ruby RAG example with a console chat app added

week5/community-contributions/ruby_rag_console_chat_app/.ruby-gemset

@@ -0,0 +1 @@
+ruby_rag_test

week5/community-contributions/ruby_rag_console_chat_app/Gemfile

@@ -0,0 +1,9 @@
+# frozen_string_literal: true
+
+source "https://rubygems.org"
+
+gem 'ruby-openai'
+gem 'chroma-db', github: 'koic/chroma', branch: 'support_chroma_v2_api' # Using fork because original one comes with outdated API
+gem 'pry-byebug'
+gem 'ruby-next' # to auto-transpile the required files from source on load. Otherwise the code from chroma-db may not work correctly
+gem 'pragmatic_segmenter'

week5/community-contributions/ruby_rag_console_chat_app/Gemfile.lock

@@ -0,0 +1,86 @@
+GIT
+  remote: https://github.com/koic/chroma.git
+  revision: c457cb0541dffeb4386a8d4d47eaaa68608d2561
+  branch: support_chroma_v2_api
+  specs:
+    chroma-db (0.8.2)
+      dry-monads (~> 1.6)
+      ruby-next (~> 1.0, >= 1.0.3)
+      zeitwerk (~> 2.6.0)
+
+GEM
+  remote: https://rubygems.org/
+  specs:
+    ast (2.4.3)
+    byebug (12.0.0)
+    coderay (1.1.3)
+    concurrent-ruby (1.3.5)
+    diff-lcs (1.6.2)
+    dry-core (1.1.0)
+      concurrent-ruby (~> 1.0)
+      logger
+      zeitwerk (~> 2.6)
+    dry-monads (1.9.0)
+      concurrent-ruby (~> 1.0)
+      dry-core (~> 1.1)
+      zeitwerk (~> 2.6)
+    event_stream_parser (1.0.0)
+    faraday (2.13.4)
+      faraday-net_http (>= 2.0, < 3.5)
+      json
+      logger
+    faraday-multipart (1.1.1)
+      multipart-post (~> 2.0)
+    faraday-net_http (3.4.1)
+      net-http (>= 0.5.0)
+    json (2.13.2)
+    logger (1.7.0)
+    method_source (1.1.0)
+    multipart-post (2.4.1)
+    net-http (0.6.0)
+      uri
+    paco (0.2.3)
+    parser (3.3.9.0)
+      ast (~> 2.4.1)
+      racc
+    pragmatic_segmenter (0.3.24)
+    pry (0.15.2)
+      coderay (~> 1.1)
+      method_source (~> 1.0)
+    pry-byebug (3.11.0)
+      byebug (~> 12.0)
+      pry (>= 0.13, < 0.16)
+    racc (1.8.1)
+    require-hooks (0.2.2)
+    ruby-next (1.1.2)
+      paco (~> 0.2)
+      require-hooks (~> 0.2)
+      ruby-next-core (= 1.1.2)
+      ruby-next-parser (>= 3.4.0.2)
+      unparser (~> 0.6.0)
+    ruby-next-core (1.1.2)
+    ruby-next-parser (3.4.0.2)
+      parser (>= 3.0.3.1)
+    ruby-openai (8.1.0)
+      event_stream_parser (>= 0.3.0, < 2.0.0)
+      faraday (>= 1)
+      faraday-multipart (>= 1)
+    unparser (0.6.15)
+      diff-lcs (~> 1.3)
+      parser (>= 3.3.0)
+    uri (1.0.3)
+    zeitwerk (2.6.18)
+
+PLATFORMS
+  arm64-darwin-23
+  ruby
+
+DEPENDENCIES
+  chroma-db!
+  pragmatic_segmenter
+  pry-byebug
+  ruby-next
+  ruby-openai
+
+BUNDLED WITH
+   2.6.8

