-
-
Saathi on Mobile lets users ask, reflect and receive guidance through a clean, focused interface.
-
spiritual guidance for modern life across desktop and mobile.
-
A responsive mobile experience designed for accessible spiritual guidance wherever users are.
-
Saathi GPT provides empathetic, structured and practical guidance through natural conversation.
-
Explore the Bhagavad Gita, Ramcharitmanas, Upanishads and Krishna stories in one experience.
-
Scriptures on Mobile makes spiritual texts, meanings and reflections easy to explore on any device.
-
Guidance in Action connects personal concerns with reflection, relevant teachings and clear next steps.
-
Story-Based Reflection transforms Krishna stories into practical lessons for modern emotions and decisions.
-
A structured scripture library with clear categories, summaries and simple navigation.
-
Visual storytelling brings Krishna’s teachings to life through illustrations, narratives and memorable lessons.
Inspiration
Many people, especially students and young adults, experience stress, loneliness, confusion, anger, fear, academic pressure, and uncertainty about important life decisions. Although spiritual texts such as the Bhagavad Gita contain valuable teachings about courage, duty, self-control, resilience, and clarity, these teachings can be difficult to discover and apply to modern situations.
Digital Krishna was inspired by the relationship between Krishna and Arjuna. Krishna did not simply give Arjuna an answer; he helped him understand his emotions, responsibilities, and choices with greater clarity.
We wanted to create a culturally grounded digital companion that makes this wisdom more accessible through empathetic conversation, verified scripture retrieval, practical reflection, multilingual interaction, and engaging stories.
Digital Krishna is not intended to replace professional medical, psychological, legal, or emergency support. It is designed as a spiritual reflection and educational platform.
What it does
Digital Krishna is an AI-powered spiritual guidance and reflective wellbeing platform that connects modern life challenges with relevant teachings from Indian scriptures.
A user can describe a situation in natural language, such as:
- “I am worried about my future.”
- “I am unable to control my anger.”
- “I feel disappointed after receiving low marks.”
- “I am confused about an important decision.”
- “I feel unmotivated and emotionally exhausted.”
The platform analyses the concern and provides a structured response containing:
- An empathetic acknowledgement of the user’s situation
- Relevant spiritual or philosophical guidance
- A verified scripture reference when available
- A simple explanation of how the teaching relates to the situation
- Practical reflection steps that can be followed in daily life
- Safety-oriented messaging for sensitive situations
Digital Krishna also includes:
- Saathi AI Companion: A conversational guidance interface
- Scripture Library: Bhagavad Gita, Ramcharitmanas, Upanishadic teachings, and curated stories
- Verse Experience: Sanskrit, Hindi, English, meaning, and practical teaching
- Krishna Stories: Stories connected with real-life emotions and values
- Animation and Narration: Visual storytelling designed to make teachings memorable
- Multilingual Interaction: Support for English, Hindi, and Hinglish experiences
- Voice Experience: Speech and narration features for accessibility
- Responsive Web Platform: Designed to work across desktop and mobile devices
The target users include students, young adults, families, spiritual learners, educators, and people seeking culturally familiar tools for reflection and personal growth.
How we built it
Digital Krishna was developed as a modular web platform with separate layers for user experience, guidance generation, scripture retrieval, response validation, safety, and media delivery.
Frontend
The user interface was developed with React and responsive web technologies. The platform contains dedicated experiences for AI guidance, scripture exploration, verse reading, stories, animations, reflections, and mobile access.
The interface uses reusable UI components, responsive layouts, clear navigation, and lightweight visual elements to provide a calm and accessible experience.
Guidance pipeline
The guidance system follows a controlled processing flow:
- The user submits a concern through text or voice.
- The request is classified into a relevant concern category.
- The system checks for sensitive or high-risk language.
- Relevant scripture content is retrieved from a curated knowledge source.
- The AI produces a structured guidance response.
- The response is validated for format, relevance, safety, and scripture integrity.
- The final result is shown with practical reflection steps.
Scripture passages are treated as protected source content. The AI is not allowed to freely rewrite a verse and present the rewritten text as an authentic quotation.
AI and knowledge architecture
The project includes a Qwen-based model workflow, structured training-data preparation, retrieval-augmented generation, schema validation, prompt-safety controls, and a private model-serving architecture.
A verified initial scripture dataset was prepared for retrieval and evaluation. The architecture can later be expanded to support a much larger reviewed scripture collection.
Codex with GPT-5.6 was used as an engineering partner to:
- Audit the existing project repository
- Understand the frontend and backend architecture
- Design the guidance pipeline
- Implement and debug application components
- Prepare data-validation and model-training workflows
- Create safety and scripture-integrity tests
- Debug Hindi and Hinglish edge cases
- Document architecture, deployment, limitations, and evaluation
Codex and GPT-5.6 assisted with development. They are not presented as the spiritual authority or as the final source of scripture.
Technology stack
The project uses technologies including:
- React and JavaScript
- Python
- Tailwind CSS and responsive CSS
- Radix UI and Lucide icons
- Zod, AJV, and backend contract validation
- PyTorch, Transformers, PEFT, and Datasets
- Qwen model workflows
- ChromaDB and LangChain-compatible retrieval
- JSON-backed application storage
- Docker and cloud deployment configurations
- FFmpeg for video and media preparation
Development during DSOC
Digital Krishna existed as an early product concept before DSOC. During the DSOC development period, the project was strengthened as a more responsible and technically structured intelligent system.
The DSOC work focuses on:
- Improving the scripture-grounded guidance architecture
- Making the AI pipeline more explainable
- Strengthening response validation
- Improving multilingual safety detection
- Separating verified scripture from generated interpretation
- Documenting the technical architecture
- Improving the responsive product experience
- Preparing the platform for evaluation and scalable expansion
This distinction is important because the project submission represents meaningful new technical development rather than claiming that every part of the platform was created from zero during DSOC.
Challenges we ran into
Preventing incorrect scripture citations
Generative AI can produce text that sounds convincing even when a quotation or reference is inaccurate. For a spiritually grounded platform, this would seriously affect user trust.
We addressed this by separating verified scripture content from generated interpretation. Retrieved verses remain controlled source records, while the model explains their relevance separately.
Balancing empathy with safety
The platform must provide compassionate responses without pretending to be a doctor, therapist, emergency service, or spiritual authority.
We introduced safety-oriented classification, response boundaries, sensitive-language checks, and messaging that encourages users to seek trusted real-world support when necessary.
Supporting English, Hindi, and Hinglish
Users may communicate in English, Hindi, Romanised Hindi, or a mixture of languages. Sensitive expressions can appear very different across these formats.
We refined classification and safety checks for multilingual and Hinglish inputs and tested the guidance pipeline with varied phrasing.
Creating useful rather than generic guidance
A response can sound positive but still fail to help the user. We therefore designed a structured format that combines empathy, context, scripture, explanation, reflection, and one or more practical steps.
Maintaining performance
The platform includes images, animations, narration, AI interaction, and scripture content. Heavy media can cause slow loading, particularly on mobile devices.
We worked on responsive layouts, media optimisation, lightweight effects, controlled asset loading, and clearer separation between core guidance features and optional media experiences.
Distinguishing development tools from the runtime
Codex and GPT-5.6 contributed to engineering, debugging, documentation, and testing. The project also contains a separate Qwen-based inference workflow.
Clearly communicating this distinction was important for transparency and technical accuracy.
Accomplishments that we're proud of
We are proud that Digital Krishna has grown beyond a basic chatbot concept into a broader spiritual learning and reflection platform.
Key accomplishments include:
- Building a working conversational guidance interface
- Connecting modern user concerns with curated spiritual teachings
- Creating a controlled scripture-retrieval workflow
- Separating authentic scripture from AI-generated explanation
- Designing structured and actionable guidance responses
- Adding multilingual English, Hindi, and Hinglish support
- Improving sensitive-language detection
- Creating scripture, story, animation, narration, and reflection experiences
- Preparing a Qwen-based model and fine-tuning workflow
- Implementing response validation and integrity testing
- Creating a responsive platform for desktop and mobile users
- Documenting the architecture, limitations, safety approach, and future roadmap
We are especially proud that the platform aims to preserve cultural context while using modern AI responsibly.
What we learned
We learned that building a meaningful AI platform requires much more than connecting a user interface to a language model.
Reliable guidance depends on:
- High-quality and reviewed source material
- Clear separation between facts, quotations, and generated interpretation
- Safety systems that operate before and after generation
- Multilingual evaluation
- Transparent model and tool disclosure
- Thoughtful user experience
- Continuous feedback from real users and subject-matter reviewers
We also learned that AI should not present itself as an unquestionable authority. It should help users reflect, understand context, explore trusted sources, and make thoughtful decisions.
The project also taught us the value of modular architecture. By separating retrieval, generation, validation, safety, media, and interface layers, we can improve one part of the platform without rebuilding the entire system.
What's next for Digital Krishna
The next phase of Digital Krishna will focus on accuracy, accessibility, evaluation, and responsible growth.
Our roadmap includes:
- Expanding the verified scripture knowledge base
- Adding expert review and source-audit workflows
- Displaying clearer chapter, verse, and source information
- Improving semantic retrieval and verse ranking
- Conducting larger multilingual evaluations
- Adding more Indian languages
- Improving voice input and natural narration
- Building consent-based personalised reflection journeys
- Adding journaling, goals, and daily practice features
- Improving accessibility for visually impaired users
- Creating dashboards for anonymous helpfulness and quality feedback
- Strengthening privacy and data-retention controls
- Optimising the platform for low-bandwidth devices
- Collaborating with educators, scholars, wellbeing professionals, and cultural experts
- Developing sustainable subscription and institutional-access models
Our long-term vision is to make ancient wisdom understandable and practically useful for modern life while preserving authenticity, user dignity, safety, and cultural responsibility.
Digital Krishna — Ancient Wisdom for Modern Challenges.
Built With
- ai
- amazon-web-services
- beanstalk
- chromadb
- css
- docker
- elastic
- face
- generation
- github
- hugging
- javascript
- langchain
- mlx
- python
- pytorch
- qwen
- radix
- react
- render
- retrieval-augmented
- tailwind
- transformers
- ui
Log in or sign up for Devpost to join the conversation.