`ConcordRefounding.compactControlPlane` re-wraps one edition per entity
when a community rotates epoch, and the ROTATOR chooses which one
survives. The receiving side had no memory: `refold()` folds only the
wraps at the current epoch's Control-Plane address and discards the prior
epoch's buffer, and `EditionFold` accepts whatever it is handed (its
no-genesis fallback anchors at the lowest version present).
So a rotator could publish only version 1 of a chain and omit version 2 —
restoring a revoked role, clearing a banlist, reverting metadata. Every
signature is genuine; this is rollback by omission, not forgery.
Adds a per-entity floor: the version AND hash last successfully folded.
- **No floor (fresh joiner)** — unchanged: genesis anchor, else the
lowest-version edition as the legitimate compaction bootstrap.
- **With a floor** — the walk is anchored AT the floor: the offered set
must contain the exact edition already folded (version and hash; a
same-version sibling is a fork, not our chain), then walks up. A head
below the floor is structurally unreachable.
- **Gap** (the floor edition is absent) — refuse, and keep the known
head. Refusing by *retaining* matters here: this fold is recomputed
from scratch each time, so letting an entity vanish would itself be a
rollback — a dropped banlist is an unban.
The floor needs no new persistence. It is derived from `heldRoots`, the
rotated-out access roots already persisted in the NIP-44 self-encrypted
kind-13302 list: the session derives each prior epoch's Control-Plane
address from them, folds oldest-first, and takes the resulting heads as
the floor. That survives both a process restart and the session rebuild
`ConcordSessionRegistry.sync` performs at exactly the moment of a
Refounding — which would have destroyed any in-session floor. If the old
planes are not served, there is no floor and behaviour is as before.
Floors are built from AUTHORITY-GATED heads, not raw ones. Without that,
any ex-member still holding a rotated-out root could mint a high-version
edition on the old plane and freeze the entity for every honest client —
a denial of service this change would otherwise have introduced. Covered
by a test.
Verified by disabling both enforcement points: 7 of 12 quartz tests and
the end-to-end commons test fail, and the ones that still pass are
exactly the non-regression cases (fresh joiner, honest compaction,
pass-through without floors).
Known limit: `AuthorityResolver.resolve` folds authorized SUBSETS of the
edition pool and does not carry floors itself; gating happens at the pool
level before the resolver sees anything. Sound, but connectivity checked
on the full set is a weaker precondition than on each subset — passing
floors into the resolver's three folds is worth a follow-up.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reviewed the previous commit's authority changes against Armada
(gitlab.com/soapbox-pub/armada), the reference Concord client we interop
with. Read for its rules only — Armada is AGPLv3 and Amethyst is MIT, so
no code was taken from it.
It confirmed the role rank gate (Armada checks both directions too: the
author must outrank the position being minted AND the standing position
being replaced) and the banlist check on rotation. It also showed one of
our rules was wrong and one gate was missing.
**Removes the owner's refusal of foreign rotations.** The previous commit
made the owner ignore any rotation it did not author, reasoning that a
BAN-holder could otherwise carry the owner onto a root of their choosing.
Armada does the opposite on purpose — "authority is the roster, never key
possession" — and it is right: an admin legitimately rotating to remove a
spammer would leave the owner alone on the dead epoch, self-inflicting
the strand the rule was meant to prevent, and diverging from the
reference implementation forks communities across clients. The threat is
better answered by the rank gate: with role editions gated, nobody can
escalate themselves to BAN, so BAN-holders are people the owner
deliberately trusted.
**Adds the missing rank gate on grants.** A grant was authorized if the
granter outranked every role it handed out — but a REVOKE carries no role
ids, and `all {}` over an empty list is vacuously true. So any
MANAGE_ROLES holder could strip anyone's roles, the owner's admins
included: promotion was gated, demotion was free. Armada treats a grant
as an action ON the member and requires outranking the target's standing
rank; this now does the same.
Both new tests were verified to fail with the corresponding check
disabled, and each has a companion asserting the legitimate case still
works (an admin can still revoke a moderator beneath it).
Also records what Armada does about exclusion, since we cannot prevent it
receiver-side: it does not try to. A rotation carries no recipient list
there either, so a stranded member instead re-resolves the invite link
they joined through and merges forward to the higher epoch ("stranded
recovery"). Amethyst has no equivalent, so a stranded member stays
stranded — noted at the call site as follow-up work.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two authority holes, plus the limit that remains.
**Role editions had no rank gate.** Grant editions are correctly gated —
a granter must hold MANAGE_ROLES *and* strictly outrank every role it
hands out. Role editions checked only that the author held MANAGE_ROLES.
Nothing stopped an authorized signer editing a role at or above its own
rank, including the role it holds itself.
So a moderator at position 5 with MANAGE_ROLES could publish one edition
on their own role's chain claiming position 1 and every permission bit,
then a second demoting the real admins beneath them — reaching full
authority over everyone but the owner in two editions. The
rogue-higher-version defence added earlier does not cover this: it drops
editions from *unauthorized* signers, and this signer is authorized.
Role editions are now gated in both directions: an author may not claim a
position at or above its own rank, may not touch a role that already sits
at or above it, and may not hand a role permission bits it does not hold
itself. Deleting keeps only the second rule, so retiring a role beneath
you still works. The owner is unaffected.
**A banned moderator could still rotate the community.** The rekey
receive path authorized the rotator with `effectivePermissions(...)`,
which ignores the banlist, rather than `hasPermission(...)`, which
excludes banned members. Now uses the latter.
**The owner no longer adopts a root someone else minted.** A rotation
replaces the community root, so a rotator who *includes* the owner as a
recipient hands themselves the keys to the owner's own community — the
owner would follow them onto an attacker-chosen epoch. The owner changes
epoch only by rotating themselves.
**Known limit, documented at the call site.** A rotation carries only
(newRoot, newEpoch, rotator) — no recipient list — so a receiver cannot
tell who was omitted. A BAN-holder can therefore still evict the owner by
leaving them out: everyone else adopts, the owner is stranded on the old
epoch. That is not fixable in the receive path; it needs a protocol
change (a recipient commitment the receiver can check, or owner
co-signing). Tracked for CORD-06.
The escalation test was verified to fail with the gate disabled, and a
companion test asserts an admin can still edit a role beneath it, so the
gate is not merely blocking everything.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three fixes to the Concord invite and moderation paths.
**Invite deep links redeemed with zero consent.** `ConcordInviteScreen`
called `joinConcordViaInvite` from a `LaunchedEffect` on open, and the
manifest registers `https://amethyst.social/invite/` as BROWSABLE. So a
link on any web page — or a QR code, or a push — silently caused a
connection to up to three ATTACKER-CHOSEN relay URLs decoded from the URL
fragment (disclosing the user's IP to a third party), a Guestbook JOIN
signed by the user's identity published to those relays, and a write to
their private community list. No tap, no preview.
The screen now opens in an awaiting-consent state and only joins from an
explicit Join button. The preview is built entirely from the link itself
— base64url and NIP-19 decoding, both pure in-memory — and touches the
network for nothing: no relay connection, no signing, no publishing. It
shows the relays it would contact so the user can see whom they'd be
talking to. The community name lives inside a bundle only those relays
can serve, so it is honestly reported as unknown until joining rather
than fetched.
**Invite expiry was decorative.** `ConcordInviteBundle.isExpired` had no
production callers at all — the only ones were in a test — so an expired
invite redeemed forever. Expiry is now enforced at `classify`, the choke
point every redeem path funnels through, with its own result and message
so the user knows to ask for a fresh link.
**Moderation read the wrong edition.** `ConcordModeration` used
`firstOrNull` over `controlEditions()`, which is in wrap-ARRIVAL order,
not the folded head. Once an entity had two or more editions the next one
chained off a stale predecessor, forking the chain at an already-used
version, and `EditionFold` then resolved the fork by `minByOrNull` on the
rumor id — a coin flip. Bans were masked by a down-only healing union;
UNBANS and role revocations were not, so they could silently fail to
apply. Both call sites now fold to the true head.
Regression tests assert the fold-head behaviour under two arrival orders
— a single order accidentally puts the head first and passes against the
buggy code.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`computeShares` paid each fee recipient `totalMilliSats * split / 100`
with no upper bound on `split`, and `split` comes verbatim from a
kind-30054 episode event that anyone can publish (`ValueTag.parse` is a
bare `fromJson` with no validation). A single `fee:true, split:1000`
recipient was therefore paid TEN TIMES the amount the user chose. The
`remainder` clamp looked like a safety net but only zeroed the honest
recipients; it never touched the fee recipients themselves.
Proven by test before fixing: `split:1000` pays 10x, and two fee
recipients at 60% each pay 1,200,000 millisats for a 1,000,000 zap.
This was reachable in the worst possible place. Streaming V4V pays every
minute, automatically, so a modest multiplier stays under a typical NWC
budget and simply runs — and the on-screen running total tracks the
INTENDED amount, so it reads "100 sats" while 1,000 left the wallet,
while the streaming error handler suppresses the toast. The ordinary zap
button reroutes through the same path for any note carrying a value
block, so it was not limited to the streaming toggle.
Fees are a percentage off the top, so each is now clamped to 100% and the
cumulative total to the remaining budget: the payout can never exceed
what the user chose.
The existing test asserted the right invariant (`sum <= total`) but only
ever ran it on well-formed input, which is why this survived.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adversarial audit of the PR's own changes (8 finder angles, verified
before fixing). Quartz core:
- fetchAll-family drains get a wall-clock ceiling (maxTotalMs, default
10x the idle window, delay()-watchdog: cancellable and virtual-time
testable). The pure idle window was unbounded when a relay trickled
events forever — sandboxed napplet queries, set -e fetches, and
marmot await stuck inside one drain. Streaming relays still finish.
- The suspending onEvent hook no longer runs inside a cancellable
timeout scope (an expiring window could cancel verifyAndStore
mid-write and silently drop a received event); the timeout is armed
only when the channels are dry (no per-message timeout-job churn).
- fetchAll is a projection over fetchAllWithHooks: fixes its
unsynchronized events/seenIds mutation from concurrent socket
threads and deletes the duplicate loop + per-event activity channel.
- publishAndConfirmDetailed regains its only-responders contract
(synthetic no-response entries no longer render as 'relay rejected
your message' in app callers); results built by pure associateWith;
shared failure-reason constants + PublishResult.isTransportFailure.
- NIP-65 mutations: split read+write r-tags for the same URL now merge
to BOTH instead of last-wins dropping a facet (+ test).
- TcpProber's 128-thread pool drains after 60s idle.
CLI:
- publishGuard: all-transport failure exits 124 as timeout; rejected/1
is reserved for an actual OK-false answer.
- --help anywhere in argv is hoisted centrally; 'amy notes post "x"
--help' prints usage instead of publishing.
- rejectUnknown false-reject traps fixed: geochat --no-fetch behind an
early return, and 13 elvis-alias short-circuit sites read eagerly.
- Aliases load once per Context and only match name-shaped inputs (no
shadowing a real npub/NIP-05/hex); stderr color requires a
positively-known terminal (TERM sniff polluted captured logs).
