v0.6.4

Custom relay endpoints that serve the same model id are now tracked independently. Previously every custom relay stored its models under a single identity, so registering a model id that another endpoint already served rebound the existing row — one enable flag, one set of scores, one cooldown shared between unrelated endpoints.

Improvements

• Each custom endpoint now keeps its own model row for a given model id, with independent enable state, reliability and speed scores, cooldowns and retirement. A relay that starts failing no longer demotes, cools down or disables the same model id on a different relay (#651, #619)
• Requesting a model by its plain id now reaches every endpoint that serves it, so failover across relays works by default. When two endpoints serve the same id, the dashboard shows each endpoint's URL alongside the model and offers per-endpoint ids for pinning one specific endpoint
• A model requested by its exact id is always tried first. Models matched only by a similar display name are used as fallback afterwards, never ahead of an exact match
• Setups with a single custom endpoint are unaffected — no new fields to fill in, no naming changes, no configuration to migrate

Existing databases migrate automatically on first start; no action is needed.

Full diff: v0.6.3...v0.6.4

macOS note: the DMG is not notarized — on first launch use right-click → Open (or xattr -d com.apple.quarantine /Applications/FreeLLMAPI.app).