Add feasibility study: JPL Scout → Kafka bridge for Rubin ToO alerting

[talister]

Summary

Adds docs/scout_kafka_bridge_feasibility.md — a feasibility study for a containerized service that polls JPL Scout and republishes new/updated/cancelled Rubin-ToO-candidate objects as a Kafka stream for the Rubin Observatory ToO system.

Key points:

• Rubin's first-year ToO operations paper (arXiv:2607.00217) explicitly notes no Kafka stream exists for hazardous asteroids and invites one; Rubin ingests all external ToO alerts via SCiMMA Hopskotch.
• Broker options evaluated: SCiMMA Hopskotch (recommended, $0) vs self-hosted Kafka on AWS (MSK ≈ $547/mo minimum; single-node ≈ $30–60/mo + ops, and Rubin would need a bespoke consumer).
• Architecture scoping concluded on a standalone TOM Toolkit project depending on tom_jpl (no logic forking; Postgres transactional outbox; ~1 GB SPICE-free image) over a Django-free micro-poller or building it into FOMO.
• Deployment via LCO's standard GitOps pattern: app repo lsst-sssc/scout-alert-bridge (public ghcr image) + deploy repo LCOGT/scout-alert-bridge-deploy (kpt/kustomize, sealed-secrets, ArgoCD, 10-min CronJob).
• Includes a v1 event/message schema proposal, failure-mode analysis, costs, open coordination questions (SCiMMA, Rubin ToO team/SCOC, CNEOS, LCO infra), and prototype milestones M0–M5.

Note: the doc is standalone Markdown (not yet in the Sphinx toctree, which is rst-only); it renders on GitHub and can be MyST-ified later if we want it in the built docs.

🤖 Generated with Claude Code

https://claude.ai/code/session_01EmBXs3EBmEmMZRPZG7pESg