Can I vibecode Afforai?

Build a comprehensive, private document research assistant and reference management workspace designed to serve as a powerful local alternative to Afforai. This tool must enable users to ingest documents, chat with their content, and manage references efficiently.
Use exactly this technology stack: Next.js 15 (App Router), TypeScript, SQLite (with Prisma), and Playwright for end-to-end testing. Integrate LangChain or LlamaIndex for document processing and vectorization, and use a local vector database like ChromaDB or a simple in-memory vector store backed by SQLite.
The primary job is to create an all-in-one workspace. The user must be able to upload files (PDF, DOCX, TXT, Markdown) or provide URLs. The system will extract the text, chunk it, generate embeddings, and store them for semantic search.
Implement a side-by-side interactive document viewer. On the left, render the uploaded document (e.g., using a PDF viewer library). On the right, provide a chat interface where the user can ask questions about the document.
The AI chat must utilize Retrieval-Augmented Generation (RAG). When the user asks a question, the system searches the vector store for relevant document chunks, injects them into the LLM prompt, and streams the answer back. The AI must provide inline citations linking directly to the specific chunks or pages of the source document used to generate the answer.
Include three distinct research modes in the chat: 'Document Mode' (queries only the uploaded files), 'Web Search Mode' (uses a search API to find real-time internet data), and 'Local Library Mode' (queries all documents the user has ever uploaded).
Implement a reference management library. When a document is uploaded, automatically attempt to extract its metadata (title, authors, publication date) using an LLM pass, and store it in SQLite. Allow the user to manually edit this metadata and organize documents into folders or tags.
Include a citation generator feature that can export the selected documents' metadata into standard formats (APA, MLA, Chicago).
Ensure the interface is modern, polished, and highly responsive. Include clear loading states during document processing, toast notifications for success/failure, and empty states for the library.
Ensure all data is stored locally. The application must not include analytics, telemetry, or third-party user accounts. Put all external API keys (like OpenAI or a web search API) in a `.env` file and provide a `.env.example`.
Deliberately exclude access to proprietary scholarly databases, team collaboration features, and live cloud syncing.
Write focused unit tests for the document chunking and metadata extraction logic. Add at least one end-to-end smoke test using Playwright that simulates uploading a text file and asking a question about it.
Create a detailed README explaining how to set up the local vector store, configure API keys, and run the application in development and production modes.