revert: simple lazy loading inside @spaces.GPU — working baseline

fix: initialize self.pipe = None in __init__ — prevent AttributeError

fix: @spaces.GPU on standalone function, not class method — ZeroGPU compat

fix: proper ZeroGPU lifecycle — try/finally CPU offload, @spaces.GPU on generate()

fix: recreate pipeline each request — ZeroGPU deallocates GPU between calls

perf: decouple CPU retrieval from GPU generation — 10x GPU quota savings

fix: torch_dtype→dtype deprecation, remove pad_token_id conflict with generation_config

fix: remove device_map conflict, load onto GPU inside @spaces.GPU context

fix: full lazy model loading for ZeroGPU — remove preload thread, defer pipeline init to first request

refactor: Gradio 5.30 compat + spaces first import + deferred torch in generation.py

refactor: ZeroGPU-compatible — spaces first import, @spaces.GPU on generate(), deferred torch/transformers

fix: dummy @spaces.GPU placeholder — no GPU allocation, no quota consumption (same pattern as neurorag-2026)

fix: keep @spaces.GPU(duration=15) — zero-a10g requires it, but with HF_TOKEN secret for authenticated quota

fix: remove @spaces.GPU — run on CPU to eliminate ZeroGPU quota errors for website visitors

fix: use GPU properly — @spaces.GPU + device_map=auto for A10G

Aethron Portfolio Agent v2: Phi-3.5-mini + FAISS + keyword fallback, HF Spaces compatible