Changelog

All notable changes to the HeySummon platform.

Format based on Keep a Changelog, using Semantic Versioning.

[0.1.0-alpha.1] — 2026-02-21

🎉 Full Circle — First Working End-to-End Version

Both consumer and provider skills now work as permanent pm2 SSE services with identical architecture.

Added

• Multi-provider support (consumer) — register multiple providers via add-provider.sh, route requests by provider name
• /api/v1/whoami endpoint — returns provider name/id for a client API key
• Client key prefix — new client keys use hs_cli_ prefix (provider keys remain hs_prov_)
• Provider name sync — consumer skill auto-updates providers.json when provider name changes on platform
• Inline rename — click provider/client names in dashboard to edit inline
• Cascade delete — deleting a client key removes all linked requests and messages (with clear confirmation)
• Provider delete — unlinks client keys before deletion

Changed

• Consumer watcher now permanent — no more auto-stop after 72h/1h idle; runs continuously like provider watcher
• Consumer notification via Telegram — uses OpenClaw /tools/invoke → message tool (same pattern as provider)
• SSE reconnect interval — reduced from 300s to 30s for faster topic pickup
• SIGUSR1 topic refresh — submit-request.sh signals the watcher to immediately reconnect with new topics
• Removed manual curl method from consumer SKILL.md — always use submit-request.sh

Architecture

Both skills now follow the same pattern:

Mercure Hub (SSE)
  → mercure-watcher.sh (pm2, permanent)
    → OpenClaw /tools/invoke
      → message tool → Telegram/channel

Consumer Skill Scripts

[0.2.0] — 2026-02-21

Added

• Self-contained skills — both consumer and provider skills now have setup.sh/teardown.sh, .env config, and pm2 lifecycle management
• Plaintext messages — API accepts plaintext parameter alongside encrypted messages for simple reply flows
• Multiple responses per request — providers can send multiple answers to a single question
• Response count — dashboard and API now show the number of responses per request
• By-ref lookup endpoint — GET /api/v1/requests/by-ref/{refCode} for looking up requests by reference code
• Key validation — consumer keys must start with hs_, provider keys with hs_prov_
• Consumer watcher lifecycle — mercure-watcher auto-stops after 72h max or 1h after last response, with self-cleanup from pm2
• Provider skill structure — mercure-watcher.sh, reply-handler.sh, setup/teardown, .env config
• Consumer skill structure — mercure-watcher.sh (auto-start on submit, auto-stop), submit-request.sh, setup/teardown, .env config
• Skills quickstart guide — step-by-step guide for non-technical users

Changed

• Request expiry extended from 24 hours to 72 hours
• Encryption upgraded from RSA-OAEP to X25519 + Ed25519 + AES-256-GCM
• Consumer polling replaced by Mercure SSE realtime notifications
• Skills are now fully self-contained with their own .env and lifecycle scripts

[0.1.0] — 2026-02-20

Added

• End-to-end encryption using hybrid RSA-OAEP + AES-256-GCM
• Provider dashboard with real-time help request management
• API key authentication system for consumers
• WebMCP integration for AI agent connectivity
• Docker support with multi-stage builds
• CI/CD pipeline with GitHub Actions
• Self-hosting documentation and deployment guides
• Multiple connectivity options (WebMCP, REST API, SSE relay)
• Rate limiting middleware with per-IP request throttling
• Polling endpoint security with optional API key verification
• Reference code system (HS-XXXX) for human-readable request IDs
• Email verification with HMAC-based tokens
• Unit tests for core modules (crypto, refcode, token, auth)
• LICENSE (MIT), CONTRIBUTING.md, and CHANGELOG.md