week5/community-contributions/ruby_rag_console_chat_app/README.md

@@ -0,0 +1,90 @@
+# Local RAG Knowledge Base with Llama 3.2, Chroma, and Ruby
+![Diagram](./doc/ruby_rag_diagram.png)
+
+## Project Overview
+
+This mini project demonstrates building a local Retrieval-Augmented Generation (RAG) system using the Llama 3.2 language model (via Ollama locally), Chroma vector database, and Ruby. The project includes:
+
+- Reading and processing a knowledge base consisting of markdown files organized in folders.
+- Splitting documents into context-preserving chunks using the `pragmatic_segmenter` gem for sentence-aware chunking.
+- Generating semantic embeddings of chunks with Llama 3.2 local model via an OpenAI-compatible API (Ollama).
+- Storing embeddings along with metadata and documents into a Chroma vector database collection.
+- Performing similarity search in Chroma on user queries to retrieve relevant context.
+- Constructing a conversational prompt by combining chat history, retrieved context, and user input.
+- Streaming responses from the Llama 3.2 model back to the console for real-time interaction.
+
+---
+
+## What Has Been Done
+
+- **Folder crawler and document loader:**  
+  Recursively read all markdown files in the `knowledge_base` directory, assigning document type metadata from folder names.
+
+- **Smart text chunking:**  
+  Integrated the [`pragmatic_segmenter`](https://github.com/diasks2/pragmatic_segmenter) gem to split texts into sentence-safe chunks (~1000 characters) with overlaps to preserve context and avoid cutting sentences unnaturally.
+
+- **Embeddings generation:**  
+  Leveraged the local Llama 3.2 model via Ollama's OpenAI-compatible streaming API to generate embeddings of all text chunks, enabling efficient semantic search.
+
+- **Chroma vector store integration:**  
+  Used the `chroma-db` Ruby gem with a locally hosted Chroma server (via Docker Compose) to store embeddings and metadata, and to perform similarity search queries.
+
+- **Interactive conversational loop:**  
+  Maintained chat history manually as an array of message hashes, combined with relevant retrieved chunks to form prompts fed into the Llama 3.2 model.
+
+- **Streaming chat responses:**  
+  Implemented real-time streaming of LLM output to the console leveraging the Ruby OpenAI gem streaming feature and Ollama's compatible API.
+
+---
+
+## Tricky Findings & Gotchas
+
+- **Ruby Next transpilation required for `chroma-db` gem:**  
+  The gem uses modern Ruby features and requires `gem 'ruby-next'` with `require "ruby-next/language/runtime"` loaded early to avoid LoadErrors.
+
+- **Chroma API version compatibility:**  
+  Different Chroma server versions expose different API versions (`v1` vs `v2`). The `chroma-db` Ruby gem expected v2 endpoints. Using matched versions of Chroma server and the gem, or a forked gem branch with v2 support, was crucial.
+
+- **Bundler context for scripts:**  
+  Running scripts must be done with `bundle exec` or with `require 'bundler/setup'` to load local gem dependencies correctly (especially forked gems).
+
+- **Manual management of conversational memory:**  
+  Unlike Python LangChain, no high-level Ruby library exists for conversation memory or RAG chains, so that had to be implemented as arrays of messages, and prompt assembly was manual.
+
+- **Text chunking with `pragmatic_segmenter`:**  
+  Using sentence segmentation improved context retention significantly over naïve character splitting, but required careful assembly of chunks and overlaps.
+
+- **Streaming outputs handled via custom block in Ruby OpenAI gem:**  
+  Streaming integration required capturing delta chunks from the streaming API and printing them in realtime, instead of waiting for full response.
+
+---
+
+## Setup Instructions
+
+### Requirements
+
+- Ruby 3.2.x
+- Bundler
+- Docker & Docker Compose
+- Ollama installed and running locally with the `llama3.2` model pulled
+- Basic terminal shell (macOS, Linux recommended)
+
+### Steps
+
+1. **Clone/Fork the repository:**
+2. Run `bundle install` to install Ruby dependencies
+3. Run `docker compose up -d` to boot up Chroma DB
+4. Run `ollama run llama3.2` to boot up Open Source LLM
+5. Run `bundle exec ruby seed.rb` to seed Chroma DB with chunks of data from `knowledge_base` folder
+6. Run `bundle exec ruby main.rb` to start actual conversation
+
+### Questions to try on
+1. What is the company name?
+2. When the company was establised?
+3. Which techologies does the company use?
+4. Tell me the emplooees` names and their titles?
+5. Who knows how to work with IOS?
+6. Tell me who was the client for web project?
+7. Is the company looking for IOS developer?
+
+Then you can compare it with the actual knowledge base.

week5/community-contributions/ruby_rag_console_chat_app/bin/console

@@ -0,0 +1,6 @@
+#!/usr/bin/env ruby
+require 'irb'
+require 'irb/completion'
+require_relative '../initializer'
+
+IRB.start

week5/community-contributions/ruby_rag_console_chat_app/docker-compose.yml

@@ -0,0 +1,12 @@
+services:
+  chromadb:
+    image: chromadb/chroma:latest
+    ports:
+      - "8000:8000"
+    environment:
+      - IS_PERSISTENT=TRUE
+      - PERSIST_DIRECTORY=/chroma/chroma
+      - ANONYMIZED_TELEMETRY=TRUE
+    volumes:
+      - ./chromadb:/chroma/chroma
+    restart: unless-stopped

week5/community-contributions/ruby_rag_console_chat_app/initializer.rb

@@ -0,0 +1,12 @@
+require 'pathname'
+require 'pry'
+require 'openai'
+require "ruby-next/language/runtime"
+require 'chroma-db'
+require 'logger'
+require 'json'
+
+Chroma.connect_host = "http://localhost:8000"
+Chroma.api_version = "v2"
+Chroma.logger = Logger.new($stdout)
+Chroma.log_level = Chroma::LEVEL_ERROR

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/company/about.md

@@ -0,0 +1,15 @@
+# About TechNova Solutions
+
+TechNova Solutions is a technology services company specializing in **web development**, **iOS mobile app development**, and **machine learning solutions**. Founded in 2015, we have grown into a trusted partner for startups, medium-sized businesses, and enterprises worldwide.
+
+## Mission
+To empower businesses with cutting-edge software solutions that combine usability, scalability, and intelligence.
+
+## Vision
+To be a global leader in delivering high-quality digital products that transform industries.
+
+## Values
+- Innovation: Continuously adopting new technologies.
+- Collaboration: Building strong partnerships with clients.
+- Excellence: Maintaining high standards in every project.
+- Growth: Creating opportunities for both employees and clients.

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/company/careers.md

@@ -0,0 +1,22 @@
+# Careers at TechNova Solutions
+
+We are always looking for passionate and skilled professionals to join our growing team. Our company culture emphasizes **learning**, **innovation**, and **collaboration**.
+
+## Current Openings
+- **iOS Developer**
+  - Skills: Swift, SwiftUI, UIKit, experience with App Store deployment.
+- **Web Developer**
+  - Skills: React.js, Node.js, TypeScript, REST API integration.
+- **Machine Learning Engineer**
+  - Skills: Python, TensorFlow/PyTorch, data pipelines, NLP, computer vision.
+- **Project Manager**
+  - Skills: Agile methodologies, client communication, technical understanding.
+
+## Why Work With Us?
+- Remote-friendly with flexible working hours.
+- Opportunities to work on international projects.
+- Professional development budget for training and conferences.
+- Inclusive and diverse work culture.
+
+## How to Apply
+Send your CV and portfolio to **careers@technova.com**

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/company/overview.md

@@ -0,0 +1,37 @@
+# Company Overview: TechNova Solutions
+
+## Industry
+Software Development and Consulting
+
+## Headquarters
+Berlin, Germany (with remote teams worldwide)
+
+## Services
+TechNova Solutions provides:
+1. **Web Development**
+   - Full-stack web solutions
+   - Modern scalable architectures
+   - E-commerce and SaaS platforms
+
+2. **iOS App Development**
+   - Custom mobile applications
+   - Swift/SwiftUI-based native solutions
+   - App Store deployment and maintenance
+
+3. **Machine Learning Development**
+   - Predictive analytics
+   - Natural Language Processing (NLP) models
+   - Computer vision and automation solutions
+
+## Client Base
+- Startups launching MVPs
+- Growing SaaS companies
+- Enterprises requiring digital transformation
+
+## Key Achievements
+- Delivered over **150 digital projects** across industries.
+- Worked with clients in **FinTech, Healthcare, and E-commerce**.
+- Recognized as a **Top B2B Tech Firm in 2023** by Clutch.
+
+## Culture
+Our team thrives on **technical challenges** and believes in **continuous improvement**. We foster a work environment where creativity meets execution.

