Feature/wsjtx sse enrichment

[ceotjoe]

What does this PR do?

Summary

This PR moves the WSJTX decode enrichment intelligence from the OHC server into the local rig-bridge process, so SSE consumers receive fully enriched events even when no server relay is configured.

• Local enrichment library (rig-bridge/lib/wsjtx-enrich.js): Maidenhead grid → lat/lon, Hz → band name, FT8/FT4 message parser (CQ/QSO type, caller, modifier, embedded grid), per-client grid cache (2 h TTL), and full enrichment helpers for DECODE, STATUS, QSO_LOGGED, and WSPR_DECODE messages.
• HamQTH callsign lookup (Phase 5, opt-in): fire-and-forget background resolution of callsign → lat/lon via hamqth.com/dxcc.php; 24 h cache, max 5 concurrent requests. On resolution a decode-update SSE event is emitted so the frontend can live-patch map pins already on screen. Toggled via a new checkbox in the rig-bridge Integrations tab.
• End-to-end wiring (rig-bridge.js): decode, status, qso, clear, wspr, and decode-update bus events are now all forwarded to SSE with their full enriched field sets (previously only a trimmed subset was forwarded, silently dropping all enrichment).
• Frontend (useWSJTX.js): handles new SSE event types — clear (per-client filtering), wspr (capped ring buffer), decode-update (async HamQTH patch). Band-change detection uses the bandChanged flag from enriched status; falls back to manual comparison on the server/relay path.
• Bug fix: wsjtx-enrich was emitting lowercase 'cq'/'qso' type strings; changed to uppercase 'CQ'/'QSO' to match the server convention and fix CQ filter + row colouring in PSKReporterPanel.
• Bug fix: ModernLayout was not passing wsjtxWspr to PSKReporterPanel — the WSPR tab now receives data in that layout.
• Enrichment on the relay path is unchanged — the rig-bridge relay queue still sends raw messages so the server can do its own enrichment without double-processing.

Test plan

• Run rig-bridge in SSE-only mode (no server relay): confirm DECODE events in PSKReporterPanel show band, grid, lat/lon, CQ/QSO type and row colour
• Enable HamQTH callsign lookup in the Integrations tab: confirm callsigns without an embedded grid eventually get lat/lon resolved and map pins appear
• Confirm WSPR tab populates in ModernLayout (was broken before this PR)
• Confirm CLEAR from WSJT-X wipes only the correct client's decodes
• Run rig-bridge with a server relay configured: confirm server-side enrichment still works and no fields are duplicated or corrupted
• Check rig-bridge status endpoint reports gridCacheSize, callsignCacheSize, and hamqthLookup flag correctly
• Verify MSHV / JTDX / JS8Call (digital-mode-base.js) receive the same enrichment as WSJT-X

Checklist

• App loads without console errors
• Tested in Dark, Light, and Retro themes
• Responsive at different screen sizes (desktop + mobile)
• If touching server.js: caches have TTLs and size caps (we serve 2,000+ concurrent users)
• If adding an API route: includes caching and error handling
• If adding a panel: wired into Modern, Classic, and Dockable layouts
• No hardcoded colors — uses CSS variables (var(--accent-cyan), etc.)
• No .bak, .old, console.log debug lines, or test scripts included

[alanhargreaves]

I'm happy that it does what it says it does. I'm not seeing any errors from the rig-bridge node.js nor extras from the clients.

I checked out the debug stuff that was in there, and the api that had some stats on port 5555. Also played with adding in the gridSource to the spots on WorldMap to verify that we were getting grid locs from different places.

I feel I was a bit close to this one to approve it, and I'd also be more comfortable with you doing the best-practice javascript once over, given there is some claude code in there.

[ceotjoe]

@accius if you don't like the .gitignore leave it. I just want to avoid that this thing (Claude.md) get's into the repo accidentally.

[accius]

Review — WSJT-X SSE enrichment

Really solid work. Local enrichment makes the SSE-only mode genuinely useful, and the separation (raw → server relay, enriched → bus) is clean. Good test plan too. A handful of issues to fix before merge:

Bugs

1. Duplicate type key in rig-bridge.js decode forwarder:

   type: msg.message.startsWith('CQ') ? 'CQ' : 'QSO',
   ...
   type: msg.type,

   Both keys are on the same object literal — the second overwrites the first and the first line is dead (and throws TypeError if msg.message is ever undefined, which is guarded elsewhere with ?? ''). Remove the first.

2. clientStates[msg.id]?.deCall / deGrid never populated. In both digital-mode-base.js and wsjtx-relay.js, clientStates[msg.id] is set to { band, dialFrequency, mode } only. The QSO_LOGGED handler falls back to clientStates[msg.id]?.deCall || null, which is always undefined → null. Either store deCall/deGrid from STATUS into clientStates, or drop the fallback and accept myCall/myGrid being null when WSJT-X doesn't supply them.

3. isGrid case-sensitivity bug: the regex /^[A-R]{2}\d{2}(?:[A-Xa-x]{2})?$/ has no i flag and is tested against the original input (not the uppercased g). A 4-char lowercase grid (em28) returns false. FT8 messages are uppercase in practice, so low-impact, but trivial to fix — test against g, or add the i flag.

4. pluginBus.on('wspr', …) forwards msg directly as data, including the internal source field and no field whitelisting — every other handler carefully selects fields. Be consistent, and be careful not to leak fields the UI shouldn't see.

Safety / robustness

5. HamQTH XML parsed by regex (<lat>([^<]+)<\/lat>). Works for the current endpoint but brittle against whitespace/attributes. Consider a minimal XML parser or at least matching against <lat[^>]*>(-?[0-9.]+)</lat> to constrain to numerics.

6. HamQTH public dxcc.php — confirm this endpoint permits unauthenticated automated use at our expected volume (5 concurrent × many users). If ToS requires a session token, this breaks silently. Worth documenting in the README.

7. No persistent HamQTH cache. Every rig-bridge restart re-queries — not a correctness issue, but if we're asking users to enable this across the fleet we may want to persist a JSON cache file.

8. Rate limiting only via HAMQTH_MAX_CONCURRENT = 5. There is no overall QPS limit; on a busy band a burst of unknown calls could trigger dozens of requests per minute. Consider a short (~2s) per-callsign cool-down plus a global QPS cap.

9. SEEN_DECODE_MAX = 2000 — reasonable, but check memory behavior on a very active 20m FT8 session over 24h. Should be fine with FIFO eviction.

Minor

10. .gitignore adds CLAUDE.md — harmless, but unrelated to this PR's topic. Happy to leave it.
11. dt is stored as a pre-formatted string (toFixed(1)) in enrichDecode, but useWSJTX and the UI may want the numeric form for sorting. Worth passing both or keeping numeric and formatting at render time.
12. Setup HTML adds hardcoded colors #1a1f2e, #2d3548, #6b7280 — matches existing style in this file, so not a regression, but a note for a future cleanup.
13. .gitignore-ing CLAUDE.md with no comment — a one-line comment would help future readers.

Checklist: all UI-related items are unchecked — please verify at least the Modern layout WSPR tab and the CQ filter in PSKReporterPanel before merge.

Overall: fix 1–3 (real bugs) and this is ready.