Can I vibecode Scite?

Build a comprehensive academic research and citation intelligence workspace, heavily inspired by the core functionalities of Scite. The primary objective is to create an advanced tool that allows researchers to manage, analyze, and synthesize scientific literature with a focus on citation context and evidence-based AI assistance.

Use this modern stack: Next.js (App Router), TypeScript, TailwindCSS for styling, Prisma with PostgreSQL or SQLite for local/self-hosted data persistence, and integrate an LLM provider (e.g., OpenAI or Anthropic via an SDK) for the AI assistant and text analysis capabilities.

Key Features to Implement:
1. **Smart Citation Extraction & Classification:** Allow users to upload PDFs or input URLs of research papers. Extract citation statements and use the LLM to classify them automatically as "supporting," "contrasting," or "mentioning." Display these contexts prominently alongside the paper's metadata.
2. **AI-Powered Research Assistant:** Create a conversational interface ("Assistant") where users can ask complex scientific questions. The assistant must query the user's uploaded library or connected scholarly APIs (like Semantic Scholar, Crossref, or OpenAlex) to provide grounded answers. Critically, every claim in the assistant's response must be backed by a specific inline citation linking back to the source text.
3. **Citation Statement Search Engine:** Implement a robust search feature that doesn't just search paper titles and abstracts, but searches within the extracted citation statements across the entire user library. Users should be able to filter search results by classification (e.g., show me only papers that 'contrast' with this specific study).
4. **Dashboard and Collections:** Allow users to organize papers into custom dashboards or collections. Provide visual summaries of the literature in a collection, such as the ratio of supporting vs. contrasting citations for key papers, helping users quickly assess the consensus in a specific field.
5. **Reference Checking Workflow:** Create a specific view where users can paste their manuscript's reference list. Parse the references and highlight any that are heavily contrasted by recent literature or have been marked as retracted (simulated via an external API or mocked data).

Technical Requirements & Constraints:
- Start from scratch and provide the complete, working project.
- Design a premium, highly responsive UI with a scholarly yet modern aesthetic. Include clear empty states, loading indicators for AI operations, and robust error handling.
- Store all user data and documents locally or in the configured database to ensure privacy.
- The AI classification and assistant features should use environment variables for API keys (`.env` file with `.env.example`).
- Implement import and export functionality for the library (e.g., BibTeX, RIS) to avoid vendor lock-in.
- Ensure the application degrading gracefully if the AI API is unavailable, keeping basic library management functional.
- Include a comprehensive README detailing setup instructions, environment variables, architecture overview, and how to run tests.
- Write Playwright E2E tests for the core workflow: uploading a document, viewing its extracted citations, and asking the assistant a question.