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/contracts/ios_project.md

@@ -0,0 +1,34 @@
+# iOS App Development Agreement
+
+**Parties:**  
+Client: Acme Innovations LLC  
+Developer: TechNova Solutions
+
+## Project Scope
+TechNova Solutions will design, develop, and deploy a custom iOS mobile application for Acme Innovations, including user authentication, dashboard, push notifications, and App Store submission.
+
+## Deliverables
+- iOS app (Swift/SwiftUI) for iPhone  
+- Admin dashboard (web-based)  
+- User and technical documentation  
+- 2 post-launch maintenance updates
+
+## Timeline
+- Project kickoff: October 1, 2025  
+- Beta delivery: December 1, 2025  
+- Final delivery: January 15, 2026
+
+## Payment Terms
+- 30% upfront  
+- 40% upon beta delivery  
+- 30% upon final delivery
+
+## Intellectual Property
+All source code and produced assets are transferred to the client upon final payment.
+
+## Confidentiality
+Both parties will maintain strict confidentiality regarding proprietary information.
+
+## Signatures  
+_Client Rep:_ _____________________  
+_TechNova Rep:_ ___________________

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/contracts/ml_project.md

@@ -0,0 +1,35 @@
+# Machine Learning Project Agreement
+
+**Parties:**  
+Client: HealthPlus Analytics  
+Developer: TechNova Solutions
+
+## Objective
+TechNova Solutions will develop a predictive analytics solution using machine learning models for healthcare data, aimed at reducing patient readmission rates.
+
+## Statement of Work
+- Data preprocessing and anonymization  
+- Model development (Python, TensorFlow, scikit-learn)  
+- Evaluation and performance tuning  
+- Deployment and staff training  
+- Integration with client’s data system
+
+## Timeline
+- Data delivery: November 1, 2025  
+- Prototype model: January 10, 2026  
+- Final delivery: February 20, 2026
+
+## Payment Schedule
+- 25% upon signing  
+- 50% upon prototype  
+- 25% upon final acceptance
+
+## Confidentiality & Data Privacy
+Strict data privacy and NDA agreement. No data shared outside of project scope.
+
+## Intellectual Property
+Custom ML models and documentation are transferred to the client at project completion.
+
+## Signatures  
+_Client Rep:_ _____________________  
+_TechNova Rep:_ ___________________

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/contracts/web_project.md

@@ -0,0 +1,35 @@
+# Web Development Services Agreement
+
+**Parties:**  
+Client: BlueSky E-Commerce  
+Developer: TechNova Solutions
+
+## Project Description
+TechNova Solutions will build a responsive e-commerce web platform with integrated payment gateway, product catalog, and analytics dashboard for BlueSky E-Commerce.
+
+## Scope of Work
+- Modern UX/UI design  
+- Frontend (React.js), backend (Node.js, PostgreSQL)  
+- Payment integration (Stripe/PayPal)  
+- Mobile-responsive design  
+- 3 months maintenance & support
+
+## Project Phases
+1. Specification & wireframing  
+2. Development & staging  
+3. Testing & deployment  
+4. Support period
+
+## Pricing & Payment
+- Total: €25,000  
+- 40% upon contract start, 40% after development milestone, 20% after deployment
+
+## Intellectual Property
+All code, documentation, and design assets become property of the client post-final payment.
+
+## Dispute Resolution
+Any disputes to be resolved via arbitration in Berlin, Germany.
+
+## Signatures  
+_Client Rep:_ _____________________  
+_TechNova Rep:_ ___________________

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/employees/alex_johnson.md

@@ -0,0 +1,54 @@
+# HR Record
+
+# Alex Johnson
+
+## Summary
+- **Date of Birth**: May 10, 1990  
+- **Job Title**: iOS Engineer  
+- **Location**: Berlin, Germany  
+
+## Career Progression
+- **March 2015**: Joined TechNova Solutions as Junior iOS Developer, focused on building and refining mobile features.  
+- **July 2017**: Promoted to iOS Engineer, led SwiftUI adoption in new projects.  
+- **November 2019**: Spearheaded the launch of the flagship iOS app with 100k+ downloads.  
+- **August 2023**: Mentors junior iOS developers and contributes to open-source Swift libraries.
+
+## Annual Performance History
+- **2021**:  
+  - **Performance Rating**: 4.6/5  
+  - **Key Achievements**: Implemented a new push notification system improving user retention by 15%.
+
+- **2022**:  
+  - **Performance Rating**: 4.8/5  
+  - **Key Achievements**: Led successful iOS 16 app updates supporting new platform features.
+
+- **2023**:  
+  - **Performance Rating**: 4.7/5  
+  - **Key Achievements**: Reduced app crash rate by 30% with detailed runtime diagnostics tools.
+
+## Compensation History
+- **2021**:  
+  - **Base Salary**: €65,000  
+  - **Bonus**: €7,000  
+
+- **2022**:  
+  - **Base Salary**: €70,000  
+  - **Bonus**: €8,000  
+
+- **2023**:  
+  - **Base Salary**: €75,000  
+  - **Bonus**: €10,000  
+
+## Other HR Notes
+- **Training Completed**:  
+  - Advanced SwiftUI Workshop (2022)  
+  - iOS Performance Optimization Course (2023)  
+
+- **Awards**:  
+  - Employee of the Quarter Q4 2022  
+
+- **Interests**:  
+  - Open-source iOS community, mountain biking.
+
+- **Feedback from HR**:  
+  - Alex is highly reliable, innovative, and a technical leader on the iOS team.

