Spaces:
Sleeping
Sleeping
Commit History
-
deploy: auto-deploy main to the Space via Actions; script ships any branch
- deploy_space.sh now defaults to the current branch (main fallback on detached HEAD) instead of the stale v2, and warns when the tree is dirty - new .github/workflows/deploy-space.yml deploys every push to main using the HF_TOKEN Actions secret (concurrency-guarded) - README + AGENTS.md deploy docs rewritten around the two pathsd4c9a3c -
finished!!
c5672b8
-
agent fully implemented, cleaned ui merged data etc
a1f52bf -
wow: staged engine-visible forecast, smoothed animated 3D seafloor, lo-fi charts, ms tool badges, warm caches, clean MCP surface
- fetch_forecast: staged generator — Open-Meteo+scoring charts land first (~80ms), GEBCO world model streams in from a parallel thread with per-engine timings in the status line - seafloor: 10x10 grid (was 9x9), scipy bicubic smoothing to 41x41 display, animated swell layer (20 frames + play button) riding the forecast's dominant period/height, lo-fi themed - charts: shared style_fig (Courier/dark-blue ink), weekend shading on score/swell/wind - tool calls: list_breaks memoized (840ms -> 0.7ms warm), per-tool ms badges on agent tool cards - ui: engine strip under the hero (world model / feeds / climate / scorer / LLM / MCP), demo.load warm thread for featured breaks - MCP: internal listeners hidden from the API — surface is exactly score_week / rank_region_week / explain_score - scripts/warm_caches.py: pre-fill forecast+seafloor caches before a demo recording - scipy added as the only new dependency69bd160 -
merge v2-agent (agent+narrator) into v2
74a6179 -
climate: climatology grounding (climate.py, build_climate.py, tests, 126 profiles)
93d288c -
Worker C: agent + narrator (llm factory, 9 tools, ToolCallingAgent, narrate, eval, dry-run tests)
bf20e78