Refresh Serenity placement plan after live review
This commit is contained in:
@@ -45,12 +45,13 @@ Text Embeddings Inference reranker for OpenWebUI RAG pipeline.
|
||||
| **Model** | `BAAI/bge-reranker-v2-m3` |
|
||||
| **Image** | `ghcr.io/huggingface/text-embeddings-inference:cpu-latest` |
|
||||
| **Port** | `9787` |
|
||||
| **Volume** | `/mnt/user/appdata/reranker` |
|
||||
| **Volume** | `/mnt/user/appdate/reranker` |
|
||||
| **LiteLLM model name** | `reranker` |
|
||||
| **API base** | `http://10.5.30.5:9787` |
|
||||
|
||||
> Moved from PlausibleDeniability (port 8787) to Serenity (port 9787) on 2026-05-09.
|
||||
> CPU-only inference is acceptable for RAG reranking workloads.
|
||||
> Live runtime currently uses `/mnt/user/appdate/reranker`; preserve that path as-is until a deliberate migration normalizes it.
|
||||
|
||||
### ARR Stack
|
||||
|
||||
@@ -79,5 +80,6 @@ Serenity hosts the secondary backup Technitium node on `10.5.30.10`.
|
||||
|
||||
- Transitional only; do not design new long-term dependencies around Serenity.
|
||||
- Near-term it continues to host storage locality, reranker, the backup Technitium node, and the Serenity-local Newt needed for Pangolin access to Serenity resources.
|
||||
- Low-risk miscellaneous apps such as Wizarr, Hawser/dockersocket, and possibly netdata should be peeled off before the storage-locality lane.
|
||||
- Once PD directly owns the storage, qBittorrent/ARR locality should move to PD and Serenity should be drained and retired.
|
||||
- Before any Pangolin cleanup, re-home Serenity resources away from stale literal-IP targets; live Newt logs show repeated health checks to old `10.5.1.5` addresses.
|
||||
- The stale Pangolin/Newt health-check drift to `10.5.1.5` was fixed on 2026-05-25; future Newt work is now a deliberate rehome/retirement design task rather than an active incident.
|
||||
|
||||
Reference in New Issue
Block a user