MemoryVerse AI '26

Reviewer Notes and Project Documentation 1. Setup and Installation Steps (Local VS Code) Reviewers can clone and spin up the project locally in less than two minutes by opening VS Code, opening the terminal, and running the following standard commands: First, clone the repository using your git link and change your directory into the project folder using the command: git clone followed by cd MemoryVerse-AI-26. Second, create a clean environment by running: python -m venv venv. Third, activate the environment. If you are on Windows, use: .\venv\Scripts\activate. If you are on Mac or Linux, use: source venv/bin/activate. Fourth, install the project's minimalist dependencies by running: pip install streamlit chromadb google-genai pypdf requests. Next, provide your Gemini API key using one of these two options: Option A (Inside the Code File): Open app.py in VS Code. Right at the top under the initialization header, paste your key directly into the string: HARDCODED_KEY = "your-actual-key-here". Remember to clear this out before making the repository public! Option B (Terminal Environment Variable): If you prefer using the terminal, run set GEMINI_API_KEY="your-key-here" on Windows CMD, $env:GEMINI_API_KEY="your-key-here" on Windows PowerShell, or export GEMINI_API_KEY="your-key-here" on Mac/Linux. Finally, launch the application by running: streamlit run app.py. The interface will instantly initialize and open a live preview in your web browser. 2. Testing Credentials and Sample Sandbox Data Because this system runs completely on a local, zero-authentication architecture built for the hackathon prototype, no login credentials or user accounts are required. Reviewers can immediately test the operational flow using the following sample data: For GitHub Sync Ingestion: Enter any active, public GitHub profile username into the input box. For example, you can use the test profile T-Sayuj to instantly test the live repository indexing pipeline. For Document Upload Ingestion: Drag and drop any standard multi-page academic PDF, such as a student resume, an internship completion certificate, or a project report document. For Semantic Search Validation: Once documents are fully processed and ingested, navigate to the search tab and try entering conversational queries like: "Show my experience managing backend databases" or "Where did I display leadership or team management?". 3. Key Architectural and Design Decisions Single-File Streamlit Architecture: The decision to consolidate the entire technical pipeline into a single app.py script eliminates boilerplate state management and complex multi-file imports. During a rapid hackathon window, this maximizes app stability and guarantees that reviewers won't hit broken local file paths. Local Vector Storage with ChromaDB: Using an embedded, local instance of ChromaDB that saves files directly inside a project folder called memoryverse_db bypasses the need for cloud database configuration. It runs entirely on your local machine with zero network latency, ensuring it won't crash due to third-party cloud service limits. The Metadata Enrichment Strategy: Traditional applications search matching literal text strings. By routing text content to the Gemini API to extract explicit summaries and metadata tag lists right at the moment of upload, the system builds an organized index that lets users filter and review skills instantly. 4. Core Features Implemented Automated Data Ingestion: The app features multi-source text processing using helper libraries to pull clean text layers out of uploaded PDFs alongside live web requests to fetch public GitHub repositories. Intelligent AI Processing: Real-time semantic parsing maps raw text inputs into clean JSON summaries and extracts key professional tags like coding languages, soft skills, and years. Contextual Vector Queries: By comparing natural language inputs against stored vector blocks, the system lets users ask conversational questions about their past experience instead of typing rigid keywords. Automated Portfolio Synthesis: With a single click, the system pulls all independent records out of the database and rewrites them into a clean, recruiter-ready professional identity document.