Update app.py

fix: pin MODEL_REVISION to stop trust_remote_code re-exec pickle error

revert: default model back to openbmb/MiniCPM3-4B

fix: RAG seeding timeout + robust agent mode with always-on corpus search

chore: runtime updates and fixes

fix(spaces): bind app to port 7860 and disable SSR to fix HF launch timeout

fix: graceful fallback when HEARTHNET_DATA_DIR is not writable

feat: P5/P6 model swap + modal fix + doc consolidation

feat: 15 targeted improvements — RAG persistence, bus failover, agent hardening, deps sync

feat(rag): docs ingestion + UI/bus enhancements (P3 continuation)

feat(openbmb): serve multiple MiniCPM models from one local endpoint

fix(chat): wire ChatService with bus in app.py so cross-node sends deliver

fix: resolve cross-node chat [echo] bug

feat: relay hub serves in-process node directly so Space answers relay RPCs

fix: self-join relay synchronously (App has no add_event_handler)

fix: Space node auto-joins its own relay hub on startup so remote peers see it

fix: Python 3.10 compat — replace datetime.UTC with timezone.utc shim

Add all-to-all internet mesh over relay hub (P1-P3) + user-story screenshot proof

Quality improvements: Unicode chars, Token class, imports, type hints, formatting

fix: HF bus /bus/v1/call uses Body(...) not Request (422 on Space FastAPI)

feat: real node-to-node HTTP peering (local <-> HF Space)

fix: disable Gradio SSR mode so Python FastAPI handles /webagent/ requests

fix: patch App.create_app at class level to inject webagent mount

fix: popup shows webagent (not Gradio); ticker speed 80s

fix: webagent at /webagent/ via lifespan mount; ticker shows real HN+BBC news

fix: serve webagent via GRADIO_ALLOWED_PATHS env var (works on HF Space)

feat: easter egg CSS injected via js_on_load (no gr.Blocks css= deprecation)

fix: CSS via gr.Blocks(css=) + serve webagent as static files

feat: WebLLM browser agent with PeerJS mesh, HybridRAG, news signals, and easter-egg ticker

feat: activate real services, fix gossip, wire sponsor backends + EventLog

fix: asyncio.get_running_loop() across backends; expand mesh/capability/model UI docs

fix: corpus dropdown, LLM error display, chat UX, invite endpoint, getting started text

fix: 0 test failures; FileService; real RagService; emergency probe; chat return

fix: @spaces.GPU wrapper for ZeroGPU; 18 real component tests

feat: HF Space shows real 7-tab node UI with SmolLM2 backend

Update app.py

feat: mesh graph tab, routing trace, QR invite, remove all fakes, specialized-node tests

Switch Space to local-first real model backend

Add ZeroGPU startup probe

Build Phase 1 HearthNet Space