[chappie-daemon]

Implements the fourteen NotImplemented event-strategy arms behind the existing select / EventStrategy seam, plus the three core pieces they call. Every arm is ported from the capture at 242a3001 and pinned by a census that now covers all 36 strategy spec files.

Closes #974.

Needs a human — access and identity

The factory's approver returned ESCALATE with all three approve conditions holding. The condition is money-access-identity, and the reason is what this branch makes live rather than how it was built: these four arms no longer answer NotImplemented, they decide things.

• join_request — NIP-42 gate, then claim_code, then admit_user
• leave_request — revocation
• report — mark_actionable_target under hide_actionable_reports

Job 007 escalated on the same condition for the storage half of this path.

Both features ship disabled (nip43.enabled: false, nip56.enabled: false, nip56.hideActionableReports: false in resources/default-settings.yaml), so the question is narrow and it is a judgement rather than a defect: is dormant-by-default enough for rewrite/main, or should those three arms be read by a person first?

Evidence, for a reader who will not open build.md

• Oracle exact: census: 658 cases, 612 mapped across 200 mapping rows to 182 test functions, 7 exempt groups holding 46 cases, 4 passed; 0 failed.
• The fixture is the capture's own output: the committed generator run against a tree extracted from 242a3001 prints 658 cases from 36 files and diffs empty against crates/nostream-relay/tests/fixtures/ts-cases.json.
• All fifteen acceptance criteria, each with the command that checks it and its output, in the approver's own run — not copied from the builder's report.
• git merge-tree --write-tree rewrite/main factory/011-relay-strategies exits 0.

Repository checks in both profiles: cargo fmt --all --check clean, both clippy passes clean with no warning line, cargo test --workspace --all-features 56 test targets, 0 failed, in debug and in release, cargo build --workspace --all-features --release clean.

What lands

area	change
crates/nostream-relay/src/strategies/	fourteen arms: contact_list, delete, dvm_job_request, ephemeral, gift_wrap, group, invite_request, join_request, leave_request, parameterized_replaceable, replaceable, report, timestamp, vanish
crates/nostream-core	nip03 (OpenTimestamps validation for kind 1040), nip56 (report target extraction), report_scoring (ReportWeight::weight)
crates/nostream-relay/src/wot.rs	WotGraph trait and UnavailableWotGraph, carried on HandlerContext beside limits
census	fixture regenerated from the capture: 658 cases, 612 mapped, 36 spec files
docs	ADR 0006 records the four decisions this job had to make; ADR 0004's census figures move 557 → 658
tests	tests/strategies.rs (100 cases) and the extended parity oracle

Review

Two rounds, four lenses each (ui and ux stood down in both, with reasons checked against the whole branch diff rather than the round's). Round 1 returned CHANGES from code and architecture — four findings, all closed and re-verified against the code in round 2. Security and performance passed in both rounds.

Verdict sequencing: the approver's decision was reached on ae6f889. The tree now differs only in the two ADR citation lines committed at 06463a5, which is why the verdict still stands; the approver checked that delta by hand.

Notes for the reader

• Commit hooks did not run on this branch. .husky/pre-commit drives the legacy pnpm lint/build/test and its .husky/_/husky.sh is absent in the worktree, so it aborts before any check. Commits used HUSKY=0 git commit --no-verify. What ran instead is the list under Evidence, in full, at the head.
• wire_parity's oversized-frame step no longer requires a 614 600-byte write to succeed. The relay refuses on the frame's declared length and may close before the client finishes writing, which is what the capture's socket does too; the step tolerates an IO error on that write and keeps the 1009 close comparison as its assertion. Measured at 24b679c: 10 failures in 20 release runs before; 0 in 30 after, with 11 of 20 writes losing the race during classification. Closes relay: the wire_parity release test fails intermittently with BrokenPipe #974.
• A kind-62 vanish request can be outrun by a concurrent publish, and reads exclude deleted rows rather than vanished authors, so the late event survives. The capture has the same window in the same order, so this port reproduces it rather than inventing a fix; filed as relay: a publish racing a kind-62 vanish request survives it #975 with the decision points written out.

[changeset-bot]

🦋 Changeset detected

Latest commit: ba95e52

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 0 packages

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[greptile-apps]

Confidence Score: 5/5

[High impact] The changes since the previous review appear safe to merge.

Summary

Implements fourteen event strategies, adds OpenTimestamps checks and report helpers, and expands the parity census.

• The relay stores, rejects, deletes, or broadcasts events according to their kind.
• NIP-43 requests can admit or revoke relay users.
• Reports can score targets and hide actionable reports from live fan-out.
• Kind 1040 events now need a matching OpenTimestamps proof.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Event passes shared checks] --> B[Select event strategy]
  B --> C[Store, replace, or delete]
  B --> D[Broadcast without storage]
  B --> E[Join or leave membership]
  C --> F[Send result]
  D --> F
  E --> F
  C --> G[Broadcast when allowed]
  G --> H[Update reports, jobs, or follows]

Loading

_{Reviews (2) · Last reviewed commit: "test(core): give the report-target fixtu..." · Reviewed by Greptile}