Building a 5-in-1 Local LLM App with Flutter Web and Flask

The idea started simple: build five small apps that demonstrate what a local LLM can do with private data — data you'd never send to a cloud API. A secret diary with sentiment analysis. A password strength checker. A confidential document summarizer. Someone looked at the prototypes and said: "These ideas are boring." They were right. Five HTML files hitting a vLLM endpoint with generic prompts isn't exactly groundbreaking. But the criticism led to something much more interesting: what if the "private data" wasn't synthetic — what if it was my own 874MB of Claude Code session history? The Pivot: Mining Your Own Development History Claude Code stores conversation history as JSONL files. Over months of heavy use, mine had grown to 874MB — containing every question I'd asked, every code snippet Claude had generated, every debugging session, every architectural decision. That's not just chat logs. That's a structured record of how I think, what problems I solve, and how my approach has evo