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/employees/emma_clark.md

@@ -0,0 +1,36 @@
+# HR Record
+
+# Emma Clark
+
+## Summary
+- **Date of Birth**: December 1, 1994  
+- **Job Title**: Junior iOS Developer  
+- **Location**: Remote (Barcelona, Spain)  
+
+## Career Progression
+- **January 2023**: Joined TechNova Solutions as Junior iOS Developer.  
+- **June 2023**: Assisted in feature development for client app projects.  
+- **September 2023**: Completed SwiftUI fundamentals training.
+
+## Annual Performance History
+- **2023**:  
+  - **Performance Rating**: 4.3/5  
+  - **Key Achievements**: Successfully delivered bug fixes that reduced crash reports by 20%.
+
+## Compensation History
+- **2023**:  
+  - **Base Salary**: €45,000  
+  - **Bonus**: €2,000  
+
+## Other HR Notes
+- **Training Completed**:  
+  - SwiftUI Fundamentals (2023)  
+
+- **Awards**:  
+  - Rookie of the Month (August 2023)  
+
+- **Interests**:  
+  - Mobile UI design, digital painting.
+
+- **Feedback from HR**:  
+  - Emma is eager to learn and shows promising potential in iOS development.

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/employees/li_wang.md

@@ -0,0 +1,54 @@
+# HR Record
+
+# Li Wang
+
+## Summary
+- **Date of Birth**: November 3, 1991  
+- **Job Title**: Machine Learning Researcher  
+- **Location**: Berlin, Germany  
+
+## Career Progression
+- **September 2016**: Joined as ML Research Intern focusing on NLP models.  
+- **March 2018**: Promoted to ML Researcher, published 3 papers in AI conferences.  
+- **July 2021**: Led development of custom ML pipeline for predictive analytics client project.  
+- **January 2024**: Mentoring new ML engineers and collaborating cross-functionally.
+
+## Annual Performance History
+- **2021**:  
+  - **Performance Rating**: 4.8/5  
+  - **Key Achievements**: Developed novel transformer model adaptation improving accuracy by 12%.
+
+- **2022**:  
+  - **Performance Rating**: 4.9/5  
+  - **Key Achievements**: Successfully delivered a computer vision system automating quality inspection.
+
+- **2023**:  
+  - **Performance Rating**: 4.9/5  
+  - **Key Achievements**: Published an internal white paper on explainable AI.
+
+## Compensation History
+- **2021**:  
+  - **Base Salary**: €68,000  
+  - **Bonus**: €7,500  
+
+- **2022**:  
+  - **Base Salary**: €75,000  
+  - **Bonus**: €9,000  
+
+- **2023**:  
+  - **Base Salary**: €80,000  
+  - **Bonus**: €11,000  
+
+## Other HR Notes
+- **Training Completed**:  
+  - AI Model Interpretability Workshop (2022)  
+  - TensorFlow Advanced Techniques (2023)  
+
+- **Awards**:  
+  - Best Paper Award, AI Conference 2022  
+
+- **Interests**:  
+  - AI ethics, chess.
+
+- **Feedback from HR**:  
+  - Li is a deep thinker and innovator, key to the ML capabilities of TechNova.

