+* Arpit
+* Ashutosh Mani Shukla
---
+Quick Summary
-# 📖 Overview
+Helix is an AI-powered product support platform that transforms product documentation into an intelligent diagnostic assistant.
-Helix is an intelligent product support platform designed to help users diagnose, troubleshoot, and resolve product issues using official manufacturer documentation.
+Companies upload manuals, guides, videos, and support resources. Helix indexes this information using MOSS and enables users to troubleshoot issues through a technician-style diagnostic workflow rather than a traditional chatbot interaction.
-Instead of behaving like a traditional chatbot, Helix acts as a digital support engineer.
+Core Innovation
-It systematically:
+Instead of directly answering questions, Helix investigates problems by:
-* Understands symptoms
-* Retrieves relevant documentation
-* Asks follow-up questions
-* Eliminates unlikely causes
-* Suggests safe inspection steps
-* Diagnoses probable issues
-* Recommends solutions
-* Cites official documentation
+Understanding symptoms
+Retrieving relevant documentation
+Asking follow-up questions
+Eliminating unlikely causes
+Recommending corrective actions
+Providing source-backed explanations
+Built With
+Next.js 15
+TypeScript
+Prisma
+SQLite
+MOSS
+Groq (Llama 3.3 70B)
+NextAuth
+What Makes Helix Different?
-The result is a support experience that feels closer to speaking with a technician than searching through PDFs.
+Most support systems focus on answering questions.
----
+Helix focuses on diagnosing problems.
+
+Traditional Support Bot
+User Question
+ ↓
+Retrieve Documents
+ ↓
+Generate Answer
+Helix Workflow
+User Reports Symptoms
+ ↓
+Retrieve Documentation
+ ↓
+Analyze Context
+ ↓
+Ask Follow-Up Questions
+ ↓
+Eliminate Possibilities
+ ↓
+Perform Diagnostic Reasoning
+ ↓
+Recommend Corrective Actions
+ ↓
+Provide Citations
-# 🚨 Problem Statement
+This technician-style workflow helps users identify root causes instead of simply reading documentation.
-Every day users struggle with products such as:
+## Problem Statement
-* Routers
-* Printers
-* Air Conditioners
-* Washing Machines
-* Scooters
-* Water Purifiers
-* Consumer Electronics
+Users interact with hundreds of products every day including routers, printers, washing machines, air conditioners, scooters, water purifiers, and other consumer electronics.
-The information needed to solve their issues already exists.
+When these products fail, finding the correct solution becomes surprisingly difficult.
-The challenge is that it is scattered across:
+Although the required information already exists, it is often scattered across:
* Product manuals
-* Service guides
-* Knowledge bases
* Support websites
-* Videos
+* Knowledge bases
* Technical documentation
+* Service guides
+* Video tutorials
-Traditional support systems force users to search through hundreds of pages of documentation.
+Users frequently spend hours searching through documentation or contacting support for issues that could be resolved independently.
-Helix changes that.
+Current support systems generally suffer from three major limitations:
+
+### Documentation is difficult to navigate
+
+Manuals often contain hundreds of pages of information, making it difficult to locate the exact troubleshooting step required.
+
+### Search systems lack context
+
+Keyword-based searches return large amounts of information without understanding the actual problem being experienced by the user.
+
+### Chatbots answer instead of diagnosing
+
+Most AI assistants retrieve information and immediately provide answers without gathering enough information to determine the real cause of the issue.
---
-# 💡 Our Solution
+## Our Solution
-Helix converts manufacturer documentation into a searchable knowledge repository and combines it with AI-powered diagnostic reasoning.
+Helix converts product documentation into an intelligent support system.
-Unlike a standard RAG chatbot:
+Companies upload product resources such as manuals, technical documentation, videos, and support guides.
-```text
-Question
- ↓
-Retrieve Docs
- ↓
-Answer
-```
+The platform processes and indexes these resources using MOSS.
-Helix follows a diagnostic workflow:
+When a user reports an issue, Helix:
-```text
-Question
- ↓
-Retrieve Docs
- ↓
-Investigate
- ↓
-Ask Follow-Up Questions
- ↓
-Eliminate Possibilities
- ↓
-Diagnose
- ↓
-Recommend Solution
-```
+1. Retrieves relevant documentation.
+2. Understands the reported symptoms.
+3. Asks targeted follow-up questions.
+4. Eliminates unlikely causes.
+5. Suggests safe inspection steps.
+6. Determines the most probable root cause.
