Ollama reloaded the chat model mid-message because plan/summary/router
calls sent divergent runner options (no num_ctx/num_batch) and omitted
keep_alive — the main cause of perceived slowness.
- share one keep-alive httpx.AsyncClient for all Ollama calls
- unify runner options (runner_options) + keep_alive on every model
call, embeddings included
- drop the blocking LLM routing fallback (pure lexical heuristic)
- run RAG recall + plan + code-model pick in parallel inside the SSE
stream, after the start event
- RAG cosine scoring off the event loop; cache /api/tags 30s and
nvidia-smi 5s; frontend polls 2s->5s, warm poll backoff, dedup
config fetch
feat: working session menu in TopBar (switch/create/rename/delete)
feat: workspace file deletion (DELETE /api/files + UI trash buttons)
docs: recommended Ollama env vars (KEEP_ALIVE, MAX_LOADED_MODELS...)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- routes/system : /api/system/hardware — liste les GPU vus par le conteneur
(nvidia-smi, multi-GPU) ou l'override GPU_VRAM_MB, et le placement réel des
modèles chargés côté Ollama (GPU/CPU/mixte + VRAM) via /api/ps.
- config : réintègre gpu_vram_mb / gpu_name (perdus lors d'une fusion).
- UI : carte MATÉRIEL dans Configuration (rafraîchie toutes les 5 s) montrant
les deux points de vue et si Ollama est local ou distant.
Répond à « l'appli est-elle réservée à la 3060 ? » (non) et permet de voir
tout de suite si Ollama charge un modèle sur CPU faute de support GPU.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SVay7z3y7q2gEe54ByAE6N
Favicon now matches the LokiMark topbar icon instead of an unrelated
arrow glyph. New /api/system/stats endpoint (psutil + nvidia-smi)
polled every 2s and shown in the topbar next to the Ollama status.
MichaelandClaude Opus 4.8