- Relay-CSV strictness unified on RawEventSupport.relayFlag (post,
graperank publish/followers/register no longer silently drop
malformed URLs); Args.timeoutMs(+OrNull) replaces 27 hand-rolled
conversions, all strict; offer/debit --timeout > 3600 rejected with
a 'looks like milliseconds' hint; NPub.create idiom; stale jq .id in
the marmot reactions harness; printUsage drift (offer pay --with,
profile --clink-offer, search --kind).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CP4kfLCa3wWtE8Khy21Pkj
The fetchAll/fetchAllWithHooks accessories wrapped their whole
collection loop in one withTimeoutOrNull, so a relay actively
streaming a large backlog was cropped mid-delivery the moment the
absolute deadline hit — even though the loop already has proper
terminal conditions (per-relay EOSE / CLOSED / cannot-connect) and the
timeout's only real job is stall detection.
timeoutMs now measures the delta since the LAST message: every event
or terminal signal resets the window (fetchAll gains a conflated
activity ping so event progress is visible to its wait loop), and only
a full window of silence ends the fetch early. fetchFirst/count keep
absolute waits (single-response — idle and absolute coincide), and
subscribe's duration timeout stays absolute by design (a live stream
has no terminal state).
Since the pages/pool helpers delegate to fetchAll, pagination inherits
the semantics. This also changes app-side callers of these accessories
— in their favor: the timeout only ever fired on slow relays, exactly
when cropping loses data.
New commonTest suite pins the behavior: a relay emitting every 200ms
under a 300ms window streams to completion (10/10 events); a stall
ends one window after the last message, not after the start; EOSE
still returns immediately. CLI docs reworded (--timeout = idle window).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CP4kfLCa3wWtE8Khy21Pkj
A relay whose domain no longer exists (lapsed registration, decommissioned
host) was treated like a busy relay: the backoff doubled from 1s and spent
about ten dials climbing to the ceiling it was always going to reach. An HTTP
upgrade rejection already jumps straight there; a name that does not resolve
deserves the same.
Matching is on the exception type rather than the message because the message
is localized and platform-specific — Android says `Unable to resolve host "x"`,
JVM on macOS says `nodename nor servname provided, or not known` — while the
class name is stable. That is why onCannotConnect appends it in the first
place. Neither message ends with "Host unreachable", so the existing check
never caught DNS failures.
Being this eager is only safe because the verdict is cheap to revisit: a DNS
answer is a property of the network, not of the relay (a captive portal or a
filtering resolver forges NXDOMAIN), and both a network-identity change and a
transport change now clear the backoff outright. The test pins that round trip,
not just the classification.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A relay's reconnect backoff was process-global and network-blind: the delay
earned on one network was still being served out on the next. The only reset
signal was OkHttpClient reference identity, which is rebuilt off the metered
bit, so it fired for wifi<->cellular and nothing else. Wifi A -> wifi B, a VPN
coming up, a captive portal clearing, and metered-wifi -> cellular all left
every relay parked on a penalty earned against a network the device had left —
up to five minutes of silence on a network that might reach the relay instantly.
The transient Off that would have reset things is swallowed by the 200ms
debounce in ConnectivityFlow, so it never rescued those cases.
Key the decision on ConnectivityStatus.Active.networkId instead, which is the
same signal SurgeDns already uses to stale its cache, and treat a genuine
network change as a full pool rebuild: every socket is bound to an interface
that no longer carries traffic, and needsToReconnect() cannot see that because
it only compares the proxy and the timeouts.
Also treat a Tor policy flip as a transport change. Flipping a Tor toggle while
Tor is already up leaves both OkHttpClient references identical, so a relay
whose transport just changed kept waiting out a backoff earned on the other
transport. Only TorRelaySettings is compared, not the relay sets that
TorRelayEvaluation also carries — those churn while an account's relay lists
load (observed firing three times in one cold start), and forgiving the whole
pool every time any list updates is far more damage than it repairs.
Adds IRelayClient.resetBackoff() (default no-op, so the existing fakes and
BleNostrClient are unaffected) rather than reusing ignoreRetryDelays, which
only skips the gate for a single attempt and still doubles the stored delay —
a relay that failed that one dial came back worse off than before.
INostrClient.resetBackoff() is deliberately separate from reconnect(): the
latter debounces, so folding this into a coalescing command would let a later
request silently drop the reset.
The decision table moves into RelayProxyClientConnector.apply() so it can be
exercised directly, without a debounce and a shared StateFlow in the way.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The kind:10002 facet-merge rules (adding a write marker to a read-only
relay promotes it to BOTH; removing one facet of BOTH demotes to the
other; removing the last facet drops the relay) lived as private
helpers in the CLI's RelayCommands. Any frontend that edits a NIP-65
list needs them, so they now live in quartz nip65RelayList as
AdvertisedRelayListMutations (applyFacet/addFacet/removeFacet/setFacet)
with commonTest coverage. Behavior unchanged; the CLI rewires to the
shared functions.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CP4kfLCa3wWtE8Khy21Pkj
Adds a confirm-then-pay flow so a 402 from a paid Blossom server can be settled
from the app instead of only being reported.
- quartz: BlossomPaymentProof (settled Cashu token / lightning preimage) with the
X-Cashu / X-Lightning retry headers; BlossomClient.mirror accepts a proof.
- BlossomPaymentHandler (Android): pays the challenge's BOLT-11 invoice via the
account's existing NIP-47 (NWC) wallet and returns the preimage — it never
handles keys or funds itself, only drives the connected wallet. Decodes the
invoice amount for display.
- Blob manager: a mirror that hits 402 now raises a payment prompt; a dialog shows
the amount and, on confirm, pays and retries the mirror, then continues with the
remaining servers. Cancel leaves the blob unmirrored.
Cashu-only servers and the composer upload path still surface a clear message;
auto-settlement there can reuse this handler next. Not yet validated against a
live paid server.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ckbnz1N94W1hnNC9xpsCNP
Extends Blossom support toward a full client on both the CLI and the mobile app.
