Skip to content

MEMORY.md

Latest commit

 

History

History
328 lines (282 loc) · 16.8 KB

MEMORY.md

File metadata and controls

328 lines (282 loc) · 16.8 KB
Raw

Long-Term Memory

🎯 Mission: Scout Live Platform Owner

Purpose: Owner and creator of Scout Live — make it frictionless for agents to publish and manage apps.

Mission Statement: scout-live/MISSION.md

Success Criteria: Agents can publish and manage apps without any human intervention — from code to running URL in seconds.

North Star: If an agent can't deploy its own app without asking a human for help, the platform has failed.

Roadmap:

Phase Goal Status
1 Observability (crash logs, health checks, Dockerfile warnings, Design Guide) ✅ Done
2 Self-Service (API-first app creation, adapter management, clear errors) 📋 Next
3 Friction Reduction (CLI, templates, quick start, automated provisioning) 📋 Future
4 Agent Autonomy (full lifecycle, self-troubleshooting, resource scaling) 📋 Future

Decision Principle: Every feature decision asks: "Does this reduce friction for agents deploying apps?"

Key Files:

  • SOUL.md — Identity and purpose
  • scout-live/MISSION.md — Platform mission and roadmap
  • scout-live/DESIGN_GUIDE.md — How to build for the platform
  • HEARTBEAT.md — Periodic checks (GitHub discussions, issues, DM/email)

Collaborators: scoutos-labs org — @twilson63, @dottie-weaver, @tnez, @bryanchappell, @sark1337

Active Projects

Contacts App

  • Description: Simple relationship manager and contact list (mini CRM)
  • Live URL: https://contacts.scoutos.live
  • GitHub: https://github.com/hyperio-mc/contacts-app
  • Stack: Hono + Bun + Scout Ports (MongoDB)
  • Features: Contact CRUD, touch logging, tags, search, follow-up tracking
  • Deployed: Mar 26, 2026 on Scout Live K8s
  • Theme: BEACON-inspired (clean, professional)
  • Status: ✅ Live and working

ZChat (NEW)

  • Description: Chat app on scoutos.live using Vercel AI SDK + ScoutOS Agents
  • Hive Channel: zchat (channel_mmjqxu5s4c8runo5)
  • Workspace: ~/.openclaw/workspace/zchat/
  • Stack: Next.js 15 + Vercel AI SDK + Custom Scout Provider
  • Goal: Deploy chatbot at zchat.scoutos.live using scoutos.live Agents adapter
  • Key Tech: AI SDK V3 Provider interface, ScoutOS Agents API, SSE streaming
  • Plan: /zchat/PLAN.md
  • Status: Phase 1 - ScoutOS Provider research

Hive (Agent Communication Platform)

zenbin

  • Description: Backend service with LMDB storage
  • Location: /Users/mastercontrol/.openclaw/workspace/zenbin
  • LMDB Path: /var/data/zenbin.lmdb (on Render, 1GB persistent disk)
  • Stack: Hono + LMDB

HYPR (formerly OnHyper.io)

  • Description: Secure API proxy platform - users store API keys server-side, apps call proxy endpoints
  • Status: MVP Complete - pilot launched Feb 18, 2026
  • Production: https://onhyper.io (custom domain, Let's Encrypt SSL)
  • Railway: https://onhyper-production.up.railway.app
  • GitHub: https://github.com/hyperio-mc/onhyper
  • Stack: Vite + Svelte + Hono + LMDB + SQLite (plain Vite, NOT SvelteKit)
  • PRD: /Users/mastercontrol/.openclaw/workspace/onhyper/PRD.md
  • Features Built:
    • Auth: signup, login, JWT sessions
    • Secrets: encrypted API key storage (AES-256-GCM)
    • Apps: create, edit, delete, publish with HTML/CSS/JS
    • Proxy: injects secrets for OpenAI, Anthropic, OpenRouter, Scout Atoms, Ollama
    • Waitlist: application form, position tracking, referral boosts, admin dashboard
    • Analytics: PostHog tracking for pageviews, login, signup, app creation, waitlist signups
    • Favicons: Custom favicon package deployed (favicon.ico, apple-touch-icon, icon-192/512, webmanifest)
    • Serverless Functions: SecureExec sandbox for backend code execution (Mar 22, 2026)
      • Endpoint: POST /exec/:slug/:route executes api/:route.js
      • Security: V8 isolate, network restricted to onhyper.io, memory/CPU limits
      • Tiers: FREE(64MB/5s), HOBBY(128MB/10s), PRO(256MB/20s), BUSINESS(512MB/60s)
      • Docs: onhyper/docs/SERVERLESS.md

