Skip to content

Latest commit

 

History

History
117 lines (80 loc) · 3.13 KB

NOTES.md

File metadata and controls

117 lines (80 loc) · 3.13 KB

Project Notes

Agentic memory for persistent context across sessions. AI agents should update this file with discoveries, decisions, and patterns.

Session Log

2026-01-18 — Convex Migration Completed

  • ✅ Completed migration from Supabase to Convex
  • Cleaned up legacy Supabase type files and deprecated API stubs
  • Updated documentation to reflect Convex patterns
  • User data now sourced from Clerk (auth) + Convex (app data)

2026-01-14 — Project Context Setup

  • Created core AI context files: CLAUDE.md, NOTES.md, spec.md, plan.md
  • Reviewed existing AGENTS.md — comprehensive, includes gold standard examples
  • Project was in active migration: Supabase → Convex (now complete)

Decisions Made

Date Decision Rationale Reference
2026-01 Use Convex over Supabase Built-in RAG, real-time, TypeScript-first .agents/research/tech-stack-evaluation.md
2026-01 Use Clerk for auth Native integration with Convex + Flowglad .agents/research/tech-stack-evaluation.md
2026-01 Use Flowglad for payments Open-source, Clerk-native .agents/research/tech-stack-evaluation.md
2026-01 Testing: critical paths only AI responses non-deterministic AGENTS.md

Patterns Discovered

Optimistic Updates

Location: lib/chat-store/chats/provider.tsx

// Store previous state → update optimistically → rollback on error

Streaming API Response

Location: app/api/chat/route.ts

// Uses Vercel AI SDK StreamingTextResponse

Hook Composition

The chat feature uses composed hooks for separation of concerns:

  • useChatCore — AI SDK integration
  • useChatOperations — CRUD operations
  • useFileUpload — File handling
  • useModel — Model selection

Known Issues

Issue Status Notes
Supabase migration ✅ Complete All data operations now use Convex
Convex schema ✅ Complete See convex/schema.ts
Legacy docs ⚠️ Partial Some docs still reference old patterns

Performance Notes

  • Identify slow API routes
  • Profile chat component rendering
  • Measure context window usage

Security Findings

  • OAuth tokens must be encrypted at rest
  • Never log API keys or session tokens
  • Rate limiting implemented in chat operations

Research To-Do

  • Verify Context7 tool status (may be deprecated)
  • Research MCP security best practices
  • Document sub-agent token economics
  • Test nested CLAUDE.md loading behavior

Useful Commands

# Development
bun run dev                    # Start dev server
bun run typecheck              # Type check specific file: tsc --noEmit src/file.ts

# Debugging
bun run lint -- --fix          # Auto-fix lint issues
bun run build                  # Check for build errors

# Git
git diff --staged              # Review staged changes
git log --oneline -10          # Recent commits

This file is maintained by AI agents. Human review encouraged.