Quartz (protocol):
- BlossomAuthorizationEvent: add t=media auth (BUD-05) and optional BUD-11
`server` domain scoping on every factory (stops replayable upload/delete tokens)
- BlossomServerUrl: mirror/media/list/report path builders, BUD-06 preflight and
BUD-07 payment header constants, and a lowercase bare-domain helper
- BlossomUploadResult: parse `ox` (BUD-05 original hash) and `nip94` (BUD-08)
- BlossomPaymentRequired: BUD-07 402 challenge model (Cashu/Lightning)
- BlossomReport: BUD-09 kind-1984 blob report reusing NIP-56 tag builders
Commons (shared JVM client, now in jvmAndroid so Android shares it too):
- BlossomClient gains mirror (BUD-04), list/delete (BUD-02), media (BUD-05),
preflight/has (BUD-06/01), report (BUD-09) and typed 402 handling
- BlossomAuth: media/list/delete passthroughs with server scoping
CLI (first-class):
- amy blossom now routes all HTTP through the shared client and adds `media`
and `report` verbs; auth tokens are scoped to --server
Android (first-class):
- uploads mirror to the user's other Blossom servers (BUD-04) best-effort
- new "Manage stored files" screen: per-server presence matrix (BUD-02 list +
BUD-01 HEAD), delete, mirror-to-missing, and report actions
Tests: quartz URL/auth/descriptor/payment parsing.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ckbnz1N94W1hnNC9xpsCNP
Web-based highlighter clients publish kind:9802 highlights with W3C Web
Annotation selectors (textquoteselector / textpositionselector /
rangeselector) instead of a NIP-84 `context` tag. These were previously
ignored, so the highlight rendered without its surrounding paragraph and
the "jump to page" link couldn't disambiguate repeated quotes.
- Parse the W3C textquoteselector into TextQuoteSelectorTag (exact/prefix/
suffix; a "-" or empty exact is treated as a placeholder since the quote
lives in .content).
- HighlightEvent.contextOrReconstructed() prefers an explicit `context`
tag and otherwise rebuilds the paragraph from prefix + content + suffix,
so the in-context bolding still works.
- Build a disambiguated Text Fragment URL (`#:~:text=prefix-,exact,-suffix`)
from the selector's prefix/suffix so the source link scrolls to the
correct occurrence.
The position/range selectors are left unparsed; they only matter for an
in-app live-page re-highlighter, which we don't have.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Di4UurD9SQGrpScX7uy2Kh
NostrSignerRemote extended NostrSigner(signer.pubKey), where `signer` is the
ephemeral NIP-46 transport keypair — so `pubKey` returned the transport key,
not the user's identity. Every self-encryption / self-authorship site keys off
`signer.pubKey`, so for bunker accounts this silently broke:
- private NIP-51 lists (private bookmarks / mute / follows / hashtags) and
NIP-37 drafts — an `if (signer.pubKey != event.pubKey)` guard short-circuits
(desktop: private bookmarks always empty);
- NIP-44 self-encrypted data (Concord list, Cashu) sealed to / read against
the wrong peer key.
Android is unaffected (no bunker path); desktop and CLI were affected.
Make `NostrSigner.pubKey` open and have `NostrSignerRemote` return the
bunker-resolved user key: `getPublicKey()` now caches it, and `bindUserPubkey()`
sets it eagerly for a reloaded account / stored identity. Internal transport
(the response-subscription `p` filter, request addressing) keeps using the
transport keypair explicitly, so it is unchanged. No-op for local/external
signers, where signer.pubKey already equals the account key.
Wired: desktop AccountManager.loadBunkerAccount binds the resolved pubkey; CLI
Context binds identity.pubKeyHex. amy's Concord-list decrypt workaround is
dropped — `newest.decrypt(ctx.signer)` now works for a bunker. Verified live:
`amy concord import` over a bunker account decrypts the kind-13302 list and
recovers Soapbox heldRoots [0,1].
Plan: quartz/plans/2026-07-17-nip46-remote-signer-self-pubkey.md
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Each kind:30382 GrapeRank card now carries two more public tags alongside
`rank`:
- `followers` — the number of the target's followers whose own score clears a
threshold (`--followers-threshold`, default 0.02), mirroring Brainstorm's
trusted-follower cutoff.
- `hops` — the shortest follow-graph distance from the observer (1 = a direct
follow), matching the `hops` field on Brainstorm's ScoreCard.
New `HopsTag` (the `followers`/`FollowerCountTag` already existed) is wired
through the ContactCardEvent tag accessors/builders. TrustGraph gains
`hopsFrom` (a follow-only BFS over the compact int-CSR) and
`trustedFollowerCounts`; the out-CSR now packs the relation code so a forward
walk can filter FOLLOW edges. The publisher's `reconcileLocal` takes a richer
`ScoredCard` and diffs the full (rank, followers, hops) triple, so a card
re-signs when any of them moves and older cards migrate onto the new tags once.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Xc3Wm4qVCrAGvSAotTUVt4
Relays that can't serve a request (a NIP-50 search-only relay pulled into
the feed, a write-only relay, a relay whose filter shape is rejected) were
being hammered with the same doomed REQs. Observed on a 40s cold start:
search.nos.today CLOSED 13-14x ("error: search filter is required") across
6 subscriptions, plus repeated "restricted: does not accept REQs" and
"unsupported: too many filters". The reconnect path replayed refused REQs on
every reconnect, and many different subscriptions kept hitting the same
capability wall.
Two complementary, purely-quartz mechanisms (so the app and Amy both benefit
with zero wiring):
- Per-subscription refusal memory (RequestSubscriptionState + PoolRequests):
a filter a relay CLOSES is not replayed to that relay across reconnects
until it meaningfully changes or a REQ succeeds (EOSE/event). Never applies
to auth-required (the auth subsystem re-signs and replays) or rate-limited
(the adaptive limiter spaces it out).
- Per-relay capability block (RelayReqRefusals): after 2 refusals, classify a
relay SEARCH_ONLY (suppress only non-search filters; genuine search REQs
still flow) or NO_READS (suppress all), from narrow substring markers that
deliberately avoid auth-conditional messages. A fully-blocked relay is
dropped from PoolRequests.desiredRelays so the pool disconnects it, closing
the idle socket rather than keeping it open with every REQ suppressed. A
SEARCH_ONLY relay stays connected while any subscription carries a search
filter for it, so the separate search path is unaffected.
