feat(api): cancel analysis runs via loopback CLI

[seonghobae]

Why

GAP-003A / issue #166 still required operators to write raw HTTP/1.1 after cancel HTTP. Operators cannot withdraw accepted or running runs from a client without guessing framing. Cancel CLI stdout stays metric-free. tepp.scientific_acceptance.v1 never appears.

Live slices that this PR does not duplicate:

• feat(analysis): bind validation runs to scientific acceptance evidence #356 analysis_engine scientific-acceptance library
• feat(api): carry scientific acceptance only on terminal results #358 tepp_api terminal-wire DTO
• feat(api): serve scientific acceptance only on loopback status GET #359 GET loopback status by id
• feat(api): record scientific acceptance via loopback lifecycle POST #360 POST running/terminal lifecycle
• feat(api): cancel accepted and running analysis runs on loopback #361 cancel HTTP
• feat(api): drive scientific acceptance via loopback CLI #362 scientific-acceptance loopback CLI (tepp-analysis-run)
• feat(api): enumerate analysis runs via loopback collection GET #368 collection GET
• feat(api): enumerate analysis runs via loopback collection CLI #371 collection CLI list (this PR stacks on that head f7d5e5a391eb2caa55bd5d337e90e719137b5041)
• feat(api): cancel analysis runs from LineageWeave and Naruon compatibility #373 cancel consumer-parity
• feat(api): retry failed and cancelled analysis runs on loopback #369 retry HTTP
• feat(analysis): bind TDT/CHRONOS composition to an analysis-run profile #364 GAP-007 TDT/CHRONOS analysis-run profile
• feat(persistence): persist idempotent analysis runs #287 GAP-003B persistence/Compose
• feat(network): replace union-find stand-in with Leiden consensus #351 Leiden

What

Ninth GAP-003A slice stacked on #371 over collection GET #368 / cancel HTTP #361 / protected main 1bc02f580cf48e1d39da239f0e818453437c31c3:

• tepp-analysis-runs cancel client of POST /v1/analysis-runs/{run_id}/cancel
• metric-free cancelled status: run_id, run_state, idempotency_key, terminal_result: null
• accepted cancel and idempotent replay; another consumer cannot cancel the first consumer's run
• tepp.scientific_acceptance.v1 never prints
• empty stdin is admitted; typed cancel JSON must match --run-id / --idempotency-key
• loopback-only; unpublished consumers, credential flags, collection pagination flags, metric stdin, and unknown verbs fail closed
• ADR 0033 (0026–0032 live on other GAP-003A PRs)

Not an ADR 0014 scientific claim. Not persistence. Not GET-by-id. Not lifecycle POST. Not cancel HTTP. Not scientific-acceptance CLI. Not collection list.

Verification

• cargo clippy -p tepp_api --all-targets -- -D warnings
• cargo test -p tepp_api --lib
• cargo test -p tepp_api --test analysis_run_cancel_cli_contract
• RUSTDOCFLAGS='-D warnings' cargo doc -p tepp_api --no-deps
• python3 scripts/validate_documentation.py
• python3 scripts/check_docstrings.py

Do not merge without two independent APPROVE reviews on this exact head and current Checks. Author COMMENTED is not independent APPROVE.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

Hour-21 exact-head review request.

Current head c891d46a57c03e71f1ca44150a00a9d71aa2f2a9.
Devin/CodeRabbit/author COMMENTED is not independent APPROVE. Copilot review requests are not APPROVE. Ruleset 18156473 needs two independent approvals. Do not self-approve. Do not merge without two independent APPROVE reviews on this exact head.

@opencode-agent review

[seonghobae]

Closing with replacement mapping to #174. The CLI is a client of the unauthenticated #361 destructive mutation and cannot make that boundary safe. Preserve its typed request/framing/stdout/refusal tests for the future authenticated Analysis Run operations CLI, but do not ship a credential-less loopback cancel command.