- OpenRouter as an alternative chat provider: server-side API key
(settings.json or OPENROUTER_API_KEY), model list with vision marks and
filter, SSE streaming, images as data URLs. Memory embeddings stay on Ollama;
park/warm LLM are no-ops for the remote provider.
- AssistentSaveKnowledgeAttach: JArray param is not supported by SwarmUI API
reflection and aborted registration of every later API call; use string[].
- SQLite bootstrap: preload native e_sqlite3 for the current OS/arch
(win/osx/linux) and resolve DllImport to it.
- settings.json: shared-read + atomic write with retry to avoid Windows
sharing violations.
- csproj: fix MSB4012 in the native SQLite copy target.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Index Assistent/books search.jsonl, expose knowledge hops/catalog in chat, and drop QLoRA/HF training stack.
Co-authored-by: Cursor <cursoragent@cursor.com>
Wrap AssistentImportHfDataset in try/catch, batch inserts in a transaction, clearer HF rows errors for gated datasets, EnsureTrainingReady with actionable message, and friendlier training UI errors instead of generic internal error.
Co-authored-by: Cursor <cursoragent@cursor.com>
Adds session_exact pinning, sampler/scheduler chips, expanded param tags with non-default highlighting, QLoRA HF presets, LoRA strength editing, SQLite bootstrap for training memory, last-job UI, and harsher critique_image QC so result review leads with defects instead of praise.
Co-authored-by: Cursor <cursoragent@cursor.com>
Chips and quick patches apply params without auto-Generate; aspect sizes match Swarm Side Length 1024. HF import shows drafts after switching filter from approved-only; training status and job polling improved.
Co-authored-by: Cursor <cursoragent@cursor.com>
Load Assistent/extensions packs between bundled and overlay, keep neutral/aggressive/dreamer, drop cinema/lewd/private shelves, and make horny slash commands schema-driven.
Co-authored-by: Cursor <cursoragent@cursor.com>
Overwrite foreign steps/cfg in sparse patches when the user did not ask for params, and skip the cold-load path when Ollama already has the chat model resident.
Co-authored-by: Cursor <cursoragent@cursor.com>
Client always merges Exact turbo|raw steps/cfg/sigma before Generate so sparse LLM omissions and leftover SD 20/7 cannot stick; Ollama park/warm skip no-op /api/ps round-trips when the chat model is already unloaded or resident.
Co-authored-by: Cursor <cursoragent@cursor.com>
Per-chat Generate session with sparse deltas; drop Cards/Civitai/wanted hops; rolling history summary via the same Ollama model with a budget chip and /compress.
Co-authored-by: Cursor <cursoragent@cursor.com>
Replace fake train loop with TRL SFTTrainer, HF column mapping with fiction preset,
safetensors to GGUF conversion, and ollama create using ollama_base plus ADAPTER.
Co-authored-by: Cursor <cursoragent@cursor.com>
Restructure UI with app-level tabs and chat history drawer; add dataset curation,
HF import, Modelfile/QLoRA hooks, and link approved samples to the agent immediately
via heard vector memory without waiting for fine-tuning.
Co-authored-by: Cursor <cursoragent@cursor.com>
Replace split Assets JS with a built bundle and aligned C#/config so SwarmUI loads one script and patch apply stays consistent; drop legacy terse persona and memory seed.
Co-authored-by: Cursor <cursoragent@cursor.com>
mrleo1nidandClaude Opus 5
Leonid PershinandCursor