CHOLA INTELLIGENCE(CI)

@John_Nikson on CHOLA INTELLIGENCE(CI) · 20 days ago

30m logged

In this update to CHOLA (CI), I focused on maximizing UI personalization and accessibility by engineering a fluid theme and language switching system. I built a dynamic glassmorphic interface that allows users to instantly toggle between an eye-strain-reducing Night mode and a clean Day theme, saving their preference natively to browser storage. Alongside the visual overhaul, I implemented a system-level dual-language pipeline that fluidly adapts formatting and input layouts between English and Tamil (தமிழ்), adjusting the underlying prompt behavior in real time so users can interact naturally in their preferred dialect.

Ship #1 Changes requested

@John_Nikson on CHOLA INTELLIGENCE(CI) · 21 days ago

🚀 What did I make?
I built CHOLA (CI)—an agile, multimodal AI workspace platform engineered to optimize local productivity and bypass the common bottlenecks of standard AI wrappers. It features a responsive glassmorphic UI with built-in Night/Day themes and dynamic language toggles supporting English and Tamil (தமிழ்).

🛠️ What was challenging?
Two major things:

The PDF "Jumbled Text" Bug: Standard LLMs frequently throw errors or read gibberish when you upload compressed PDFs. To fix this without paying for heavy cloud OCR servers, I injected PDF.js natively into the client header. Now, it unzips complex document text layers instantly inside the browser.

Context Amnesia: I hated that opening a new chat tab meant the AI forgot everything I was working on. I had to build a dual-layer memory system that logs project milestones to localStorage and dynamically serializes and injects that historical context into the model’s systemInstruction block on launch.

🏆 What are you proud of?
--The Plus (+) Icon Modality Router: Bypassing strict text-model limitations by embedding the FLUX Engine right into the input tray. Clicking the expand button instantly switches modes, rendering high-fidelity graphics with canvas-bypassing direct downloads built right into the output bubble.

--Zero-Server Architecture: Keeping the heavy lifting (document parsing and memory tracking) entirely client-side to keep latency incredibly low and API token budgets predictable.

🧪 What should people know so they can test your project?
-Test the PDF Engine: Drop a text-heavy PDF into the chat and watch it instantly extract and analyze the data right inside your browser window.

-Try Image Gen: Click the Plus (+) icon in the input tray to toggle the FLUX pipeline, drop a creative prompt, and test the direct download button on the generated image.

-Test the Memory: Tell the AI a specific fact about your current project, open a brand new chat thread, and see how it carries that context over using the remembrance system!

To Be Continued............
CHOLA v1.5 Coming Soon..

1 devlog
1h

Try project → See source code →

Open comments for this post

@John_Nikson on CHOLA INTELLIGENCE(CI) · 21 days ago

33m 14s logged

CHOLA (CI) — Devlog #01:
Zero-Server PDF Parsing & Cross-Chat Memory.,

Hey Hack Club! 👋 I’m John Nikson, and I’m building CHOLA (CI)—an agile, multimodal AI workspace optimized for local productivity.Here is what I just shipped:

📂Zero-Server PDF Extraction: Fixed the classic “jumbled text error” on PDF uploads. I injected PDF.js directly into the client header. Now, it dynamically unpacks document text layers right in the browser, completely bypassing slow, expensive cloud OCR pipelines.
🎨 Expandable Modality Menu: Integrated the FLUX Engine natively. Clicking the Plus (+) icon swaps modes instantly, routing image prompts to the rendering cluster with canvas-bypassing direct download tools built into the UI.
🧠 Cross-Chat Remembrance: Built a dual-layer memory system. High-level project milestones are logged to localStorage. When a new thread starts, CHOLA serializes and injects that historical context straight into the model’s systemInstruction block—remembering your goals across separate chats without bloating your token budget.
🌍 Localized UX: Wrapped it in a responsive glassmorphic UI with Night/Day themes and dynamic system toggles for English and Tamil (தமிழ்).

CHECK IT OUT ON: https://chola-intelligence.vercel.app

Built with Groq API, HTML5, and JavaScript. Let me know what you think! 👇