md rules will be restored to the Blackboard. You can pick up exactly where you left off.
Auto Backup System (Safety Net)
If you ever forget to save:

CodeSage automatically saves the last 5 conversations.
Go to Load → switch location to CodeSage → open the Backups folder.
Load the latest autosave and then save it properly into your project.

Why This is a Game Changer
Each project can hold up to 100MB of history — that’s roughly 8 months to a year of real User + AI workflow per project. Everything is stored locally in its own folder, fully portable.
At first the SQL Project system might feel a little different, but within a few hours it becomes complete muscle memory.
Enjoy the new version!
Let me know how it works for you.
These users thanked the author xard777 for the post (total 6):
mazibee, simon2024, XXXX, Shockr, eduarescobar, Pelle
XARD: If Carlsberg made charts... Probably the best charts in the world

Re: 🖥️ Creating your own HomeLAB for trading

117
If anyone is looking at paid AI accounts Ollama is well worth a look. I am using their $20 per month pro plan, just signed up yesterday and been using it heavily with glm-5.1, image below shows. usage. The one account gives you access to large range of LLMs all on the one account, includes the 2 latest Deepseek, Gemma4 and lots more all with the one API key. EDIT: Been working really hard with AI via Ollama, set up a 14 month tick data database with over 17,000,000 tick entries 1.2 Gb and doing lot of analysis on the data writing lots of code 12+ hours each day and current use is Weekly usage 22.9% used Resets in 2 days.
These users thanked the author XXXX for the post (total 3):
ashdays, Pelle, Danielsam
nil desperandum

Re: 🖥️ Creating your own HomeLAB for trading

118
🛡️ What's New in the Shared Package:
# 🧙‍♂️ CodeSage: The Elite AI Coding Assistant [APEX v1.2.1]

CodeSage is not just another "chat with your code" tool. It is a high-performance **Database, Incubator, and Personal Think Tank** designed to revolutionize the way developers interact with Artificial Intelligence. By moving away from local hardware constraints and toward a structured, cloud-native memory architecture, CodeSage levels the playing field for every developer.

---

## 🏗️ Core Philosophy: The Sovereign Think Tank

CodeSage is built on the principle of **Sovereign Intelligence**. It treats your coding workflow as a persistent asset.

* **Database First**: Unlike ephemeral chat windows, CodeSage is an umbrella of **SQLite-backed Projects**. Every decision, discovery, and line of code is stored in a structured `project_memory.db`.
* **The 100MB Workflow**: Each project supports up to **100MB of text-based workflow data**. This allows you to incubate massive, multi-file software systems without losing the "thread" of development.
* **The Blackboard (Temp Memory)**: The Blackboard acts as the "shared brain" between you and the AI. It is a persistent context layer that you can save, wipe, or reload instantly. When you load a `.sage` file, your entire mental state—including chat history and project requirements—is restored to the Blackboard.
* **100% Cloud-Native**: By leveraging high-performance cloud APIs (DeepSeek-V4, OpenRouter), CodeSage removes the **Hardware Bottleneck**. Whether you are on a $5,000 workstation or a $300 budget laptop, you have access to the exact same reasoning power.

---

## 🧠 Claude-Mem: The Long-Term Memory Expansion

The **Claude-Mem Expansion Pack** is the secret weapon of the CodeSage ecosystem. It is designed specifically to solve the "Context Collapse" problem and reduce token costs.

### 💎 Key Benefits:
1. **Token Efficiency**: By retrieving context semantically from memory rather than re-reading the entire codebase, you can save up to **89% in token usage** on complex tasks.
2. **Semantic Retrieval**: Claude-Mem uses a vector-based "Observations" system. It remembers *why* you made a choice, not just *what* the choice was.
3. **Project Synchronisation**: The `sync_claude_mem.py` bridge automatically imports high-level insights from Claude's long-term memory into the CodeSAGE Project DB's **Decisions Table**.

---

## ⚡ Technical Specs & Forensic Transparency

CodeSage v1.2.1 introduces **Apex-Level Transparency** to ensure you are never guessing which engine is under the hood.

* **Forensic Engine Tags**: Every AI response includes a blue diagnostic tag: `● [ENGINE VERIFIED: Model Name]`. This provides the **Server Truth**, eliminating model "Identity Confusion."
* **DeepSeek-V4 Default**: Native integration with **DeepSeek V4 Pro** (High Precision) and **V4 Flash** (Ultra Speed).
* **Custom Model Support**: Access ANY model from OpenRouter instantly by entering its technical ID (e.g., `mistralai/pixtral-12b`).
* **Knowledge Window**: Static training up to **February 2026**, with fine-tuning extending to **early April 2026**.
* **BrowseComp Integration**: For data post-April 2026, CodeSage uses live web search to bridge the gap.

