YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
AURA Concurrent Sessions — Implementation Package
Implementation files and documentation for transforming AURA from a single-mission-at-a-time model to concurrent session-per-input, based on analysis of 5 codebases.
Files
| File | What It Is |
|---|---|
| RESEARCH_ANALYSIS_VERIFICATION.md | Research report — Hermes internals, Cursor-Cats agent pattern, Clicky/OpenClicky UI analysis. Every claim has a file:line citation. |
| AURA_REFACTOR_PLAN.md | Full dev plan — architecture comparison, sprint breakdown, design system spec, open decisions. |
| CONCURRENT_SESSIONS_IMPLEMENTATION.md | 10-step wiring guide — exact code changes to integrate the new files into AURAStore and views. |
| Sources/AURA/Stores/MissionSession.swift | Per-input session model. Each instance owns one aura-hermes Process. Drop into Sources/AURA/Stores/. |
| Sources/AURA/Stores/MissionSessionManager.swift | Session collection manager. spawnSession() returns immediately — no blocking. Drop into Sources/AURA/Stores/. |
Repos Analyzed
- AURA — the app being refactored
- hermes-agent — the AI agent backend AURA invokes
- Cursor-Cats — reference for per-input agent spawning
- clicky — reference for UI polish and design system
- openclicky — reference for multi-session management and agent dock
Key Finding
Hermes is safe to run concurrently. Each aura-hermes chat invocation creates a separate Python process with a unique session_id, independent SQLite WAL writes, and no shared in-process state. The bottleneck is entirely in AURA's Swift layer — a single missionStatus enum and a guard missionStatus != .running check in AURAStore.startMission().
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support