mirror of
https://github.com/Kpa-clawbot/meshcore-analyzer.git
synced 2026-06-03 13:44:20 +00:00
Kpa-clawbot
412a8fdb8f
## Summary Augments the shared `HopResolver` with neighbor-graph affinity data so that when multiple nodes match a hop prefix, the resolver prefers candidates that are known neighbors of the adjacent hop — instead of relying solely on geo-distance. Fixes #528 ## Changes ### `public/hop-resolver.js` - Added `affinityMap` — stores bidirectional neighbor adjacency with scores - Added `setAffinity(graph)` — ingests `/api/analytics/neighbor-graph` edge data into O(1) Map lookups - Added `getAffinity(pubkeyA, pubkeyB)` — returns affinity score between two nodes (0 if not neighbors) - Added `pickByAffinity(candidates, adjacentPubkey, anchor, ...)` — picks best candidate: affinity-neighbor first (highest score), then geo-distance fallback - Modified forward and backward passes in `resolve()` to track the previously-resolved pubkey and use `pickByAffinity` instead of raw geo-sort ### `public/live.js` - Added `fetchAffinityData()` — fetches `/api/analytics/neighbor-graph` once and calls `HopResolver.setAffinity()` - Added `startAffinityRefresh()` — refreshes affinity data every 60 seconds - Both are called from `loadNodes()` after HopResolver is initialized ### `test-hop-resolver-affinity.js` (new) - Affinity prefers neighbor candidate over geo-closest - Cold start (no affinity data) falls back to geo-closest - Null/undefined affinity doesn't crash - Bidirectional score lookup - Highest affinity score wins among multiple neighbors - Unambiguous hops unaffected by affinity ## Performance - API calls: 1 at load + 1 per 60s (no per-packet calls) - Per-packet resolve: O(1) Map lookups, <0.5ms - Memory: ~50KB for 2K-node graph --------- Co-authored-by: you <you@example.com>