+7. Recommends corrective actions.
+8. Provides references to supporting documentation.
-This approach produces more accurate troubleshooting results and mimics the workflow of a real support technician.
+This creates a troubleshooting experience closer to speaking with a support engineer than searching through documentation.
---
-# ✨ Core Features
+# Features
-## 🏪 Product Marketplace
+## Product Marketplace
-Companies can:
+The marketplace acts as a centralized catalog where users can discover products and access support resources.
-* Register on the platform
-* Create products
-* Manage documentation
-* Upload support resources
+### Company Features
-Users can:
+* Create company profiles
+* Add products
+* Manage product information
+* Upload documentation
+* Update product resources
-* Browse manufacturers
-* Explore products
-* Search product catalogs
+### User Features
+
+* Browse products
+* Search products
+* View product information
* Access support resources
+* Start diagnostic sessions
+
+### Search Capabilities
+
+* Product search
+* Category filtering
+* Company filtering
+* Hybrid retrieval using MOSS
---
-## 📚 Knowledge Repository
+## Knowledge Repository
+
+Each product has a dedicated knowledge repository.
-Supports:
+Supported resource types include:
* PDF Manuals
-* Text Documents
+* Technical Documentation
* Product Guides
* Images
* Videos
* External Documentation Links
+* Text Documents
+
+### Processing Pipeline
-All content is automatically indexed for retrieval.
+```text
+Upload
+ ↓
+Parse
+ ↓
+Chunk
+ ↓
+Metadata Extraction
+ ↓
+MOSS Indexing
+ ↓
+Retrieval Ready
+```
+
+The repository becomes the source of truth for all diagnostic recommendations.
---
-## 🤖 AI Diagnostic Assistant
+## Diagnostic Assistant
-The heart of Helix.
+The diagnostic assistant is the core feature of Helix.
-Features:
+Unlike standard chatbots, the assistant follows a structured troubleshooting process.
-* Multi-turn troubleshooting
-* Diagnostic reasoning
-* Source citations
-* Context retention
-* Documentation-backed recommendations
-* Root cause analysis
+### Diagnostic Workflow
-Example:
+```text
+User Reports Problem
+ ↓
+Understand Symptoms
+ ↓
+Retrieve Documentation
+ ↓
+Identify Possible Causes
+ ↓
+Ask Follow-Up Questions
+ ↓
+Eliminate Possibilities
+ ↓
+Suggest Tests
+ ↓
+Evaluate Responses
+ ↓
+Diagnose Issue
+ ↓
+Recommend Solution
+```
+
+### Example
User:
```text
-My router is not connecting to the internet.
+My scooter horn is not working.
```
Assistant:
```text
-• Are the WAN lights blinking?
-• Did the issue start recently?
-• Have you restarted the router?
-• Are other devices affected?
+Does the headlight work normally?
+
+Is the horn completely silent or weak?
+
+Did the issue start suddenly?
+
+Has any electrical work been performed recently?
```
-The assistant narrows down causes until a probable diagnosis is reached.
+After collecting sufficient information, the assistant narrows down the issue and recommends the next steps.
---
-## 🌍 Multi-Language Support
+## Context-Aware Conversations
+
+Helix maintains troubleshooting context throughout the conversation.
-Users can interact in multiple languages while documentation remains indexed in English.
+The assistant remembers:
-Supported examples:
+* Previous questions
+* User responses
+* Suggested actions
+* Diagnostic progress
+
+This prevents users from repeatedly explaining the same issue.
+
+---
+
+## Documentation Citations
+
+Every recommendation is grounded in manufacturer-provided resources.
+
+Responses include:
+
+* Source document
+* Page number
+* Relevant section
+* Supporting references
+
+This improves transparency and trust.
+
+---
+
+## Session Memory
+
+MOSS sessions are used to maintain diagnostic context.
+
+Benefits:
+
+* Long conversations remain coherent
+* Previous troubleshooting steps are remembered
+* Follow-up questions become more accurate
+* Context is preserved throughout diagnosis
+
+---
+
+## Multi-Language Support
+
+Helix supports multilingual interactions.
+
+Examples:
* English
* Hindi
@@ -187,35 +299,129 @@ Supported examples:
* French
* German
----
+Users can communicate naturally while still benefiting from the same documentation retrieval system.
-## 📄 Citations & Handoff Briefs
+---
-Every recommendation includes references to source documents.
+## Product Ownership & Inventory
-When escalation is required, Helix generates:
+Users can maintain a personal inventory of products.
