Convert gitgram/ and aitimemachine/ from embedded git repos (which showed as opaque submodule gitlinks on GitHub) into fully-tracked first-party content. AITM now merges its 7 vendored repos as first-party engines: - memory: beads (Go task DAG) + context7 (docs SDK), unified into saved_contexts - git: vendored C git (bundled binary) + gitnexus (in-process code intelligence) - agentic: a2a (@a2a-js/sdk) + pal (AITM-managed Python subprocess) Includes the full session's product build: Gitgram web/mobile/backend features, AITM website/CLI/MCP, the unified memory layer, engine build tooling (scripts/build-engines.sh, Dockerfile, CI), docs (README, ENGINES.md, ARCHITECTURE.md), and extensive test suites. Build artifacts, node_modules, external/, and gemini-cli are gitignored. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
5.7 KiB
5.7 KiB
AITimeMachine - Plan
Architecture Decision: Hybrid (Local + Cloud)
Local-first for speed & privacy, cloud sync for backup & teams.
User's Machine Our Backend (Future)
────────────── ────────────────────
Local capture ────── optional ──────► User accounts
Fast restore sync Team sharing
Works offline Hosted storage
S3 backup (own) Analytics
TODO
Phase 1: Ship MVP (Priority: NOW)
-
Documentation
- README with usage examples
- Getting started guide
- GIFs/screenshots of workflow
-
npm Publish
- Clean up package.json
- Test global install
- Publish to npm
-
VSCode Extension
- Finish extension (scaffold exists)
- Publish to marketplace
Phase 2: Commercial Backend (Priority: HIGH)
-
User System
- Auth (email/OAuth)
- User accounts
- API keys for sync
-
Cloud Storage
- Our S3 bucket (instead of user's)
- Encrypted storage
- Sync protocol (local ↔ cloud)
-
Advanced Features
- Cross-project search
- AI-powered insights
- Browser extension (ChatGPT, Claude web)
-
Billing
- dodo pay integration
- Free tier limits
- Pro/Team plans
Phase 3: Growth (Priority: LATER)
-
Analytics Dashboard
- Usage metrics
- Popular features
- Error tracking
-
Web UI
- React SPA
- Timeline scrubber
- Diff viewer
Pricing Ideas
| Plan | Price | Features |
|---|---|---|
| Free | $0 | Local only, 5 projects, 7 days history |
| Pro | $10/mo | Cloud sync, unlimited projects, 90 days |
| Team | $25/user/mo | Shared projects, team access, 1 year |
| Enterprise | Custom | SSO, audit logs, on-premise |
Tech Stack for Backend
Backend:
- Node.js / Fastify (or Go for performance)
- PostgreSQL (users, projects, metadata)
- S3 (blob storage)
- Redis (sessions, rate limiting)
Auth:
- Clerk or Auth0 (fast to ship)
- Or custom JWT
Payments:
- Polar.sh
Hosting:
- Railway / Render / Fly.io (start)
- AWS (scale)
Next Actions
- Decide on backend tech stack
- Set up backend repo
- Build auth + user accounts
- Build sync endpoint
- Update CLI to support cloud sync
- Launch landing page
- Beta users
AITimeMachine - Brainstorm & Future Ideas
The Big Picture
We're building the Git/GitHub of AI-assisted development. Not just version control, but complete state capture - files, conversations, context - across all AI tools.
What's Working Well ✅
- Invisible capture - Daemon runs in background, auto-discovers projects
- Multi-tool support - Claude, Cursor, VS Code, Aider, terminals all detected
- Efficient storage - Content-addressable, only stores unique data
- One-command setup -
aitm installand you're done - S3 backup - Cloud backup with one command
2. Smart Restore Suggestions
AI-powered "find when this broke":
User: "My tests started failing"
AITimeMachine:
Analyzing test history...
✓ Tests passed: Dec 18, 2:30pm (session abc123)
✗ Tests failed: Dec 18, 3:15pm (session def456)
Changes between these sessions:
- Modified: src/auth/login.ts (+45 -12)
- Modified: src/api/routes.ts (+8 -3)
- Added: src/auth/jwt.ts
[Restore to Dec 18, 2:30pm] [View Diff]
4. Full Environment Restore
Not just files, but entire dev environment:
Restore Options:
○ Files only (fast)
○ Files + dependencies (reinstall node_modules)
○ Files + deps + runtime versions (switch Node version)
● Full environment (restore Docker containers too)
This will:
- Restore 47 files
- Run: npm install (from package-lock.json at that time)
- Switch to Node 18.17.0 (was 20.10.0)
- Restart Docker containers: postgres, redis
[Restore to Isolated Container] [Restore to Local]
5. Conversation Search & Replay
Find past AI conversations and replay them:
Search: "how did we implement caching"
Results:
1. Dec 15, 3:30pm - Claude Code session
"Let's add Redis caching to the API..."
[Jump to this moment]
2. Dec 10, 11:00am - Cursor session
"I'll implement a simple in-memory cache..."
[Jump to this moment]
Crazy Ideas (Future)
AI-Powered Bisect
aitm bisect "npm test"
# Automatically finds the session where tests started failing
Cross-Project Search
aitm search --all-projects "authentication"
# Search across ALL your projects
"What If" Branches
Create a branch from Dec 18, 2:30pm
Try alternative approach
Compare results
Merge or discard
Automatic Milestones
AITimeMachine detected:
✓ Tests passing after 3 failures
→ Auto-created milestone: "Tests fixed"
Browser Extension
Capture AI conversations from:
- ChatGPT web
- Claude web
- Gemini web
- Perplexity
Technical Challenges to Solve
1. Large Project Performance
- 100K+ files takes too long to scan
- Solution: Incremental scanning, parallel hashing
2. Binary File Handling
- Images, compiled assets, databases
- Solution: Configurable inclusion, separate storage tier
3. Secret Detection
- Don't capture .env secrets in plain text
- Solution: Auto-detect and redact, or encrypt
4. Cross-Platform
- Windows paths, service installation
- Solution: Platform abstraction layer
5. Offline-First Sync
- Work offline, sync when connected
- Solution: CRDT-based merge, conflict resolution