perf: greedy decoding + inference_mode for coherent fast answers on CIF path

fix: ThreadingHTTPServer so health checks never block during inference + all CPU threads for generation

fix: increase inference timeout 30s->180s, reduce nSamples 3->1 and maxTokens 400->100 for CPU feasibility

fix: complete model server HTTP handler, remove duplicate daemon, extend health timeout to 900s

fix: extend daemon health-probe timeout 90s→300s for 515MB model download+load

fix: remove circular isCheckpointReady() guard from startModelDaemon — daemon now self-starts

fix: aiScore ReferenceError — declare variables before if/else block

feat: AI/Search % badge, search-conditioned model, daemon self-fetches checkpoint via hf_hub_download

search-guided model: fetch first, condition model on results, maxTokens=400, usedModel flag

model-first: generate tokens directly, web search for grounding only

Fix: remove broken getCodeIntro ref, restore working code template flow

Io HST Search Engine — initial HuggingFace Space

Io HST Search Engine — initial HuggingFace Space