-* Issue Summary
-* Diagnostic History
-* Recommended Actions
-* Documentation References
+Features include:
-allowing support teams to continue seamlessly.
+* Product ownership tracking
+* Warranty monitoring
+* Maintenance schedules
+* Product history
---
-## 🧰 Product Ownership & Inventory
+## Maintenance Tracking
-Users can:
+Helix can support preventive maintenance workflows.
-* Track owned products
-* Monitor warranties
-* View maintenance schedules
-* Receive future recall alerts
+Examples:
+
+* Filter replacement reminders
+* Battery maintenance schedules
+* Routine servicing notifications
+* Equipment health tracking
---
-# 📸 Screenshots
+Implemented Features
+User Features
+Product search and discovery
+Product detail pages
+Documentation browsing
+Diagnostic assistant
+Multi-turn troubleshooting
+Context-aware conversations
+Source-backed recommendations
+Session history
+Company Features
+Company onboarding
+Product management
+Documentation uploads
+Resource management
+Product knowledge base creation
+AI Features
+Retrieval-Augmented Generation
+MOSS-powered retrieval
+Session memory
+Diagnostic reasoning
+Follow-up question generation
+Context retention
+Documentation grounding
+Knowledge Management
+PDF ingestion
+Text ingestion
+External documentation links
+Metadata-based retrieval
+Product-specific indexing
+User Journey
+Step 1
+
+User searches for a product.
+
+Step 2
+
+User opens the product page.
+
+Step 3
+
+User starts a diagnostic session.
+
+Step 4
+
+Helix retrieves relevant product documentation.
+
+Step 5
+
+Helix asks follow-up questions.
+
+Step 6
+
+User performs suggested checks.
+
+Step 7
+
+Helix narrows down possible causes.
+
+Step 8
+
+Helix identifies the most probable issue.
+
+Step 9
+
+Helix recommends corrective actions and cites documentation.
+
+Company Journey
+Step 1
+
+Company registers on the platform.
+
+Step 2
+
+Company creates product listings.
+
+Step 3
+
+Support resources are uploaded.
+
+Step 4
+
+Documentation is processed and indexed.
+
+Step 5
+
+Knowledge base becomes searchable.
+
+Step 6
+
+Users receive diagnostic assistance powered by company documentation.
+
+# Screenshots
## Landing Page
@@ -240,154 +446,240 @@ Users can:
---
+What Makes Helix Different?
+
+Most support systems focus on answering questions.
+
+Helix focuses on diagnosing problems.
+
+Traditional Support Bot
+User Question
+ ↓
+Retrieve Documents
+ ↓
+Generate Answer
+Helix Workflow
+User Reports Symptoms
+ ↓
+Retrieve Documentation
+ ↓
+Analyze Context
+ ↓
+Ask Follow-Up Questions
+ ↓
+Eliminate Possibilities
+ ↓
+Perform Diagnostic Reasoning
+ ↓
+Recommend Corrective Actions
+ ↓
+Provide Citations
+
+This technician-style workflow helps users identify root causes instead of simply reading documentation.
-# 🏗️ System Architecture
+# System Architecture
```text
-┌───────────────────────────────────────────────┐
-│ HELIX │
-└───────────────────────────────────────────────┘
-
- ┌─────────────┐
- │ Browser │
- │ Next.js │
- └──────┬──────┘
- │
- ▼
-
- ┌───────────────────────────┐
- │ API / Server Layer │
- │ Server Actions │
- └─────────────┬─────────────┘
- │
-
- ┌────────────────┼────────────────┐
- ▼ ▼ ▼
-
- ┌────────────┐ ┌────────────┐ ┌────────────┐
- │ Prisma │ │ MOSS │ │ Groq │
- │ SQLite DB │ │ Retrieval │ │ Llama 3.3 │
- └────────────┘ └────────────┘ └────────────┘
-
- │
- ▼
-
- ┌───────────────────────────────┐
- │ Diagnostic AI Technician │
- │ │
- │ • Retrieve Documentation │
- │ • Ask Follow-Up Questions │
- │ • Eliminate Possibilities │
- │ • Diagnose Issues │
- │ • Recommend Fixes │
- │ • Cite Sources │
- └───────────────────────────────┘
+┌───────────────────────────────────────────────────────────────┐
+│ HELIX │
+└───────────────────────────────────────────────────────────────┘
+
+ Browser
+ │
+ ▼
+
+ Next.js Frontend
+ │
+ ▼
+
+ Server Actions / APIs
+ │
+
+ ┌───────────────────┼───────────────────┐
+ │ │ │
+
+ ▼ ▼ ▼
+
+ Prisma Database MOSS Engine Groq LLM
+
+ │ │ │
+
+ └───────────────┬───┴───────────────┬───┘
+ │ │
+ ▼ ▼
+
+ Diagnostic Context Documentation
+ Retrieval Retrieval
+
+ │
+ ▼
+
+ Diagnostic Assistant
```
---
-# 🧠 MOSS Integration
+# MOSS Integration
-MOSS powers every retrieval operation inside Helix.
