server-v0.0.5

kulikov0/whitelist-bypassserver-v0.0.5Jul 10, 2026by Dhravya

AI Summary

Self-hosted supermemory now supports flexible embedding models, allowing users to choose between local ONNX, OpenAI, or Google APIs. The release introduces safety mechanisms for upgrades to prevent vector mixing during model switches and introduces a new plan locking mechanism.

Key Highlights

Pluggable embeddings support for local ONNX (default), OpenAI, OpenAI-compatible, and Google APIs
Safe upgrades that handle legacy stores and fail fast on model dimension mismatches
Configuration via SUPERMEMORY_EMBEDDING_* environment variables

Breaking Changes

Model switching behavior changed to fail fast on dimension mismatch instead of mixing vectors
Introduction of a new embedding-plan.json file and locking mechanism

New Features

Local ONNX embedding model support
Remote embedding API integration (OpenAI, Google)
Embedding plan locking mechanism
Legacy store handling for populated stores without a lock

Full Release Notes

Self-hosted supermemory can now use local or remote embedding models.

- **Pluggable embeddings** — local ONNX (default) or OpenAI / OpenAI-compatible / Google via `SUPERMEMORY_EMBEDDING_*`; first-boot picker + plan lock in `embedding-plan.json`
- **Safe upgrades** — populated stores without a lock assume legacy `local · Xenova/bge-base-en-v1.5 · 768d`; same-dimension model switches fail fast instead of mixing vectors

```sh
curl -fsSL https://supermemory.ai/install | bash
```

```sh
supermemory-server upgrade
```