fix(recall): default embed model to Qwen3-Embedding-8B 52b7250 Running verified art87able commited on Jun 15
honest NVIDIA path (Nemotron-Mini-4B 5/5) + bake-off table + Modal runtime serving + 3-tier fallback + SFT dataset 9ff6699 verified art87able commited on Jun 15
deploy: multi-backend (minicpm/nemotron/finetuned) + 3-tier ZeroGPU->serverless->local-finetune fallback; finetune pipeline fe7f8fd art87able commited on Jun 15
tags+evidence: achievement:welltuned + sponsor:modal (Modal-trained LoRA published to HF, verified) + NVIDIA NIM recipe 8bb1ebe verified art87able commited on Jun 15
tags+evidence: sponsor:openbmb + sponsor:nvidia (MiniCPM + Nemotron via Nebius, live-verified) 03d1090 verified art87able commited on Jun 15
tags: add track:wood (multi-track) + achievement:llama (offgrid runs via llama.cpp) 042d5eb verified art87able commited on Jun 15
feat: similar-task recall via Nebius embeddings (keyless; recall degrades off without NEBIUS_API_KEY) 79e91df verified art87able commited on Jun 15
re-recorded demo: fixed layout + off-brand theme, Jamie VO dcc3436 verified art87able commited on Jun 15
off-brand: distinct identity — Fraunces gradient wordmark, gradient buttons, layered depth, focus rings, de-branded chrome 4f454e1 verified art87able commited on Jun 15
deploy: step-row layout fix (spotlight row no longer collapses) 4bf70b8 verified art87able commited on Jun 14
submission: demo+social links, tiny-titan/llama/offgrid/openai tags, four backends 077a011 verified art87able commited on Jun 14
chore: trim dev artifacts from Space (full project on GitHub: art87able/unstuck) 3b985a9 verified art87able commited on Jun 9
fix(zerogpu): load model with .to(cuda) instead of device_map 012de31 verified art87able commited on Jun 9