Phase 2c follow-up: InfoSpec schema extension for fetch_config knobs

[gregoryfoster]

Context

Phase 2c dropped the `watches.fetch_config` JSONB column. The InfoSpec v1 JSON Schema doesn't yet model the knobs it replaced — `headers`, `ignore_patterns`, `viewport_width`, `viewport_height`, browser fetcher hints, etc. Watcher currently fail-fasts (or silently ignores) operator attempts to set them.

Listed under "Open follow-ups (deferred)" in docs/plans/2026-05-04-watcher-phase2c-cutover-plan.md wrap-up.

What to do

• Audit the pre-Phase-2c `fetch_config` shape (see legacy migration files for the exact JSON). Inventory of fields that need to come back: at minimum `headers`, `ignore_patterns`, `viewport_width`, `viewport_height`, `render` (for the future Playwright fetcher).
• Extend the InfoSpec v1 JSON Schema's `target.fetch` block to accept these. Keep field-level defaults at the consumer (per design doc decision §5: "DEFAULT_FETCH_RENDER = False" etc.).
• Wire each knob through Watcher's fetcher / extractor:
  • `headers` → `HttpFetcher` outbound headers
  • `ignore_patterns` → already partially wired via `HtmlExtractor` ignore selectors; verify post-Phase-2c
  • `viewport_width` / `viewport_height` → screenshot capture (`src/core/screenshot.py`)
  • `render` → reserved for Add Playwright headless browser fetcher #3 (Playwright fetcher); document but don't implement consumption
• Validation: existing single-URL specs without any `target.fetch` block must remain valid (all fields optional).
• Tests: round-trip each new knob; verify the consumer applies it.

References

• Phase 2c plan wrap-up: docs/plans/2026-05-04-watcher-phase2c-cutover-plan.md
• Design § 5 ("Outbound transport") — consumer-side defaults: docs/plans/2026-05-03-information-source-specifications-design.md
• Parent: Information Source Specifications & Watcher boundary realignment #138

[gregoryfoster]

Migrated to CannObserv/archiver#5 during the Archiver service extraction (#149). Path references in the body have been updated to the new archiver repo layout. Closing here.