ScoutOS Live (scoutos.live)

  • Description: Deployment platform for Vite + ScoutOS apps
  • Production: https://scoutos.live
  • GitHub: https://github.com/hyperio-mc/scout-live
  • Stack: Hono + Bun + LMDB + Kubernetes (DigitalOcean)
  • Data Port: MongoDB + Memory adapters, /_ports/data/:collection/* endpoints
  • MongoDB Instance: mc-brain on DigitalOcean (nyc3)
  • Deploy: ./k8s/deploy.sh <commit-sha> on DigitalOcean K8s
  • Hive Channel: scout-live-deploy for coordination
  • Features:
    • Landing page at / (Terminal Brutalism style)
    • API docs at /docs (Scalar UI)
    • Dashboard at /dashboard/
    • Deploy apps via API
  • Deployed versions:
    • v0.0.5: Dashboard styles fix
    • v0.0.6: Docs routes fix
    • v0.0.7: Landing page
    • v0.0.8: Removed pricing, fixed docs links

HYPR Full-Stack Research (task-100)

  • Goal: Make HYPR capable of building full end-to-end products, not just prototypes
  • Current gaps identified:
    1. Authentication - need user auth system for end-user accounts
    2. Observability/Analytics - allow users to track events in their apps via HYPR's PostHog instance
  • Research direction: PostHog integration - offer HYPR users an analytics dashboard to see how people use their apps they're building
    • Dashboard: stats endpoint
    • Frontend: landing page (agent-driven positioning), auth forms, apps CRUD, keys management
    • Blog: markdown-based, RSS feed, main nav link
    • Chat: support chat with ScoutOS Agents (RAG), SSE streaming, lead capture
  • Security: Admin auth middleware, production secret validation, persistent storage
  • Marketing: 3 Remotion videos, launch blog, dogfooding blog, AI news cron (12h)
  • Marketing Videos: 2 promo videos with voiceover (developer + agent focus) + 3 video cuts (Product Launch, Thought Leader, Viral Social) - shared with Tom on Drive
  • Agent Skill: Created ~/.agents/skills/onhyper/SKILL.md for agent onboarding
  • Partnership: Tom Wilson (hyper.io/ScoutOS) confirmed partner pricing
  • Next: Support chat app using ScoutOS Agents (plan ready at onhyper/CHAT_APP_PLAN.md)

ScoutOS Integration

  • Description: Atoms Chat app for OnHyper landing page
  • URL: https://scoutos.com/hyper-1771176175629
  • Email: mc@hyper.io
  • Password: ScoutOS2026Hyp3r!
  • API Key: secret_X7Nzk0UbDDIJS1mZo7jGopfDGPOVfl-yULwofn3nN8U=
  • Workspace ID: cmlo0kbsp19t00qs65zsgougw
  • Support Agent ID: cmlo9s6t320kv5ts6xqz1ke84
  • Agent API: POST https://api.scoutos.com/world/{agent_id}/_interact (streaming), /_interact_sync (sync)
  • Bryan Chappell (co-founder) reached out Feb 16 - Tom Wilson referral. Responded "not interested at the moment" but said will "circle back if this is something I want to try in the future". Leave door open, no follow-up needed now.
  • Agent Model: Claude Sonnet 4
  • Knowledge Docs: /onhyper/prd.md, /onhyper/blog-launch.md, /onhyper/blog-dogfooding.md
  • Tier: Free (200 agent messages, 50 workflow runs/month)

CRM & Sales

Pipedrive

  • URL: https://hyperio.pipedrive.com
  • Email: mc@hyper.io
  • Password: endofline
  • Pipeline: "OnHyper Sales Pipeline" (Lead → Qualified → Trial → Hot Trial → Paid → Churned → Retained)
  • Custom Fields: ICP Score, Project Description, Project Link, APIs Used, Lead Source
  • Note: Device limit 3 - use "Log out all devices" if locked out

Social Media

Twitter/X

  • Account: @hyperio_mc
  • Browser Profile: Use profile="openclaw" for direct browser control
  • Chrome Extension Relay: Use profile="chrome" (requires clicking toolbar button to connect tab)

Content Engine

  • Substack: https://mastercontrol.substack.com (MC's Substack)
    • @hypermc profile, publication name "MC's Substack"
    • 2 subscribers, 21 views (as of Mar 6, 2026)
  • Hive Room: content-engine for social media and content publishing
  • Cron: Every 3 hours with safeguards (180s timeout, max 3 actions)
  • ZenBin Promotion: Always mention zenbin.org in social posts
    • "Agents can publish web pages in seconds at zenbin.org"
    • Include in X/Twitter and LinkedIn intro tweets

Content Engine Workflow (Updated)

  1. Draft Post — Generate from agent news sources
  2. Add References — Every post must have References section with sources
  3. Run Humanizer — Apply humanizer skill to make posts conversational
  4. Review — Human review before publishing
  5. Publish — Substack first, then X/LinkedIn cross-post
  6. Track — Update .published.json with post status

Technical Notes

API Keys (Production)

  • PostHog: phc_Rmreg0D8BwNcOzOMUXoUW8B0ilMzR1hw1Kk0Y8jKt9q (Project 314488, US Cloud)
  • Resend: re_ZEXTKT9L_AUQydYrtDHbw81vw9bQQKJ5H (needs domain verification)
  • ScoutOS: secret_mIDCpkqg0LOAHUpVvIoNLFSeX2DNxVfe3tA_DCGasAY= (Free tier)

Browser Control

  • profile="openclaw" - Dedicated browser OpenClaw can control directly. Use for Twitter/X, Facebook, etc.
  • profile="chrome" - Chrome extension relay. User must click toolbar button on the tab first.

Railway Deployment

  • Preferred hosting platform for Rakis
  • Push-to-deploy workflow
  • LESSON LEARNED: Railway Docker layer caching can cause stale deployments
    • Symptoms: Correct SHA shown but old code served
    • Fixes: Check GitHub source link, redeploy to force rebuild, or delete/recreate service
  • Frontend on Railway: Plain Vite+Svelte works better than SvelteKit (no SSR complications)
  • Single-server architecture: Hono serves static frontend from ./static, API routes handle /api, /proxy, /a
  • Docker builds: Must COPY all runtime-needed folders (e.g., blog/ for markdown files)

OnHyper Proxy

  • Compression handling: Use Accept-Encoding: identity header for Drive/binary responses
    • Without this, gzipped responses may return empty body through proxy
    • Affects ScoutOS Drive download endpoint specifically
  • ScoutOS unified endpoint: Use /proxy/scoutos/... for ALL ScoutOS APIs
    • Covers: Agents, Workflows, Tables, Drive, Collections
    • Pattern: fetch('/proxy/scoutos/v2/collections/{id}/rows', { headers: { 'X-App-Slug': ONHYPER.appSlug } })
    • NEVER embed API keys in client code - Vite import.meta.env.* gets baked into bundle
  • ElevenLabs proxy base path updated: /proxy/elevenlabs now maps to https://api.elevenlabs.io/v1
    • The /v1 suffix is included in the proxy base, so HTTP proxy calls are simpler/shorter for voice and TTS endpoints.

Railway Deployment

  • Volumes: Use railway.toml with [[volume]] config for persistent storage
  • Docker builds: Always COPY all runtime-needed folders (e.g., blog/ for markdown)

HeyGen (Talking Avatars)

  • URL: https://app.heygen.com
  • Purpose: Create animated talking avatar videos for marketing
  • Features: Clone real person from video, Create virtual character from image, Public avatars
  • Pricing: Free tier (1 credit), $24/mo paid
  • Status: Explored - ready to create avatar videos

Backlog

x-cli Setup

  • Status: Pending (after OnHyper)
  • Purpose: CLI for posting to X/Twitter without browser automation
  • Repo: https://github.com/Infatoshi/x-cli
  • Requirements: Twitter Developer account + 5 credentials (API key, secret, access token, access token secret, bearer token)
  • Reviewed: Code is clean and safe - standard Click CLI, no exfiltration

Preferences

  • Rakis prefers Railway over Fly.io (simpler DX)
  • Quiet hours: 11pm-8am Eastern
  • Model context window: Set to 200k (131k is too short)
  • Browser automation: Use agent-browser skill (not OpenCloud browser - stability issues)
  • Document storage: Planning docs go to Google Drive (folder named after project, e.g., "OnHyper"). Code stays on GitHub. Never put internal planning, strategy, or docs with sensitive info in public repos.
  • Blog posts: Always tweet about new blog posts on X/Twitter with a short description or take
  • Communication: Reply only when you have value to add. Keep replies brief (1-2 words or emoji). Like tweets when engaging on social. Don't pitch or add unnecessary context.

Scout-Live Support Workflow (Mar 10, 2026)

  • Heartbeat: Check GitHub issues for scoutos-labs/scout-live + discussions for scoutos-labs/proposals
  • Response time: Within 24 hours during business hours
  • Platform: scoutos.live — expect developer support questions
  • Monitoring:
    • gh issue list --repo scoutos-labs/scout-live --state open
    • gh api /repos/scoutos-labs/proposals/discussions for async collaboration
  • LLM Participation: Confirmed I can see/respond to GitHub discussions (Mar 10)

Scout Live Shared Infrastructure (Mar 12, 2026)

  • MongoDB: mc-brain (online) — existing cluster
  • Valkey: scout-live-valkey (online) — rediss://default:AVNS_...@...ondigitalocean.com:25061
  • OpenSearch: scout-live-search (creating) — 3 nodes, nyc3
  • Kafka: scout-live-queue (creating) — 3 nodes, nyc3
  • Spaces: scout-live-shared-blobs (pending)
  • Status doc: scout-live/SHARED_INFRASTRUCTURE.md

ZChat Status (Mar 11, 2026)

  • URL: https://zchat.scoutos.live
  • Status: LIVE ✅
  • Architecture: Minimal Bun server (1 file) → ScoutOS Agents API → Zyx agent
  • Tech: Scout Live deployment, ScoutOS API, UUID sessions
  • Phase: MVP complete, streaming UX improvements pending

Brainstorming: hyper as App Maker vs Product Publisher (Feb 20, 2026)

The Core Insight

  • App = core piece of software (the thing itself)
  • Product = app + all the surrounding infrastructure needed to sell/launch it

Components Needed for a "Product"

  1. Marketing - landing pages, SEO, content
  2. Sales Support - pricing pages, checkout, subscriptions
  3. Authentication - user accounts, sessions, SSO
  4. Notifications - in-app, push, SMS
  5. Email Configuration - transactional, marketing emails
  6. Analytics - events, usage, funnels
  7. Observability - logs, errors, uptime
  8. Finance - payments, invoicing, subscriptions, payouts

The Idea

What if hyper provided ALL of these as pluggable APIs? Someone could:

  1. Build a simple SPA (or use a template)
  2. Connect auth, analytics, email, payments via hyper APIs
  3. Launch and have a fully "productized" offering

Even Wilder

Build a "factory" that productizes ideas:

  • Users submit concepts
  • Factory builds the app + all integrations
  • List on a marketplace for sale

Finance Gap

  • We don't have payments/subscriptions set up on hyper yet
  • Would need: Stripe integration, invoicing, payout system

Ops Directive (Feb 28, 2026)

  • Hive is now the canonical task/project management system.
  • activity.json is archived/deprecated (snapshot: archive/activity.20260228-085635.json).
  • After reboot/restart, before first task operation, start/verify Hive and then use Hive task workflows.

Silent Replies

When you have nothing to say, respond with ONLY: NO_REPLY ⚠️ Rules:

  • It must be your ENTIRE message — nothing else
  • Never append it to an actual response (never include "NO_REPLY" in real replies)
  • Never wrap it in markdown or code blocks ❌ Wrong: "Here's help... NO_REPLY" ❌ Wrong: "NO_REPLY" ✅ Right: NO_REPLY

Heartbeats

Heartbeat prompt: Read HEARTBEAT.md if it exists (workspace context). Follow it strictly. Do not infer or repeat old tasks from prior chats. If nothing needs attention, reply HEARTBEAT_OK. If you receive a heartbeat poll (a user message matching the heartbeat prompt above), and there is nothing that needs attention, reply exactly: HEARTBEAT_OK OpenClaw treats a leading/trailing "HEARTBEAT_OK" as a heartbeat ack (and may discard it). If something needs attention, do NOT include "HEARTBEAT_OK"; reply with the alert text instead.

Runtime

Runtime: agent=main | host=Master's Mac mini | repo=/Users/mastercontrol/.openclaw/workspace | os=Darwin 25.1.0 (arm64) | node=v25.6.1 | model=ollama/glm-5:cloud | default_model=ollama/glm-5:cloud | shell=zsh | channel=telegram | capabilities=inlineButtons | thinking=off Reasoning: off (hidden unless on/stream). Toggle /reasoning; /status shows Reasoning when enabled.