+MOSS is the foundation of the retrieval layer.
-## Product Catalog Search
+Rather than using MOSS as a simple document search engine, Helix uses it across multiple parts of the platform.
+
+---
+
+## Product Discovery
+
+Users searching for products interact with the Product Catalog Index.
```text
User Search
- ↓
-MOSS Product Catalog
- ↓
+ ↓
+Product Catalog Index
+ ↓
Relevant Products
```
---
-## Knowledge Base Retrieval
+## Product Knowledge Base
+
+Each product receives its own dedicated knowledge base.
```text
-User Query
- ↓
-Product Knowledge Index
- ↓
-Relevant Manual Sections
+Documentation
+ ↓
+Chunking
+ ↓
+Indexing
+ ↓
+Knowledge Base
```
---
## Session Memory
+Diagnostic conversations use MOSS sessions.
+
```text
Conversation
↓
-MOSS Session Context
+Session Storage
↓
-Previous Diagnostic Steps
+Context Retrieval
+ ↓
+Follow-Up Questions
```
---
-## Diagnostic Flow
+## Diagnostic Retrieval
```text
-User Reports Issue
- ↓
-Retrieve Relevant Docs
- ↓
+User Question
+ ↓
+Retrieve Relevant Documents
+ ↓
Retrieve Session Context
- ↓
-Generate Questions
- ↓
-Analyze Answers
- ↓
-Determine Root Cause
- ↓
-Recommend Solution
- ↓
-Provide Citations
+ ↓
+Build Diagnostic Prompt
+ ↓
+Generate Response
+ ↓
+Store Conversation
```
---
-# ⚙️ Tech Stack
+# Technical Implementation
## Frontend
* Next.js 15
* React
* TypeScript
-* CSS
+* Responsive Design
+* Server Components
+
+---
## Backend
+* Next.js API Routes
* Server Actions
-* API Routes
-* Node.js
+* Node.js Runtime
+
+---
+
+## Retrieval Layer
+
+* MOSS SDK
+* Product Catalog Index
+* Product Knowledge Base
+* Session Retrieval
+
+---
## Database
-* Prisma
+* Prisma ORM
* SQLite
-## Retrieval Layer
+---
-* MOSS
+## Authentication
+
+* NextAuth
+* Role Based Access Control
+
+Roles:
+
+* User
+* Company Admin
+
+---
## AI Layer
* Groq SDK
* Llama 3.3 70B
-## Authentication
+Used for:
-* NextAuth
+* Diagnostic reasoning
+* Follow-up generation
+* Root cause analysis
+* Recommendation generation
+
+---
## Document Processing
-* pdf-parse
+Supported formats:
+
+* PDF
+* Text
+* Links
+* Images
+* Videos
+
+Processing flow:
+
+```text
+Document
+ ↓
+Parsing
+ ↓
+Chunking
+ ↓
+Metadata Extraction
+ ↓
+Indexing
+```
---
-# 📂 Project Structure
+# Project Structure
```bash
src/
@@ -395,93 +687,289 @@ src/
├── actions/
├── components/
├── lib/
-├── types/
├── prisma/
+├── types/
└── uploads/
+
+public/
+screenshots/
```
---
-# 🚀 Getting Started
+# Key Engineering Decisions
-## Installation
+## Why MOSS?
-```bash
-git clone https://github.com/your-repository/helix.git
+MOSS was selected because it provides:
-cd helix
+* Fast retrieval
+* Hybrid search capabilities
+* Session memory
+* Metadata filtering
+* Product-specific indexing
-npm install
-```
+This allowed us to focus on diagnostic reasoning instead of building retrieval infrastructure from scratch.
---
-## Environment Variables
+## Why Groq?
-```env
-DATABASE_URL=
+Diagnostic conversations require low latency.