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/employees/oliver_smith.md

@@ -0,0 +1,54 @@
+# HR Record
+
+# Oliver Smith
+
+## Summary
+- **Date of Birth**: February 14, 1989  
+- **Job Title**: Fullstack Engineer  
+- **Location**: Berlin, Germany  
+
+## Career Progression
+- **May 2013**: Joined TechNova as Junior Web Developer.  
+- **December 2016**: Became Fullstack Engineer handling Node.js backend and iOS frontend integration.  
+- **September 2019**: Led cross-platform product integration projects.  
+- **July 2023**: Leads DevOps initiatives for CI/CD pipeline improvements.
+
+## Annual Performance History
+- **2021**:  
+  - **Performance Rating**: 4.5/5  
+  - **Key Achievements**: Automated deployment pipelines improving release frequency by 25%.
+
+- **2022**:  
+  - **Performance Rating**: 4.7/5  
+  - **Key Achievements**: Improved API response times by 35%.
+
+- **2023**:  
+  - **Performance Rating**: 4.8/5  
+  - **Key Achievements**: Led migration to containerized microservices architecture.
+
+## Compensation History
+- **2021**:  
+  - **Base Salary**: €70,000  
+  - **Bonus**: €7,000  
+
+- **2022**:  
+  - **Base Salary**: €75,000  
+  - **Bonus**: €8,000  
+
+- **2023**:  
+  - **Base Salary**: €80,000  
+  - **Bonus**: €9,000  
+
+## Other HR Notes
+- **Training Completed**:  
+  - Kubernetes & Docker Training (2022)  
+  - Advanced DevOps Practices (2023)  
+
+- **Awards**:  
+  - TechNova Team Player Award 2023  
+
+- **Interests**:  
+  - Open-source contributor, drone photography.
+
+- **Feedback from HR**:  
+  - Oliver’s versatile skill set bridges frontend, backend, and infrastructure seamlessly.

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/employees/sophia_martinez.md

@@ -0,0 +1,54 @@
+# HR Record
+
+# Sophia Martinez
+
+## Summary
+- **Date of Birth**: August 22, 1988  
+- **Job Title**: Senior Web Developer  
+- **Location**: Remote (Madrid, Spain)  
+
+## Career Progression
+- **January 2014**: Started as Junior Frontend Developer working on React projects.  
+- **June 2016**: Promoted to Web Developer handling full-stack responsibilities with Node.js backend.  
+- **May 2019**: Led e-commerce platform redeployment for major client.  
+- **February 2023**: Became Senior Web Developer and technical team lead.
+
+## Annual Performance History
+- **2021**:  
+  - **Performance Rating**: 4.7/5  
+  - **Key Achievements**: Spearheaded transition to microservices architecture.
+
+- **2022**:  
+  - **Performance Rating**: 4.9/5  
+  - **Key Achievements**: Delivered a complex multi-tenant SaaS platform on time.
+
+- **2023**:  
+  - **Performance Rating**: 5.0/5  
+  - **Key Achievements**: Implemented comprehensive CI/CD pipelines improving deployment speed by 40%.
+
+## Compensation History
+- **2021**:  
+  - **Base Salary**: €72,000  
+  - **Bonus**: €8,500  
+
+- **2022**:  
+  - **Base Salary**: €78,000  
+  - **Bonus**: €10,000  
+
+- **2023**:  
+  - **Base Salary**: €85,000  
+  - **Bonus**: €12,000  
+
+## Other HR Notes
+- **Training Completed**:  
+  - Kubernetes Administration (2022)  
+  - Advanced React Patterns (2023)  
+
+- **Awards**:  
+  - TechNova Innovation Award 2023  
+
+- **Interests**:  
+  - Technical blogging, hiking.
+
+- **Feedback from HR**:  
+  - Sophia is a strategic thinker and a force-multiplier for the web team.

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/products/ios_app_dev.md

@@ -0,0 +1,29 @@
+# iOS App Development Services
+
+## Overview
+TechNova Solutions offers custom native iOS app development using the latest Swift and SwiftUI technologies to build performant, user-friendly applications for iPhone and iPad.
+
+## Key Features
+- Native UI/UX design optimized for iOS platform  
+- Integration with Apple services: HealthKit, Push Notifications, Siri Shortcuts  
+- Robust authentication and security features  
+- App Store submission support and maintenance  
+- Continuous updates for compatibility with the latest iOS versions
+
+## Typical Use Cases
+- Consumer mobile apps  
+- Enterprise tools and productivity apps  
+- E-commerce and retail apps  
+- Fitness and wellness tracking
+
+## Technologies
+- Swift & SwiftUI  
+- Combine for reactive programming  
+- Core Data and CloudKit for persistence  
+- XCTest for automated testing
+
+## Benefits
+- Faster time-to-market with modern workflows  
+- High app performance and responsiveness  
+- Access to native hardware capabilities  
+- Dedicated support throughout app lifecycle

