Gita AI - An AI-powered workspace for authentic Gita insights.

Three months post-launch, Gita AI V2 evolves from a simple query engine into an elite, stateful spiritual workspace balancing engineering with scriptural reverence. This release introduces context-hydration memory middleware, native TTS auditory sadhana playback, a luxury split-pane 18-chapter codex, snapshot "Wisdom Stories," and an optimized Neon PostgreSQL journaling ecosystem. Ancient lineage meets high-performance modern full-stack architecture. Om Tat Sat. ๐Ÿ™๐Ÿฆš

Add a comment

Replies

Best
Hey Product Hunt community! ๐Ÿ‘‹ I'm Sameer, the creator behind Gita AI V2. After seeing how organically my previous technical project, OrchestraML, was embraced by the community, I wanted to build something deeply personal that honors spiritual lineage while pushing the absolute limits of stateful full-stack engineering. Version 1 established a secure, foundational framework for scriptural RAG-based search. However, when dealing with a text as profound as the Srimad Bhagavad Gita, structural precision must always override AI hallucination. A single misinterpreted verse fundamentally alters the weight of the guidance. With Version 2, the goal was to transition the application from a basic, stateless query-and-response loop into an elite, highly stateful spiritual workspace. Here is how the 50-50 balance of spiritual intent and core engineering looks under the hood: ๐Ÿง  Stateful Conversation Memory: I engineered a custom context-hydration middleware layer that tracks session query parentage across 4-5 continuous turns, allowing for deep, flowing spiritual exploration without premature cache dropouts. ๐ŸŽ™๏ธ Auditory Sadhana Playback: Honoring the oral tradition of the text, V2 integrates a native Text-to-Speech (TTS) rendering pipeline directly into the chat interface and library panels for original shlokas and AI context synthesis. ๐Ÿ“– Advanced Cloud Journaling: Backed by an optimized Neon PostgreSQL database architecture, the updated system uses real-time text-search indexing and multi-variant filtering parameters to keep personal reflections structured and retrievable. โณ UI/UX & Latency Masking: We completely overhauled the user experience with an interactive split-pane 18-chapter codex, premium typography pairing (Newsreader + Lora), custom monthly recap "Wisdom Stories," and structural skeleton loaders to eliminate waiting anxiety during deep vector generation across Qdrant DB. The entire ecosystem was built using React, FastAPI, Neon PostgreSQL, and Qdrant vector spaces. Developing this update forced me to grow significantly in managing strict production branch pipelines, schema migrations, and secure authentication tracking via Clerk. I would highly value your architectural critiques, database suggestions, and overall product feedback. Welcome to the Sanctum! Om Tat Sat. ๐Ÿ™๐Ÿฆš

How does the context-hydration memory actually work across sessions, is it stored on my device or on your servers, and can I export or wipe the journal data whenever I want?

ย Thanks for the excellent architectural and privacy question!

To give you the exact breakdown: The active conversation context is hydrated and processed server-side through our FastAPI backend and securely mapped to your unique session token via Clerk authentication. Your persistent journal repository runs on a relational Neon PostgreSQL database instance to ensure your data stays synchronized smoothly across different devices.

Regarding data sovereignty: User autonomy is incredibly important when building a personal digital sanctuary. In our new V2 Settings control panel, I explicitly implemented native 'Export Journal Data' and 'Clear Chat History' mechanics. You can instantly download a complete backup of your reflections or completely wipe your data from our servers whenever you choose. Let me know if you get a chance to test it out!

curious how the memory middleware handles contradictions across sessions, like if i ask about karma in week one and get a different framing in week three, does it flag the shift or just keep building on whatever i said last

ย that is a fantastic question regarding conversational depth and context management!

The memory engine is mapped directly to your active session workspace, which means it remains intact regardless of the time elapsed-whether you return to the dialogue in 5 minutes or 3 weeks.

When it comes to shifting topics or exploring a concept like Karma from a different perspective over time, the system is designed to be adaptive rather than rigidly strict. It evaluates the context of your new input against the broader active thread state. It wonโ€™t flag a contradiction or force you into a loop; instead, it gracefully refines the conversational focus based on your latest intent while keeping the historical continuity alive.

This allows your spiritual exploration to evolve naturally as your own thoughts change over time. I'd love to hear how the conversational flow feels on your end as you explore different chapters!

๐Ÿ’ก Bright idea

The split-pane codex sounds beautiful for cross-referencing verses. One thing that would really deepen my practice is a daily shloka delivery that ties into the memory layer, so Gita AI surfaces a relevant verse based on whatever I've been journaling about that week, rather than a random rotation. Would make the spiritual side feel more personalized and alive.

this is an absolute gold-tier feature recommendation! Thank you so much for this feedback.

In this V2 release, our newly integrated Daily Wisdom Card operates on a generalized baseline. However, your idea of running a semantic aggregation over the user's past week of vector logs in their Personal Journal to dynamically surface an aligned daily shloka is brilliant. It completely shifts the interface from an active query workflow into a passive, deeply personalized meditative experience.

I am instantly writing this semantic aggregation algorithm into our V3 development roadmap. Thank you for helping shape the future architecture of the Sanctum!

ย did you get a chance to sign in and test out the live workspace firsthand, or are you reviewing based on the landing page and description?

how does the context-hydration memory actually decide what's worth keeping long-term versus discarding, and is there any way to manually curate what sticks around across sessions?

ย love this question! You've perfectly captured the exact balance we tried to strike between automation and human curation.

Here is how we divided those two boundaries under the hood: The short-term conversational context is automated entirely by the middleware's sliding window to keep the immediate conversational thread focused.

For long-term memory curation, we pass the control completely back to you. If the AI provides an insight or a specific shloka that resonates deeply, you can use the interactive 'Star Message' feature. This instantly flags the record and writes it directly to your permanent cloud relational ledger inside My Journal, ensuring it acts as a long-term baseline anchor for your spiritual history across all future sessions.

The split-pane codex layout with the 18-chapter structure feels genuinely considered, not just a wrapper around a chatbot. Nice to see the journaling backend getting real engineering love too.

Thank you so much, Fevzi! Hearing this means the world to me as a developer. โ€‹Avoiding the typical 'AI chatbot wrapper' trap was the absolute driving force behind the V2 architecture. I wanted to build an immersive digital sanctuary, which is why we completely overhauled the 18-chapter library into that custom split-pane workspace layout to balance typography and readability. โ€‹I'm especially glad you noticed the journaling system! Instead of just dumping text into local storage, I built the journal on a relational Neon PostgreSQL backend framework with optimized indexing for real-time multi-variant filtering and text searches. โ€‹Really appreciate you taking the time to notice those details. Enjoy exploring the Sanctum!