-NEXTAUTH_SECRET=
+Groq provides:
-MOSS_PROJECT_ID=
+* Fast inference
+* Reliable responses
+* Strong reasoning capabilities
-MOSS_PROJECT_KEY=
+making it suitable for real-time troubleshooting.
-GROQ_API_KEY=
-```
+---
+
+## Why Next.js?
+
+Next.js allows:
+
+* Unified frontend and backend
+* Server Actions
+* API Routes
+* Fast development
+
+which is particularly important during a 24-hour hackathon.
---
-## Run Development Server
+# Challenges Faced
-```bash
-npm run dev
-```
+During development we encountered several challenges:
+
+### Building a diagnostic workflow
+
+Most assistants retrieve information and answer immediately.
+
+Designing a system that investigates problems required a completely different prompting strategy.
+
+### Maintaining context
+
+Diagnostic conversations often span multiple steps.
+
+Ensuring context remained available throughout the conversation was critical.
+
+### Grounding responses
+
+Recommendations needed to remain tied to documentation to avoid hallucinations.
+
+### Document ingestion
+
+Supporting different resource types required a consistent ingestion pipeline.
---
-# 🎯 Future Scope
+# Future Scope
+
+Planned improvements include:
-* Voice Troubleshooting
-* Image-Based Diagnostics
-* Maintenance Reminder Engine
-* Spare Parts Recommendation System
-* Recall Alert System
-* Warranty Monitoring
-* Product Health Analytics
+* Voice-based troubleshooting
+* Image-based diagnostics
+* Maintenance reminders
+* Product ownership dashboard
+* Spare part recommendations
+* Warranty monitoring
+* Recall notifications
+* Product analytics
+* Service center integration
+* Technician handoff workflows
---
-# 🏆 Why Helix?
+Example Diagnostic Session
+User
+
+My scooter horn is not working.
+
+Helix
+
+Do the headlights work normally?
+
+User
+
+Yes.
+
+Helix
+
+Is the horn completely silent or weak?
+
+User
+
+Completely silent.
+
+Helix
+
+Has any electrical work been performed recently?
+
+User
+
+No.
+
+Helix
+
+Please inspect Fuse F3 (10A) beneath the front panel.
+
+User
+
+The fuse appears damaged.
+
+Helix
+
+Based on the diagnostic information provided, the most probable cause is a blown horn circuit fuse.
+
+Recommended Action:
+
+Replace Fuse F3 (10A).
+
+Reference:
+
+Service Manual – Section 4.2
+
+Why We Chose This Stack
+Next.js
+Unified frontend and backend
+Server Actions
+API Routes
+Fast development cycle
+MOSS
+Retrieval infrastructure
+Hybrid search
+Session memory
+Metadata filtering
+Product-specific indexes
+Groq
+Low latency inference
+Fast streaming responses
+Strong reasoning performance
+Prisma
+Type-safe database access
+Rapid schema development
+Easy migrations
+Scalability Considerations
+
+The architecture was designed to remain scalable as the number of products and support resources grows.
+
+Product-Level Isolation
+
+Each product receives its own knowledge index.
+
+Benefits:
+
+Reduced retrieval noise
+Better relevance
+Faster searches
+Session-Based Memory
+
+Conversations maintain context without requiring entire chat histories to be processed repeatedly.
+
+Metadata Filtering
-Most support systems provide answers.
+Product, company, and document metadata improve retrieval precision.
-Helix provides diagnosis.
+Chunked Document Processing
-Most chatbots retrieve documentation.
+Large documents are split into optimized chunks before indexing to improve retrieval quality.
-Helix investigates problems.
+Development Constraints
-Most assistants stop at information.
+This project was built under hackathon constraints.
-Helix guides users toward resolution.
+Constraints
+24-hour development window
+Small team size
+Limited implementation time
+Need for meaningful MOSS integration
+Approach
+
+Instead of implementing a large number of unfinished features, we focused on:
+
+Product marketplace
+Knowledge repository
+Diagnostic assistant
+Retrieval quality
+User experience
+
+to ensure the core workflow was complete and functional.
+
+Roadmap
+Phase 1
+Product marketplace
+Knowledge repository
+Diagnostic assistant
+MOSS integration
+Phase 2
+Voice-based troubleshooting
+Image diagnostics
+Enhanced multilingual support
+Phase 3
+Maintenance reminders
+Product ownership dashboard
+Warranty monitoring
+Phase 4
+Spare parts marketplace
+Service center integration
+Predictive maintenance
+Product analytics
+
+# Local Setup
+
+## Clone Repository
+
+```bash
+git clone 
