Execution Blueprint — Confidant
The coordinate map every other document references. Confidant is an AI chat product, so the phases stand up accounts and billing first, then the streaming chat loop, then the AI layer (model choice, thinking, cost controls), then the privacy and personalization that make it yours.
P1 — Foundation & Accounts
- {P1S1} Project scaffold, CI, environments
- {P1S2} Authentication and sessions
- {P1S3} User profile and per-user data isolation
- {P1S4} Subscription billing (Stripe)
P2 — Chat Core
- {P2S1} Conversation + message data model
- {P2S2} Streaming chat loop against the Claude API
- {P2S3} Chat UI: message thread + composer
- {P2S4} Conversation history sidebar
P3 — AI Layer
- {P3S1} Model selection (fast vs. most capable) and system prompt
- {P3S2} Streaming responses with adaptive thinking
- {P3S3} Token metering, usage limits, and prompt caching for cost
- {P3S4} Regenerate / stop / edit-and-resend
P4 — Privacy & Personalization
- {P4S1} Private-by-default storage and data controls (export / delete)
- {P4S2} Custom instructions and saved personas
- {P4S3} Full-text conversation search
- {P4S4} Attachments (paste an image or PDF to ask about it)
P5 — Polish & Launch
- {P5S1} Settings and account management
- {P5S2} Usage dashboard
- {P5S3} Launch hardening: rate limits, abuse controls, observability