[KGKallasmaa]

keenable fetch now takes one or more URLs and the v1 Fetch options, sent to POST /v1/fetch.

Do not merge or release until POST /v1/fetch is public. It sits behind the per-org fetch_post flag (backend stack #772 → #773 → #774 → #776 → #777 → #778, plus #775).

Behaviour

• When POST is used: with more than one URL, or with any of --offset, --no-text, --summary, --schema-file, --highlights [QUERY], --links N, --images N, --max-age, or --timeout.
• When GET is still used: a single URL with only today's flags (--live, --prompt, --max-chars, --session-id) keeps calling GET /v1/fetch, so nothing changes for current users.
• Old flags on POST: --prompt becomes the summary instruction, and --live becomes max_age: "0s".
• --schema-file PATH: must hold a JSON object; it's sent as content.summary.output_schema. A bad file exits 2.
• Output:
  • YAML mode prints the API response as is.
  • -p prints each page (text, or the summary, pretty-printed when it's JSON), a hint with --offset when there's more text, and one warning per failed URL.
  • Exit code 6 when every URL failed.
• Transport: the daemon and the direct HTTP path both gain a fetch_post command that POSTs the JSON body. Daemon and client builds must match, so an older daemon refuses the request and gets restarted.

Checks

• Not built locally: this repo's rule is that Rust builds and tests run on the devbox, and it's unreachable today, so CI (fmt, clippy -D warnings, cargo test) is the build check here.
• New unit tests: build_fetch_body (defaults, every option, no text with default highlights), every_url_failed and read_schema_file.
• To do: e2e cases in tests/e2e/test_fetch.py, once the flag is on for the e2e key's org.

Linear: KEE-1403 (Fetch POST) and KEE-1438 (CLI).

🤖 Generated with Claude Code