Adds UNSUPPORTED to MachineReadablePrefix (relays send "unsupported:"; parse()
returned null on it before).
Device-verified: search.nos.today now Connecting -> OnOpen -> 2x Closed ->
Disconnected; sendit.nosflare.com (NO_READS) -> 3x Closed -> Disconnected;
feed event volume unchanged (kind-1 from 17 relays) - no coverage loss.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Third refinement from the Primal comparison — and the one that needed an
architecture change, not just UI.
Quartz: NostrConnectSignerService now fans each request into a child coroutine
under a Semaphore(maxConcurrentHandles=16) instead of handling them inline, so a
request awaiting a consent prompt no longer blocks other clients' auto-allowed
traffic and several prompts can be pending at once. Intake (dedup, staleness,
rate-limit, seen-id persistence) stays on the single consumer. Two guards keep
it safe: BunkerRequestProcessor serializes the actual crypto with a Mutex
(authorization — the prompt — runs unlocked, only sign/encrypt/decrypt holds the
lock) so an external NIP-55 signer never sees concurrent IPC ops; and
Nip46PermissionAuthorizer serializes first-connect consent so two connects can't
stack dialogs. Covered by BunkerRequestProcessorConcurrencyTest (crypto never
overlaps; a blocked prompt doesn't stall another client's signing).
Amethyst: SignerConsentCoordinator is now a shared pending StateFlow; one
SignerConsentActivity observes it and shows the rich single-request dialog (1
pending) or a batched checkbox list with select-all + a Remember toggle +
Allow/Deny selected (>1). Dismissing the sheet denies every still-open request
(fail closed).
Needs on-device validation (burst batching, no concurrent external-signer IPC,
fail-closed on dismiss) — see the device checklist.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
Reject sign/encrypt/decrypt when the identity signer is not writeable — the
account was logged out, is read-only, or lost its external (NIP-55) signer —
returning an `account unavailable` error instead of prompting the user or
hanging on a key that can't be used. Checked before authorization, so no
dialog is raised for a key we can't sign with. Public reads (get_public_key,
ping, get_relays) stay ungated.
Test: a non-writeable signer refuses a sign request without invoking the
signer or the authorizer.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
The 30s window shrank the restart re-sign problem but couldn't close it: a
relay replays stored ephemeral requests on re-subscribe, and the in-memory
dedup set is wiped on restart, so anything within the window came back.
Persist the recently-serviced kind-24133 event ids (bounded to 128) and seed
the service's dedup set from them on start, so a replay after an app restart
is dropped by EXACT event id. Chosen over a created_at high-water mark on
purpose: a global timestamp floor would wrongly drop a second connected app
whose clock lags behind another's, whereas id-matching is immune to client
clock skew. The `since` filter still bounds how far back relays replay.
- AccountSettings.nip46SeenRequestIds (persisted via putStringSet) + host-side
bounded LinkedHashSet, fed to NostrConnectSignerService.initialSeen and
advanced through onHandledId.
- Tests: a fresh request whose id was serviced last session is not repeated;
the serviced id is reported for persistence.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
Users on multiple relays were being asked to sign the same request repeatedly,
some minutes old. Root cause: kind-24133 is ephemeral, but many relays store
and REPLAY it every time the signer re-subscribes (a relay-set change,
reconnect, toggle, or rotation), and the in-memory dedup set is scoped to one
run() call so it's wiped on restart — the old requests then get signed again.
Gate requests by created_at (default 120s window):
- a `since` on the subscription filter so compliant relays never replay old
stored events, and
- a receive-side staleness drop for relays that ignore `since`.
The window must exceed realistic client/relay clock skew so a genuinely fresh
request is never dropped. Also corrected the seenCap KDoc, which called the
event-id dedup set a "request-id" set.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
NIP-46 lets the remote-signer (transport) key differ from the user's identity
key. Previously the bunker advertised and wrapped everything with the identity
key, so anyone watching the inbox relays could see kind-24133 traffic addressed
to the user's real npub and infer "this npub runs a bunker".
Now each account gets a dedicated, persisted transport keypair:
- NostrConnectSignerService wraps/unwraps the kind-24133 envelope with a
`transportSigner`; BunkerRequestProcessor keeps the identity signer for the
actual sign/encrypt/decrypt and answers get_public_key with the real npub
(disclosed only to a connected client, over the encrypted channel).
- The host mints + persists the transport key lazily (accounts that never
enable the signer mint nothing), advertises it in bunker:// and the
nostrconnect ack, and listens p-tagged to it.
- AccountSettings/LocalPreferences persist nip46TransportKey so the advertised
address stays stable across restarts.
Bonus: because the envelope is now wrapped with a LOCAL key, external NIP-55
(Amber) accounts no longer round-trip the external app for envelope crypto —
only the genuine signing request does. A new test asserts get_public_key
returns the identity, never the transport key. Unreleased feature, so no
migration needed. The CLI bunker keeps using the operator's own key (dev tool).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
Follow-ups to the audit:
- Abuse protection: the signer service now bounds its event queue
(DROP_LATEST) and rate-limits per author BEFORE decrypting — decryption can
be an external-signer (NIP-55) IPC round-trip, so a looping or hostile client
can no longer force one per event or grow the queue without limit. Fixed
window (default 40 requests / 10s per author, oldest authors evicted). The
limiter is touched only by the single consumer coroutine, so it needs no
locking. Covered by a headless test.
- logout now clears the client's persisted metadata/relays too (not just the
ledger grant), so a disconnected app stops being listened for after restart.
Not changed: get_public_key/ping stay ungated — gating them behind a prior
connect risks breaking clients that discover the pubkey at connect time, and
the pubkey is already public, so the enumeration leak is negligible.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
- NostrConnectSignerService now bounds its request-id dedup set (LinkedHashSet
with an evicting cap) so a long-lived bunker can't leak memory on the ids it
has seen.
