Remove internal docs from public repo and tighten privacy for HF.

Fix Ask Orion output mismatch and polish UI per copy guide.

Strengthen Nemotron expert voice with richer facts and fallback.

Fix UX bugs: constellation guide nav, weather/score alignment + forecast date match, persistent telemetry chips, Ask Orion placeholder hide, NL city handling + custom update, skyview constellation highlight, no "perfect" on low scores.

Sharpen Nemotron expert voice and mark Wave A2 complete in roadmap.

Support weather forecast scoring and adjust summary in Ask Orion (find_best_date)

Implement off-brand Midnight Oasis layout with Control Room sidebar, telemetry HUD, and concurrent model tabs

Plan loading state + visible LLM status (Orion's Cottage)

Make LLM polish write fresh narratives and unload Ollama after each call.

Add inference gateway, single-model planner, and honest field notes at obs time.

Implement P0: fix Ollama polish for all three models and show inference status in UI.

Sprint B + launch robustness + cleanup.

Fix review items A1-A5: astronomy correctness, scoring, reliability, UX, and location timezones.

Implement Phase 6: date picker, custom location geocoding, and NL query.

Implement Phase 5: Targets Tab and Constellation Browser

Implement Phase 0: Add starplot visualization and 3-model concurrency

Final Hackathon Polish and README for HF Spaces

Implement Field Journal layout and Red-Light Mode MVP

Fix reference in agent/tools.py

Cutover to deterministic-first router and update polish prompts

Extract local LLM client adapter

Clean up router and UI toggles

Add local-first router orchestration

Define agent tools and prompt rules

Add Phase 1 Modal toggle and tests

chore(phase1): project setup with folder structure, requirements, working Gradio skeleton, core/agent/data stubs, and placeholders