mirror of
https://github.com/hanzoai/desktop.git
synced 2026-10-30 14:58:42 +00:00
f56b876472
* feat(desktop): embed native Hanzo engine + default to Zen models - Wire embedded_engine (engine fork, path dep) behind a `native-engine` Cargo feature (off by default): OpenAI-compatible server on 127.0.0.1:36900 + Tauri IPC commands, sharing one engine instance. - Port embedded engine to the engine 0.8.1 builder API. - Default to Zen models on HF: zen-nano-0.6b (chat) + zen3-embedding-small (embedding), with zen5 as flagship. - De-bundle ollama: only hanzo-node is a sidecar; ollama is external/ configurable (like LM Studio). - Docs: integration done; Zen5 = Qwen3.5 MoDE (not Qwen3.6). * fix(desktop): repoint native engine to renamed hanzo crates Engine fork was rebranded mistralrs-* -> hanzo-* (v0.9.0): - deps: mistralrs/mistralrs-core -> hanzo (SDK) + hanzo-engine (core) - imports: mistralrs:: -> hanzo::, mistralrs_core:: -> hanzo_engine:: - dev server/devUrl port 1420 -> 1421 * chore(desktop): lock hanzo/hanzo-engine native-engine deps * fix(desktop): correct Zen catalog to live hf.co/zenlm naming Repo models.ts was stale. Use verified HF ids + size-based naming: - embedding: zen-embedding-0.6B (default) / 4B / 8B (drop zen3-*, -small/-medium) - chat: zen-nano-0.6b (default), zen-5 flagship (was zen5), zen-5-coder, zen-eco-4b-instruct - add gguf_repo for prequantized local loading (zen-embedding-0.6B-GGUF, zen-5-gguf, ...) * Commit local working changes --------- Co-authored-by: Antje Worring <[email protected]> Co-authored-by: Hanzo Dev <[email protected]>