- Add NIP-46 `logout`: the processor recognises the method, acks it, and calls
a new Nip46RequestAuthorizer.onLogout hook (default no-op) so a host can
revoke the app's grant when it disconnects.
- New NostrConnectSignerServiceTest drives full request→reply round trips
(connect/sign/logout + drop-if-not-addressed) through the service over a fake
relay client with passthrough signers — headless proof of the subscribe →
decrypt → dispatch → publish wiring.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
Adds the bunker/signer half of NIP-46 as reusable, signer-agnostic quartz
components so Amethyst can act as a remote signer for other apps:
- BunkerRequestProcessor: turns a decrypted BunkerRequest into the
BunkerResponse the client expects, performing the work through whatever
NostrSigner the account uses (local keypair or NIP-55 external app).
Signing/encryption/decryption are gated through a Nip46RequestAuthorizer;
public reads (get_public_key/ping/get_relays) are not.
- Nip46RequestAuthorizer: the permission boundary the host app plugs its
own trust model into (connect validation + per-op authorization).
- NostrConnectSignerService: subscribes to kind-24133 requests on a relay
set, decrypts, dispatches to the processor, and publishes the reply.
- NostrConnectURI: KMP-safe parse/build for bunker:// and nostrconnect://
pairing URIs (percent-encoded), shared by CLI/desktop/Android.
Unit tests cover the dispatch/authorization matrix with a fake signer (no
crypto) and the URI round-trips.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015FHr2mu5SiHwYNR7evYUuF
A live relayop.xyz kind-33301 invite bundle (vsk=6) decrypted correctly but
failed to open: its JSON diverged from quartz's CommunityInvite model on two
CORD-05 wire details, so decodeOrNull returned null and the invite reported
Unreadable ("This invite link can't be opened...") instead of joining.
- InviteChannel.key was required; a public channel (e.g. an unencrypted
`general`) carries no delivered grant key and some reference clients omit
the field. Default it to "" so a keyless channel no longer rejects the
whole bundle.
- icon was modeled strictly as an ImagePointer object; relayop emits a bare
public URL string for an unencrypted icon. Add LenientImagePointerSerializer
(a JsonTransformingSerializer) that lifts a string into ImagePointer(url=...)
on read while still serializing the canonical object form on write.
Adds a regression test driving the real live bundle + its fragment token,
asserting it now classifies as Live. This is a distinct interop case from the
existing vsk=8 mis-posted-registry test: here the sub-kind, token, and crypto
are all correct — only the JSON schema was too strict.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018iLfo68yVtG2ALkKB8tpAa
Addresses the four findings from the branch audit:
- Make the kotlinx codec (LimitsKSerializer, the iOS/native incoming path)
as lenient as the Jackson path: mistyped fields degrade to null, non-int
array elements are skipped, and a payload-less ["LIMITS"] frame yields an
empty message instead of throwing. Guard the payload access in
MessageKSerializer and MessageDeserializer likewise.
- Make the Jackson reads (LimitsDeserializer) type-checked so an explicit
JSON null or wrong-typed value stays null ("unspecified — keep previous")
instead of coercing to false/0. Both codecs now behave identically.
- LimitsMessage -> data class, so StateFlow.distinctUntilChanged in
RelayLimitsTracker suppresses no-op emissions when a relay re-advertises
identical limits, and tests get value equality.
- Drop the stale "NIP-22" labels (LIMITS is nostr-protocol/nips#1434, not
NIP-22) from LimitsKSerializer, MessageKSerializer and MessageSerializer.
Tests: added mistyped/null-field and payload-less coverage on both the
Jackson and kotlinx paths, plus a value-equality check. quartz:jvmTest
(RelayWireErgonomicsTest 11, KotlinSerializationMapperTest 57,
RelayLimitsTrackerTest 5) and amethyst play compile green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01464jkunWPtYhTReoc3fUQQ
Cleanup after adding LimitsMessage + the client-side cache:
- Remove the unused experimental LIMITS prototype
(experimental/limits/Limits.kt + LimitProcessor.kt). Its @Serializable
model duplicated LimitsMessage (minus auth_for_read/auth_for_write and
the Message wiring); the processor's clamp/reject logic is superseded by
the server-side LimitsPolicy and will be reincarnated as pure helpers on
LimitsMessage. Both were prototypes with no references (preserved in git
history).
- Rename the client accessory RelayLimits -> RelayLimitsTracker so it no
longer collides on simple name with the relay-server-side
nip01Core.relay.server.policies.RelayLimits (the operator-configured
limits a relay enforces and advertises). Updates AppModules and the test.
No behavior change. quartz:jvmTest (RelayLimitsTrackerTest 5/5) and
amethyst play-flavor compile are green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01464jkunWPtYhTReoc3fUQQ
Adds RelayLimits, a passive connection-listener accessory (modeled on
RelayAuthenticator) that caches the latest LIMITS each relay advertises
and publishes it as a Compose-stable StateFlow, so consumers can read a
relay's current rights/limits instead of only observing the raw message.
- RelayLimits: caches LimitsMessage per NormalizedRelayUrl, exposes
limitsFlow (StateFlow), get(url) and snapshot(); connection-scoped
(entry dropped on disconnect so stale limits don't leak).
- LimitsMessage marked @Immutable for Compose stability in the flow map.
- Wired into AppModules next to relayStats (Amethyst.instance.relayLimits).
- Tests: per-relay caching, later-replaces-earlier, independent relays,
drop-on-disconnect, and ignore non-LIMITS messages.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01464jkunWPtYhTReoc3fUQQ
Adds support for the relay-to-client LIMITS frame, which advertises the
connection's current rights and limits (can_read/can_write, auth
requirements, max_message_length, max_subscriptions, max_filters,
max_limit, POW, rate limits, required tags, etc.). Relays such as
wss://pipe.imwald.eu/ send it on connect and whenever rights change,
and Amethyst previously logged it as an unsupported message.
