Skip to content

SuperInstance/makerlog-ai

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

26 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Makerlog.ai

Voice-first development assistant with gamified quota harvesting.

Talk through your ideas all day. Wake up to generated assets.

The Core Loop

┌─────────────────────────────────────────────────────────────────┐
│  DAYTIME: Vibe through ideas (voice → transcript → vectors)    │
│  ──────────────────────────────────────────────────────────────│
│  "I'm thinking about a pixel art style for the game..."        │
│  "Maybe the API should have a /harvest endpoint..."            │
│  "The onboarding flow needs to feel more playful..."           │
├─────────────────────────────────────────────────────────────────┤
│  EVENING: Commit compute (review detected opportunities)       │
│  ──────────────────────────────────────────────────────────────│
│  System detects from your conversations:                       │
│  • 12 pixel art icon variations to generate                    │
│  • API endpoint boilerplate to scaffold                        │
│  • 3 onboarding copy alternatives to write                     │
├─────────────────────────────────────────────────────────────────┤
│  OVERNIGHT: Batch execution (using Cloudflare free tier)       │
├─────────────────────────────────────────────────────────────────┤
│  MORNING: Review generated assets (keep/discard/iterate)       │
└─────────────────────────────────────────────────────────────────┘

Why Voice-First?

Most AI tools require you to stop what you're doing and type a prompt. Makerlog lets you talk while you work—walking, cooking, on a boat, whatever. Your ideas accumulate in a vector database, and the system figures out what to generate overnight.

90% audio-first development. Your phone is for occasional QC and iteration.

Features

Voice Conversation Engine

  • 🎤 Push-to-talk recording — Hold to speak, release to send
  • 📝 Real-time transcription — Cloudflare Whisper
  • 🧠 Contextual responses — AI that remembers your conversation
  • 🔊 Text-to-speech — Hear responses without looking at screen
  • 🔍 Semantic search — Find anything you've ever said (Vectorize)

Opportunity Detection

  • Auto-detect generative tasks — From natural conversation
  • 🎯 Confidence scoring — Only surfaces high-value opportunities
  • ✏️ Refine prompts — Edit before committing to generation
  • 📋 Queue management — Prioritize what matters

Gamified Quota Harvesting

  • 📊 Quota dashboard — Real-time Cloudflare free tier usage
  • 🌾 Harvest mode — Use remaining quota before midnight reset
  • 🏆 Achievements — XP, levels, streaks
  • 🔥 "Use it or lose it" — Alerts when quota is about to reset

Architecture

┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│   React App     │────▶│  Cloudflare     │────▶│   Workers AI    │
│   (Voice UI)    │     │  Worker (API)   │     │                 │
│                 │     │                 │     │  • Whisper STT  │
│ • Push-to-talk  │     │  • Transcribe   │     │  • Llama chat   │
│ • TTS playback  │     │  • Respond      │     │  • BGE embed    │
│ • Opportunities │     │  • Detect opps  │     │  • SDXL images  │
└─────────────────┘     └────────┬────────┘     └─────────────────┘
                                 │
        ┌────────────────────────┼────────────────────────┐
        ▼                        ▼                        ▼
┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│       D1        │     │    Vectorize    │     │       R2        │
│   (Database)    │     │  (Embeddings)   │     │   (Assets)      │
│                 │     │                 │     │                 │
│ • Conversations │     │ • Message       │     │ • Audio files   │
│ • Messages      │     │   embeddings    │     │ • Generated     │
│ • Opportunities │     │ • Semantic      │     │   images        │
│ • Tasks         │     │   search        │     │ • Exports       │
└─────────────────┘     └─────────────────┘     └─────────────────┘

Quick Start

Prerequisites

  • Node.js 20+
  • Cloudflare account (free tier works)
  • Wrangler CLI: npm install -g wrangler

1. Clone & Install

git clone https://github.com/SuperInstance/makerlog-ai.git
cd makerlog-ai
npm install

2. Authenticate with Cloudflare

wrangler login

3. Create Vectorize Index

wrangler vectorize create makerlog-conversations --dimensions=768 --metric=cosine

4. Deploy the API Worker

cd workers/api
npm install
npm run deploy

5. Deploy the Dashboard

cd ../..
npm run build
wrangler pages deploy dist --project-name makerlog

API Reference

Voice Endpoints

Method Endpoint Description
POST /api/voice/transcribe Upload audio, get transcript + AI response

Conversation Endpoints

Method Endpoint Description
GET /api/conversations List all conversations
GET /api/conversations/:id Get conversation with messages
POST /api/conversations Create new conversation
POST /api/search Semantic search across conversations

Opportunity Endpoints

Method Endpoint Description
GET /api/opportunities List detected opportunities
POST /api/opportunities/:id/queue Queue for overnight generation
POST /api/opportunities/:id/reject Skip this opportunity
POST /api/opportunities/:id/refine Edit prompt before queueing

Task & Quota Endpoints

Method Endpoint Description
GET /api/quota Current quota usage
GET /api/tasks List all tasks
POST /api/tasks Queue a task manually
POST /api/harvest Execute all queued tasks
GET /api/digest Daily summary

Cloudflare Resources

Resource Type Purpose
makerlog-db D1 Database Conversations, messages, tasks
makerlog-cache KV Namespace Quota caching, sessions
makerlog-assets R2 Bucket Audio files, generated assets
makerlog-conversations Vectorize Semantic search embeddings

Roadmap

Phase 1: Voice-First MVP ✅

  • Push-to-talk recording
  • Whisper transcription
  • Contextual AI responses
  • Opportunity detection
  • Basic quota tracking

Phase 2: Desktop Connector

  • Local runtime for file access
  • Watch filesystem for code changes
  • Execute generated code locally
  • Sync with cloud state

Phase 3: Multi-Modal

  • Image upload for context
  • Screen sharing for QC
  • Generated asset preview on phone
  • Iteration via voice commands

Philosophy

The best interface is no interface. You shouldn't have to stop thinking to interact with AI. Makerlog is designed for people who have ideas while their hands are busy—fishermen, parents, commuters, anyone who thinks better out loud than typing.

Your free tier quota is a resource. Use it or lose it. Makerlog makes sure every token, every image generation, every API call counts toward something you actually want.

License

MIT


Built for makers who think out loud. 🎤

About

AI tools for tracking and analyzing Makerlog productivity.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors