timur/home

Monday morning meeting #1

New issue

Open

opened 2026-03-30 09:13:41 +00:00 by timur · 0 comments

timur commented

2026-03-30 09:13:41 +00:00

Owner

📝 Meeting Notes — Hero System (Proc, Proxy, Scripts, Broker)

1. Core Architecture Clarifications

Service → Actions → Runs → Jobs Model

Services are composed of actions
Running a service:
- Creates a run
- Which creates jobs
Runs can also exist without services (direct execution)

Self-Start Mechanism

Services should self-register via:
binary --start
This:
- Creates services
- Registers jobs in Hero Proc
Not handled by UI or server → handled by the binary itself

2. Key Problems Identified

2.1 Runs / Jobs Visibility Issues

Current UI:
- Runs are unclear / unnamed
- Hard to understand what is executing

Fix needed:

Runs should have:
- Name (preferably service name)
Jobs / runs should be:
- Human-readable
- Traceable

2.2 Cleanup / UX Overload

System becomes overwhelming quickly

Need:

Shortcuts to:
- Clean up runs/jobs
- Hide irrelevant history
- Reset state easily

2.3 Service Lifecycle Consistency

When a service is killed:
- Associated runs should also stop/cleanup

3. Proxy System (Major Upgrade 🚀)

New Proxy Capabilities

Acts like Cloudflare-style gateway
Supports:
- Remote tunnels
- Port forwarding
- Domain mapping
- Local socket mapping
- TCP/UDP handling

Example Flow

Remote ports (e.g. 8443, 5353) forwarded
DNS:
- UDP 53 → TCP forwarding via daemon
Enables:
- Remote → local routing seamlessly

Key Insight

👉 This becomes a universal access layer:

External → internal services
Works over Mycelium
Acts as entry point into node

Status

Early stage (not production ready)
Missing:
- Authentication integration
- Blockchain login
- Access control routing

4. Authentication Direction

Backend Philosophy

❌ No users/groups in backend
✅ Use claims-based model

Implication

Backend stays stateless + simple
Auth handled at:
- Proxy layer
- Entry points

5. Inspector vs Proxy

Inspector

Main tool for:
- Debugging
- Development
- Viewing services/jobs

Proxy

Only needed for:
- External access
- Networking
- Tunnels

👉 Important distinction:

You do NOT need proxy for local dev

6. Major Priority: Dev Environment Automation

Current State

Everything started manually ❌

Target State

Create Hero Scripts (Rhai scripts) to:

Structure

/scripts
/backend
/frontend

Capabilities

Run all services:
run_all
Build everything
Start Hero Proc jobs
Clone repos if needed
Support branch selection

Key Design Principle

Scripts should:
- Use _lib bindings
- Create jobs (not run services directly)
- Be composable

Optional Enhancements

build_all
download_all
Environment bootstrap

7. Deployment Strategy Discussion

Current

Dev mode:
- Build from source
- Git-based

Future

Need:
- Release binaries
- Stable builds

Constraint

Not stable enough yet → stay on:
- development branch builds

Alternatives

Prebuilt image (Mick’s approach)
Direct binary distribution (preferred long-term)

8. Hero Runner / Execution Optimization

New Idea

Lightweight execution layer:
- Run jobs without full runs/services overhead
Possible execution modes:
- Direct (Hero Proc)
- Hero Runner

Goal

Reduce:
- Memory usage
- Execution overhead

9. Broker (Hero Brok) Status

Issues

Not fully stable
Limited real-world testing
Problems with:
- Complex tasks
- Script regeneration

Technical Changes

Removed:
- MCP-on-MCP logic (was unnecessary)
Need to:
- Reintroduce OpenRPC services properly

10. UI / Developer Experience Issues

Problems

API keys unclear
Models unclear
Backend configs confusing
No guidance in UI

Required Fixes

Predefine:
- Providers (OpenRouter, etc.)
Auto-expose:
- Available models based on keys
Improve:
- Naming
- Discoverability
- Defaults

# 📝 Meeting Notes — Hero System (Proc, Proxy, Scripts, Broker) ## 1. Core Architecture Clarifications ### Service → Actions → Runs → Jobs Model * Services are composed of actions * Running a service: * Creates a run * Which creates jobs * Runs can also exist without services (direct execution) ### Self-Start Mechanism * Services should self-register via: binary --start * This: * Creates services * Registers jobs in Hero Proc * Not handled by UI or server → handled by the binary itself --- ## 2. Key Problems Identified ### 2.1 Runs / Jobs Visibility Issues * Current UI: * Runs are unclear / unnamed * Hard to understand what is executing Fix needed: * Runs should have: * Name (preferably service name) * Jobs / runs should be: * Human-readable * Traceable --- ### 2.2 Cleanup / UX Overload * System becomes overwhelming quickly Need: * Shortcuts to: * Clean up runs/jobs * Hide irrelevant history * Reset state easily --- ### 2.3 Service Lifecycle Consistency * When a service is killed: * Associated runs should also stop/cleanup --- ## 3. Proxy System (Major Upgrade 🚀) ### New Proxy Capabilities * Acts like Cloudflare-style gateway * Supports: * Remote tunnels * Port forwarding * Domain mapping * Local socket mapping * TCP/UDP handling ### Example Flow * Remote ports (e.g. 8443, 5353) forwarded * DNS: * UDP 53 → TCP forwarding via daemon * Enables: * Remote → local routing seamlessly ### Key Insight 👉 This becomes a universal access layer: * External → internal services * Works over Mycelium * Acts as entry point into node --- ### Status * Early stage (not production ready) * Missing: * Authentication integration * Blockchain login * Access control routing --- ## 4. Authentication Direction ### Backend Philosophy * ❌ No users/groups in backend * ✅ Use claims-based model ### Implication * Backend stays stateless + simple * Auth handled at: * Proxy layer * Entry points --- ## 5. Inspector vs Proxy ### Inspector * Main tool for: * Debugging * Development * Viewing services/jobs ### Proxy * Only needed for: * External access * Networking * Tunnels 👉 Important distinction: * You do NOT need proxy for local dev --- ## 6. Major Priority: Dev Environment Automation ### Current State * Everything started manually ❌ ### Target State Create Hero Scripts (Rhai scripts) to: #### Structure /scripts /backend /frontend #### Capabilities * Run all services: run_all * Build everything * Start Hero Proc jobs * Clone repos if needed * Support branch selection --- ### Key Design Principle * Scripts should: * Use _lib bindings * Create jobs (not run services directly) * Be composable --- ### Optional Enhancements * build_all * download_all * Environment bootstrap --- ## 7. Deployment Strategy Discussion ### Current * Dev mode: * Build from source * Git-based ### Future * Need: * Release binaries * Stable builds ### Constraint * Not stable enough yet → stay on: * development branch builds --- ### Alternatives * Prebuilt image (Mick’s approach) * Direct binary distribution (preferred long-term) --- ## 8. Hero Runner / Execution Optimization ### New Idea * Lightweight execution layer: * Run jobs without full runs/services overhead * Possible execution modes: * Direct (Hero Proc) * Hero Runner ### Goal * Reduce: * Memory usage * Execution overhead --- ## 9. Broker (Hero Brok) Status ### Issues * Not fully stable * Limited real-world testing * Problems with: * Complex tasks * Script regeneration ### Technical Changes * Removed: * MCP-on-MCP logic (was unnecessary) * Need to: * Reintroduce OpenRPC services properly --- ## 10. UI / Developer Experience Issues ### Problems * API keys unclear * Models unclear * Backend configs confusing * No guidance in UI ### Required Fixes * Predefine: * Providers (OpenRouter, etc.) * Auto-expose: * Available models based on keys * Improve: * Naming * Discoverability * Defaults ---

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

timur/home#1

No description provided.

Rows
Columns