- New LimitsMessage model in quartz commonMain with every optional field.
- Jackson decode/encode (LimitsDeserializer + MessageSerializer) and the
kotlinx-serialization path (LimitsKSerializer + MessageKSerializer).
- Logs a summary line in RelayLogger.
- Tests: production payload parse, array/tag fields, empty object, and
round-trip parity between the Jackson and kotlinx codecs.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01464jkunWPtYhTReoc3fUQQ
Bugs:
- Metadata edit could re-root a subgroup or drop its children on a load race:
the edit ViewModel snapshotted parent/children at prefill and overrode the
Account-level live-read defaults. Children are no longer snapshotted (Account
reads the live child list at save time), and the parent is only overridden
when the user actually re-parents (parentTouched) — a plain rename can't
re-root or orphan children anymore, even if metadata hadn't loaded yet.
- Parent selector card cached a null channel via remember() and never
refreshed, so the parent's name/picture never loaded and the warm-up never
mounted. Now get-or-create + warm + observe the metadata flow.
- previousEventRefs could let a note with an unresolved author slip past the
self-exclusion and reference the sender's own event. Now requires a resolved
author.
- Assigning a relay-defined role replaced the member's whole role set while the
menu implied additive; now keeps existing roles (entry.roles + role.name).
- GroupNAddrInvite now also accepts a bare `invite=<code>` remainder if the `?`
is stripped upstream (+ test).
Performance:
- Subgroups bar mounted a full warm-up (metadata + content) subscription per
child chip — up to ~21 relay subscriptions per open group. Replaced with one
relay-directory subscription; chips read from cache.
- Parent picker recomputed the whole candidate scan on every recomposition
(each search keystroke) via a produceState initial-value argument; the scan
now lives only in the producer with a cheap empty initial.
spotless clean; quartz tests green; amethyst compiles.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Qst2JsmNYMvXitv2vxo4S
Follows up the subgroup protocol work with four NIP-29 compliance/UX gaps.
previous timeline references (spec §Timeline references):
- RelayGroupChannel.previousEventRefs(): first-8-char id prefixes of the most
recent events seen from the host relay, excluding the sender's own, capped at
the spec's 50-event window. Only draws from events actually received in the
channel so the host relay is known to have them.
- Populate the `previous` tag on all outgoing group events: kind-9 chat and
replies, kind-1111 minichat comments, kind-11 threads, and group replies.
Custom roles (kind 39003):
- Route SupportedRolesEvent onto the channel (LocalCache.consume + a
RelayGroupChannel.supportedRoles field) instead of only storing it.
- Members screen: when the relay advertises a role set, offer those roles when
assigning (admins), and show each member's real relay-assigned role label
instead of collapsing everything to admin/moderator. Falls back to the
built-in admin/moderator shortcuts when no 39003 is published.
Subgroup navigation (spec §Subgroups):
- RelayGroupSubgroupsBar: a self-hiding bar under the pinned bar showing a
breadcrumb up to the parent group and chips for the child subgroups (in the
relay's `child` order), each opening that group on the same host relay.
naddr invite codes (spec §Group identifier):
- GroupNAddrInvite parses the `naddr1…?invite=<code>` suffix; both the tap
handler (ClickableRoute) and the deep-link handler (MainActivity) now feed it
into the kind-9021 join request so a shared invite naddr auto-joins.
Tests for the naddr invite parser. spotless clean; quartz tests green;
amethyst compiles.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Qst2JsmNYMvXitv2vxo4S
NIP-51's kind:10000 mute list defines four entry types — `p` (pubkeys),
`word`, `e` (threads) and `t` (hashtags). Quartz parsed only the first
three, so `t` hashtag mutes written by other clients were silently
dropped: uncounted, invisible, and never applied to filtering.
Quartz:
- Add HashtagTag (`"t"`) implementing the MuteTag sealed interface, and
register it in MuteTag.parse/isTagged so it round-trips like the other
entry types.
- Add mutedHashtags()/mutedHashtagIds() TagArray helpers.
Filtering (commons):
- Add hiddenHashtags to LiveHiddenUsers plus isHashtagHidden(), and hide
notes carrying a muted hashtag in Note.isHiddenFor() (exact, case-
insensitive `t`-tag match — distinct from the existing substring word
scan).
Amethyst:
- Aggregate HashtagTag entries from the mute/block lists in
HiddenUsersState.
- MuteListState.hideHashtag/showHashtag + Account and AccountViewModel
wrappers, and observeUserIsMutingHashtag.
- Surface a Mute/Unmute hashtag action in the hashtag screen's options
overflow menu.
Tests: HashtagTagTest (parse/round-trip/MuteTag dispatch) and
NoteIsHiddenForTest cases for muted-hashtag hiding.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017giudm3gXumsxZmd3uMQc8
Implements the NIP-29 Subgroups feature merged upstream: groups can now be
organized into a parent/child tree, scoped per host relay.
Quartz:
- Add `parent`/`child` tag classes and TagArray (builder) helpers.
- GroupMetadataEvent (39000): parent()/children()/isRoot() accessors and
build params.
- EditMetadataEvent (9002): parent()/children() accessors and build params
(a 9002 re-carries the full child list, per spec, or the relay rejects it).
- SubgroupTree: assembles a relay's flat 39000 set into the hierarchy —
structure follows each group's parent tag, sibling order follows the
parent's child-tag order, orphans surface as roots, and malformed cycles
are broken rather than looping.
- NIP-11: advertise/detect subgroup support via `nip29: { subgroups: true }`,
with a `subgroups()` builder DSL helper.
- Tests for tag round-trips, tree assembly, ordering, orphans and cycles,
plus NIP-11 serialization.
Amethyst:
- RelayGroupChannel: parentGroupId()/childGroupIds()/isSubgroup() reading the
latest metadata.