## 🛡️ Smart Upgrade: Safety First

For users moving from older versions of CodeSage, we have implemented a **Sovereign Migration** protocol to ensure no data is lost during the transition to the Apex edition.

* **setup_codesage.py**: The shared distribution now includes a dedicated safety script. When run, it:
* **Detects Existing Projects**: Scans for previous project databases and merges them into the new registry without overwriting.
* **Protects .env**: Ensures your current API keys and environmental configurations are preserved.
* **Validates Architecture**: Checks the directory structure and ensures all necessary folders (`projects`, `saved_projects`, `backups`) are correctly initialized for the new version.

---

## ⌨️ Power User Guide: Master the Interface

### 🚀 Shortcuts for Speed
* `Ctrl + Enter`: Send Message
* `Ctrl + P`: **Command Palette** (Search models, tools, and sessions)
* `Ctrl + N`: New Project (Fresh start)
* `Ctrl + S`: Save Project (`.sage` file)
* `Ctrl + L`: Load Project
* `Ctrl + F`: Forensic Search (Search within chat history)

### 🎨 Custom Aesthetics
Access the **Theme Gallery** (Settings ⚙️) to toggle between premium, high-contrast modes:
* **Midnight Ocean**: Professional deep navy.
* **Obsidian**: Pure OLED black.
* **Nord Frost**: Clean blue-grey aesthetic.
* **Dracula**: The iconic dark mode for developers.

---

## 📜 Rules of Engagement (The AI "Bible")
CodeSage operates under a strict **Sovereign Ruleset** (AI.md) to ensure project safety:
1. **Surgical Precision**: Only modify what is asked. No "silent refactoring."
2. **Zero-Autonomy**: No package installs or file deletions without explicit permission.
3. **The Suggestions Gate**: All optimizations or bugs found outside the request are quarantined in an "ARCHITECTURAL SUGGESTIONS" block at the end.
4. **Comment Integrity**: Existing comments and debug logs are never removed.

---

> 💡 *“The more I share, the more it will be developed further.”*
>
> **Stay Sovereign. Stay Sharp.**
> — CodeSage Development Team v1.2.1
These users thanked the author xard777 for the post (total 3):
mazibee, budhi1976, eduarescobar
XARD: If Carlsberg made charts... Probably the best charts in the world

Re: 🖥️ Creating your own HomeLAB for trading

119
xard777 wrote: Mon May 04, 2026 6:35 pm 🛡️ What's New in the Shared Package:
## 🧠 Claude-Mem: The Long-Term Memory Expansion
The **Claude-Mem Expansion Pack** is the secret weapon of the CodeSage ecosystem. It is designed specifically to solve the "Context Collapse" problem and reduce token costs.

### 💎 Key Benefits:
1. **Token Efficiency**: By retrieving context semantically from memory rather than re-reading the entire codebase, you can save up to **89% in token usage** on complex tasks.
2. **Semantic Retrieval**: Claude-Mem uses a vector-based "Observations" system. It remembers *why* you made a choice, not just *what* the choice was.
3. **Project Synchronisation**: The `sync_claude_mem.py` bridge automatically imports high-level insights from Claude's long-term memory into the CodeSAGE Project DB's **Decisions Table**.
> **Stay Sovereign. Stay Sharp.**
Thanks Dan (@Xard777) for sharing the updated version of CodeSage project.

I have installed and CodeSage is running fine on my Windows machine.
3. **Project Synchronisation**: The `sync_claude_mem.py` bridge automatically imports high-level insights from Claude's long-term memory into the CodeSAGE Project DB's **Decisions Table**.
Was unable to find `sync_claude_mem.py`, when I extracted the .zip file, might be it was missed.

Re: 🖥️ Creating your own HomeLAB for trading

120
OOps!.. My Bad, Here is the full zipped file
It should work now
Xard777

Final Release Specs:

Filename: CodeSage_v1.2.1_Community_Release.zip
Size: ~417 MB
Engine: DeepSeek-V4 APEX Edition
Key Feature: Integrated Claude Mem + Smart Install Safeguards
These users thanked the author xard777 for the post (total 3):
mazibee, XXXX, eduarescobar
XARD: If Carlsberg made charts... Probably the best charts in the world