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/products/ml_solutions.md

@@ -0,0 +1,36 @@
+# Machine Learning Solutions
+
+## Overview
+TechNova Solutions delivers custom machine learning models and AI systems to help businesses unlock insights, automate processes, and drive innovation.
+
+## Key Capabilities
+- Predictive analytics and forecasting  
+- Natural Language Processing (NLP) for text analysis and chatbots  
+- Computer vision for image processing and quality control  
+- Recommendation engines and personalization  
+- Automated data preprocessing and feature engineering
+
+## Project Workflow
+1. Data collection and cleaning  
+2. Model development and training  
+3. Evaluation and performance tuning  
+4. Deployment and integration  
+5. Ongoing monitoring and updates
+
+## Typical Applications
+- Customer behavior prediction  
+- Fraud detection  
+- Medical diagnostics and image analysis  
+- Automated content moderation  
+- Supply chain optimization
+
+## Technologies
+- Python, TensorFlow, PyTorch  
+- Scikit-learn, Pandas, NumPy  
+- Cloud ML platforms (AWS SageMaker, GCP AI Platform)  
+
+## Benefits
+- Data-driven decision making  
+- Process automation and efficiency  
+- Enhanced customer experience through personalization  
+- Competitive advantage via AI innovation

week5/community-contributions/ruby_rag_console_chat_app/knowledge_base/products/web_dev_services.md

@@ -0,0 +1,37 @@
+# Web Development Services
+
+## Overview
+Our full-stack web development solutions provide scalable, responsive, and secure web applications tailored to diverse business needs, from startups to enterprises.
+
+## Core Services
+- Frontend development using React.js, Vue.js, or Angular  
+- Backend APIs and microservices with Node.js, Express, and PostgreSQL  
+- E-commerce platform design and integration  
+- Content management system (CMS) development  
+- Cloud deployment and DevOps automation
+
+## Features
+- Mobile-first responsive design  
+- Secure authentication and role-based access  
+- API integrations with third-party services  
+- Real-time data synchronization and notifications  
+- Scalable architecture for high-traffic applications
+
+## Typical Clients
+- SaaS businesses  
+- Online retailers and marketplaces  
+- Media and content platforms  
+- Financial and healthcare organizations
+
+## Technologies
+- JavaScript/TypeScript  
+- React, Vue, Angular  
+- Node.js, Express  
+- PostgreSQL, MongoDB  
+- Docker, Kubernetes
+
+## Benefits
+- Custom solutions built for your unique business challenges  
+- Improved user engagement and retention  
+- Scalable systems ready for growth  
+- Expert support and maintenance

week5/community-contributions/ruby_rag_console_chat_app/main.rb

@@ -0,0 +1,68 @@
+require_relative 'initializer'
+
+chat_history = [
+  { role: "system", content: "You are helpful assistant. If you do not know the answer, do not make it up, just say you do not know." }
+]
+openai = OpenAI::Client.new(uri_base: 'http://localhost:11434/v1', access_token: 'ollama') # LLaMa locally
+
+def get_embeddings(text, client)
+  response = client.embeddings(parameters: { model: "llama3.2", input: text })
+  response["data"].first["embedding"]
+end
+
+def build_prompt(chat_history:, context_chunks:, user_message:)
+  context_text = context_chunks.join("\n---\n")
+
+  <<~PROMPT
+    Here is an additional context from our knowledge base:
+    #{context_text}
+
+    New question from a user:
+    User: #{user_message}
+
+    Please answer using the context above.
+  PROMPT
+end
+
+collection_name = "my_collection"
+collection = Chroma::Resources::Collection.get(collection_name)
+
+puts "Welcome to our little experiment!"
+
+while true
+  user_message = gets.chomp
+
+  if user_message == 'exit'
+    puts "Ending the session"
+    exit 0
+  end
+
+  query_embedding = get_embeddings(user_message, openai)
+  results = collection.query(query_embeddings: [query_embedding], results: 15)
+  retrieved_chunks = results.map(&:document)
+
+  prompt = build_prompt(
+    chat_history: chat_history,
+    context_chunks: retrieved_chunks,
+    user_message: user_message
+  )
+  chat_history << { role: "user", content: prompt }
+
+  response_message = ""
+  openai.chat(parameters: {
+    model: "llama3.2",
+    messages: chat_history,
+    stream: proc { |chunk, _bytesize|
+      delta = chunk.dig("choices", 0, "delta", "content")
+      next unless delta
+      print delta
+      response_message << delta
+    }
+  })
+  chat_history << { role: "assistant", content: response_message }
+  puts
+end
+
+
+
+

