About Geetanjali

Why we built this and how we think about it.

Vision

Leaders face ethical dilemmas with no clear answers. Traditional advice is generic or disconnected from enduring wisdom. Geetanjali bridges this gap by grounding AI-powered guidance in the Bhagavad Geeta.

What we aim to be:

A trusted resource for ethical decision-making
Accessible to anyone, regardless of technical or spiritual background
Honest about limitations—AI can inform, not decide

What we won’t do:

Replace human judgment or professional advice
Make definitive moral pronouncements
Collect or sell user data

Design Principles

Principle	Implementation
Grounded, not hallucinated	Every recommendation cites actual verses via RAG
Local-first	Runs entirely on your machine with Ollama
Privacy by default	No account required; anonymous usage supported
Open source	MIT licensed; inspect, modify, self-host
Honest uncertainty	Low-confidence responses flagged for review

FAQ

General

Who is this for? Anyone facing ethical decisions—managers, founders, professionals, students. You don’t need prior knowledge of the Bhagavad Geeta.

Is this religious instruction? No. We treat the Geeta as philosophical literature offering ethical frameworks. The focus is practical wisdom, not doctrine.

How is this different from ChatGPT? Geetanjali is specialized. Instead of general knowledge, it retrieves specific verses relevant to your dilemma and cites them. No hallucinated quotes.

Privacy

What data do you collect?

Anonymous users: Cases stored locally in browser, no server-side tracking
Registered users: Email, cases, and preferences synced to server
No third-party analytics or advertising trackers

Can I delete my data? Yes. Account deletion removes all associated cases, outputs, and preferences. For anonymous usage, clear browser storage.

Where does processing happen?

Self-hosted: Everything runs on your machine
Cloud: LLM calls go to Ollama (local) or Anthropic (API); case data stays in your PostgreSQL instance

Usage

Can I use this for my team/organization? Yes. Self-host for full control, or use the public instance for individual use.

What languages are supported? English. Sanskrit verses include transliteration and translation.

Are the recommendations prescriptive? No. We present options with tradeoffs. The choice remains yours.

What if the guidance seems wrong? Use the feedback mechanism to flag unhelpful responses. Low-confidence outputs are automatically flagged for review.

Technical

Why Ollama instead of a cloud LLM? Privacy, cost, and latency. Local inference keeps your ethical dilemmas on your machine.

Can I use Anthropic Claude instead? Yes. Set LLM_PROVIDER=anthropic and provide your API key. Ollama remains the fallback.

How accurate is the verse retrieval? We use semantic search (sentence-transformers) with confidence scoring. Results below 0.7 confidence are flagged.

What’s the uptime guarantee? None for the public instance. Self-host for reliability requirements.

Accessibility

We design for everyone. Accessibility is a feature, not an afterthought.

Current Status

Feature	Status
Keyboard navigation	Full support across all interactive elements
Screen reader compatibility	ARIA labels, landmarks, live regions
Color contrast	Designed for WCAG 2.1 AA (4.5:1 minimum)
Focus indicators	Visible focus rings on all interactive elements
Reduced motion	Respects `prefers-reduced-motion`
Text scaling	Supports browser zoom and responsive sizing
Dark mode	Full theme support with system preference detection
Skip link	Keyboard users can skip to main content
Focus trapping	Modal dialogs trap focus appropriately

Keyboard Shortcuts

Key	Action
`Tab`	Navigate forward through interactive elements
`Shift+Tab`	Navigate backward
`Enter/Space`	Activate buttons and links
`Escape`	Close modals and dropdowns
`←/→`	Navigate between verses in Reading Mode
`J/K`	Vim-style navigation (next/previous verse)
`Space`	Toggle translation in Reading Mode

Known Limitations

Verse images: Generated share images have alt text, but embedded Sanskrit may not be fully accessible
PDF export: Not tagged for accessibility; use Markdown export for screen readers
Mobile gestures: Swipe navigation in reading mode may conflict with screen reader gestures

Reporting Issues

If you encounter accessibility barriers:

Open an issue with label accessibility
Include device, browser, and assistive technology used
Describe expected vs. actual behavior

Contact

Issues: GitHub Issues
Security: See Security Policy
General: Contact form at /contact on the running application