explAIn is a two-person messaging app built on a single idea: an LLM should be scaffolding for a human conversation, not a substitute for it. Every 1:1 conversation has three voices — you, a private advisor that helps you refine your own drafts before you send, and a shared mediator with a switchable persona that processes messages between both parties. The non-negotiable rule is transparency: every transformation an LLM makes is visible to the people it affects. There is no hidden ghostwriter and no black box — the “show your work” panel is the literal product.
This wiki documents the product thesis, the LLM-assisted messaging model, the architecture, and the design decisions behind explAIn at a portfolio level. The canonical source of truth is docs/design.md in the (private) main repository.
Quick Links
- Product Vision — the thesis and the transparency contract
- The Messaging Model — advisor, mediator, and persona behaviors
- Architecture — stack, message lifecycle, and data model
- Developer Guide — repo layout, running locally, deploying
- Design Decisions — scope, roadmap, and trade-offs
Key facts
- What it is: a 1:1 messaging app where each party gets a private LLM advisor and both share a persona-driven LLM mediator, with full transparency of the shared layer.
- Stack: Expo + Expo Router + React Native + NativeWind (one codebase for iOS, Android, Web); Supabase (Auth, Postgres, Realtime, Edge Functions); Vercel AI SDK v5 for LLM routing; pnpm + Turborepo monorepo.
- LLM provider: DeepSeek in v0.1; Anthropic, OpenAI, and Google adapters scaffolded in
packages/llmfor later. - Status: v0.1 in active scaffolding — core advisor + Moderator persona loop on Expo web. The other four personas, additional providers, and native builds are on the roadmap.
- Repository: github.com/blamechris/explAIn (private).
The product uses LLMs to help users write. This documentation is authored solely by the project owner; no part of it is attributed to an AI.