week5/community-contributions/ruby_rag_console_chat_app/seed.rb

@@ -0,0 +1,129 @@
+require_relative 'initializer'
+require 'pragmatic_segmenter'
+
+folders = Dir.glob("knowledge_base/*").select { |f| File.directory?(f) }
+
+documents = []
+
+folders.each do |folder|
+  doc_type = File.basename(folder)
+  Dir.glob("#{folder}/**/*.md").each do |filepath|
+    content = File.read(filepath, encoding: 'utf-8')
+    doc = {
+      content: content,
+      path: filepath,
+      metadata: { "doc_type" => doc_type }
+    }
+    documents << doc
+  end
+end
+
+# chunk_size and chunk_overlap are configurable for getting better results
+def split_text(text, chunk_size: 1000, chunk_overlap: 200)
+  chunks = []
+  start = 0
+  while start < text.length
+    finish = [start + chunk_size, text.length].min
+    chunks << text[start...finish]
+    break if finish == text.length
+    start += (chunk_size - chunk_overlap)
+  end
+  chunks
+end
+
+def split_text_by_sentence(text, chunk_size: 1500, chunk_overlap: 200)
+  ps = PragmaticSegmenter::Segmenter.new(text: text)
+  sentences = ps.segment
+  chunks = []
+  current_chunk = ""
+  sentences.each do |sentence|
+    if (current_chunk + sentence).length > chunk_size
+      chunks << current_chunk.strip
+      # For overlap, take last N chars from current_chunk (optionally at sentence boundary)
+      overlap = current_chunk[-chunk_overlap..-1] || ""
+      current_chunk = overlap + sentence
+    else
+      current_chunk += " " unless current_chunk.empty?
+      current_chunk += sentence
+    end
+  end
+  chunks << current_chunk.strip unless current_chunk.empty?
+  chunks
+end
+
+chunks = []
+
+# documents.each do |doc|
+#   split_text(doc[:content]).each_with_index do |chunk, idx|
+#     chunks << {
+#       content: chunk,
+#       metadata: doc[:metadata].merge({ "chunk_index" => idx, "path" => doc[:path] })
+#     }
+#   end
+# end
+
+documents.each do |doc|
+  split_text_by_sentence(doc[:content]).each_with_index do |chunk, idx|
+    chunks << {
+      content: chunk,
+      metadata: doc[:metadata].merge({ "chunk_index" => idx, "path" => doc[:path] })
+    }
+  end
+end
+
+puts "Chucks count: #{chunks.count}"
+puts "Document types found: #{chunks.map { _1[:metadata]['doc_type']}.uniq.join(', ') }"
+
+# 1. Set up OpenAI client (replace with RubyLLM or HTTP if using HuggingFace)
+# openai = OpenAI::Client.new(access_token: ENV['OPENAI_API_KEY']) # OpenAI API, remotely
+openai = OpenAI::Client.new(uri_base: 'http://localhost:11434/v1', access_token: 'ollama') # LLaMa, locally
+
+# 2. Get embeddings for each chunk
+def get_embedding(text, client)
+  response = client.embeddings(parameters: { model: "llama3.2", input: text })
+  response["data"].first["embedding"]
+end
+
+# Check current Chrome server version
+version = Chroma::Resources::Database.version
+puts version
+
+collection_name = "my_collection"
+collection = begin
+  Chroma::Resources::Collection.get(collection_name)
+rescue Chroma::APIError => e
+  nil
+end
+
+if collection
+  puts "Collection already exists"
+  puts "Do you want to reset it? (y/n)"
+  answer = gets.chomp
+  if answer == 'y'
+    Chroma::Resources::Collection.delete(collection_name)
+    puts 'Collection deleted'
+    exit 0
+  end
+end
+
+puts "Creating collection - #{collection_name}"
+collection = Chroma::Resources::Collection.create(collection_name, { lang: "ruby" }) unless collection
+
+chunks.each do |chunk|
+  chunk[:embedding] = get_embedding(chunk[:content], openai)
+end
+
+# 4. Insert into Chroma
+embeddings = chunks.each_with_index.map do |chunk, idx|
+  Chroma::Resources::Embedding.new(
+    id: "chunk-#{idx}",
+    embedding: chunk[:embedding],
+    metadata: chunk[:metadata],
+    document: chunk[:content]
+  )
+end
+
+collection.add(embeddings)
+puts "Vectorstore created with #{embeddings.size} documents"
+
+# Now 'chunks' is an array of hashes with chunk[:] and metadata.