- Account.editRelayGroupMetadata: preserve the group's current parent and full
children list on a plain metadata edit so an admin renaming a subgroup no
longer detaches it (or gets rejected for dropping children).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011Qst2JsmNYMvXitv2vxo4S
Follow-up making the invite redeemer match the CORD-05 §2 spec for the
addressable invite coordinate (33301, link_signer, d=""):
- vsk=6 → live bundle (open with the link token)
- vsk=9 → revocation tombstone: the newest event wins, so a tombstone buries
even a stale, still-openable copy on another relay ("a fetcher finds the
grave instead of keys"). Amethyst previously never checked for this, so a
revoked link failed generically.
- anything else present (e.g. a mis-posted registry vsk=8, the shape of the
relayop.xyz link that hung) → unreadable
- nothing on any relay → absent
New pure `ConcordInviteBundle.classify(wraps, token): InviteBundleStatus` in
quartz (next to parse/validate), wrapped by `ConcordActions.classifyInvite`,
and mapped by `Account.joinConcordViaInvite` to the `ConcordInviteResult`
cases — including a new `Revoked` outcome with its own message and no futile
retry. Crypto is unchanged and already matches the spec
(hkdf(token,'concord/invite-key') → NIP-44 → snake_case CommunityInvite).
Adds ConcordInviteClassifyTest covering live / revoked (order-independent) /
unreadable / absent, plus the real relayop.xyz vsk=8 event → Unreadable.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015KngFNwrQDLYa9QW1f5RRD
Makes location channels reachable and manageable natively:
- quartz GeohashChannelLevel: the named Bitchat precision levels
(region2/province4/city5/neighborhood6/block7/building8) with cellFor
truncation, so one location fix yields the whole ladder. Tested.
- LocationState.preciseGeohashStateFlow: an 8-char (building) location flow
alongside the existing 5-char one (untouched, so the "around me" feed is
unchanged); channels truncate it per level.
- NewGeohashChatScreen (Route.NewGeohashChat): join a cell from your current
location (region -> building, with place names via LoadCityName) or by typing
a geohash. Joining adds it to the kind-10081 geohash list (followGeohash) and
opens the chat. Includes a "Your channels" section that lists joined cells with
open + Leave (unfollow) -- the management surface the list previously lacked
(only the per-cell Follow toggle on the notes screen existed).
- Adds a "Location channel" option to the Messages "+" new-conversation chooser.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0172JoMccseEKenyWan6txWV
Implements NIP-29 message pinning (nostr-protocol/nips#2379):
Protocol (quartz):
- GroupPinnedEvent (kind 39005): relay-signed pinned-message list, `d`
group id + ordered `e` ids.
- UpdatePinListEvent (kind 9010): moderator `update-pin-list` write,
carries the full list so pin/unpin/reorder/clear are one submission.
- Register both in EventFactory; add a pinnedEventIds tag helper.
Model + cache:
- RelayGroupChannel now folds the pin list (pinnedEventIds / isPinned)
with the same createdAt-supersede guard as the roster.
- LocalCache consumes 39005 into the channel and stores the 9010 write;
39005 added to the group's metadata REQ filter so pins load.
Publish path:
- Account.pin/unpin/updateRelayGroupPins + AccountViewModel wrappers.
UI (non-intrusive, self-hiding):
- Collapsed pinned-message bar under the top bar: shows the current pin,
N-count cycling, tap to jump to the message in-feed (hoisted jump
request threaded through the shared chat feed view). Renders nothing
when the group has no pins.
- Moderator-only Pin/Unpin action under "Show more" in the chat message
bottom drawer, gated on membership.canModerate().
- Small pin glyph on pinned bubbles' footer.
Tests: quartz build/parse round-trip for both kinds; channel pin-fold
supersede/replace/clear semantics.
Geohash chat stays anonymous by default (posting under the real npub would
broadcast the user's location tied to their public identity), but instead of a
random per-device seed the throwaway identity is now derived from the account:
- GeohashKeyDerivation.accountSeed(privKey) = HMAC(privKey, versioned label);
per-geohash keys are HMAC(seed, geohash). The result is stable across the
user's devices and recoverable from the account, yet publicly unlinkable to
their npub and across cells (all one-way).
- GeohashChatIdentity picks the account-derived seed for local-key accounts and
falls back to the random encrypted device seed for NIP-46/NIP-55 signers whose
key we can't reach.
Tests cover determinism, account-specificity, and cross-cell unlinkability.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0172JoMccseEKenyWan6txWV
Adds the wire-level protocol for interoperating with Bitchat's Nostr location
channels:
- GeohashChatEvent (kind 20000): plain-text public geohash message with a single
exact ["g", geohash] tag plus optional ["n", nickname] and ["t","teleport"].
- GeohashPresenceEvent (kind 20001): presence heartbeat carrying only the g tag.
- GeohashKeyDerivation: deterministic, unlinkable per-geohash ephemeral identity
(HMAC-SHA256(deviceSeed, geohash||counter) with retry + SHA-256 fallback).
- Registers both kinds in EventFactory; NIP-13 PoW reuses the existing PoWTag.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0172JoMccseEKenyWan6txWV
Concord/NIP-C7 image messages from interop clients (e.g. Ditto/Soapbox
Armada) attach the image purely as a NIP-92 `imeta` tag and leave the
message content empty. Amethyst's shared renderer only shows media whose
URL appears in the text, so those images never rendered — even though the
encrypted blob's key/nonce were already registered for transparent
decryption at ingest.
RenderChat now appends any imeta URL missing from the content before
handing it to the media renderer (symmetric to ChannelChat.imageMessage,
which appends the URL on send), so imeta-only attachments display.
Also forward the thumbhash on the encrypted-image imeta: FileHeader
computes both blurhash and thumbhash, but only blurhash was being sent.
encryptedImageImeta now emits an additive `thumbhash` field and
toConcordImeta forwards the computed value, so receivers can paint a
placeholder while the blob decrypts.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016TGjBiJL2fvTR37eePBKrG
Vitor PamplonaandClaude Opus 4.8
