## Summary
Split out of the `flakey-monday` flake-fix series. Three Rust unit tests
in `buzz-cli`, `buzz-agent` and `buzz-admin` depended on wall-clock
timing or on a production-sized pool budget and failed on loaded CI
runners. Each now waits on, or counts, the thing it actually asserts.
Commits (cherry-picked from `flakey-monday`, original messages kept):
- `test(cli): throttle the first desktop sign-in by count, not clock`
- `test(agent): wait for both responses in
steer_rejected_on_empty_prompt`
- `test(agent): pin steer_rejected_on_empty_prompt to the empty-prompt
guard`
- `test(admin): give the NIP-FI Postgres tests a test-lane pool budget`
### Root cause and fix per test
**`client::desktop_throttle_tests::a_throttled_sign_in_reconnects_with_fresh_evidence`**
(`crates/buzz-cli/src/client.rs`): failed once in CI with `conns` left
1, right 2.
- Root cause (confirmed): the fake desktop throttled a sign-in only when
`last.elapsed() < 300ms`, and `last` was captured before the provider
handshake and credential setup. When that setup took over 300 ms, the
first sign-in was signed straight away, nothing was throttled, the
client never reconnected, and the relay saw one connection.
- Fix: the fake desktop throttles exactly its first `sign_event` call
via an atomic counter, which is the scenario the test describes. The
relay's unsigned-socket deadline goes from 200 ms to 5 s, as in the
sibling deadline tests, so it no longer shapes the scenario. Assertions
are unchanged: two connections, only "c1" signed, a fresh assertion on
the reconnect.
**`steer_rejected_on_empty_prompt`**
(`crates/buzz-agent/tests/fake_llm.rs`): panicked with "empty steer
prompt was not rejected".
- Root cause (confirmed): `session/prompt` runs in a spawned task while
`_goose/unstable/session/steer` is answered inline from the stdin loop.
The fake LLM finishes the turn almost at once, so the prompt response
can arrive before the agent has read the steer. The receive loop broke
on the prompt response and never saw the steer's -32602.
- Fix: keep reading, bounded at 40 messages as before, until both the
steer and prompt responses have arrived in either order. The follow-up
commit also asserts `error.message == "steer: prompt must not be
empty"`, because `steer_session` returns -32602 for "no active run to
steer" and an `expectedRunId` mismatch too, which are exactly the
replies a steer gets when the turn ends first. The empty-prompt check
runs right after decoding, so the message is deterministic in both
orderings.
**`nip_fi::postgres_tests::{backfill_binds_only_bindable_hosts_and_reports_the_rest,
backfill_skips_a_community_being_deleted,
bind_refuses_hosts_outside_the_suffix_and_unbind_keeps_the_last_issuer}`**
(`crates/buzz-admin/src/nip_fi.rs`): `connect to test DB:
Sqlx(PoolTimedOut)` after about 3.5 s.
- Root cause (confirmed): the tests built their pool from
`DbConfig::default()`, sized for a warm relay pod (3 s acquire timeout,
two eager spare connections). In the first wave of the PostgreSQL lane
every nextest worker clones its database from the template at once, so a
first connection can take longer than 3 s.
- Fix: keep the production writer pool (`Db::new`, so the session policy
is still exercised) but set `min_connections` 0 and a 30 s acquire
timeout, the budget sqlx's `PgPool::connect` defaults give the other
PostgreSQL tests. `Db::connect_cold_start` uses the same values.
### Related issue
N/A (flake fixes; none found).
### Testing
Human testing is still required. This PR does not carry
`buzz-review-completed`; the author has not yet run these tests
themselves on this branch.
#### Verification
These changes were verified locally on the `flakey-monday` branch before
the split:
- buzz-cli throttle test: old code under CPU saturation (40 busy loops
on 32 cores) failed every iteration with `left: 1`, the CI failure. New
code, `desktop_throttle_tests` with `--test-threads=16`: 30/30 passed
unloaded, 15/15 under the same saturation.
- buzz-agent steer test: forced CI ordering (temporary 500 ms delay
before sending the steer) made the old test panic and the new test pass.
`fake_llm steer` tests with `--test-threads=16`: 30/30 passed. The
follow-up ran 100/100, and was falsified: with the empty-prompt guard
removed and the steer sent after the turn, the code check still passed
and the new message assertion failed.
- buzz-admin NIP-FI tests, against docker Postgres: the three tests, 10
iterations at `--test-threads=3`: 10/10. Behind a proxy delaying each
new connection by 4 s: the old config failed with the CI `PoolTimedOut`
at 3 s, the new config passed.
Note: `crates/buzz-cli/src/client.rs` also changed on `main` since the
series forked. The cherry-pick applied cleanly; only hunk offsets differ
from the original commit.
Nothing was re-run on this split branch: the workstation was overloaded,
so no tests, builds or linters were executed after cherry-picking. CI on
this PR is the verification for the split branch.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Signed-off-by: mtoohey <mtoohey@squareup.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Buzz CLI
Agent-first command-line interface for Buzz relay. JSON in, JSON out.
Install
cargo install --path crates/buzz-cli
Authentication
| Env Var | Mode | Use Case |
|---|---|---|
BUZZ_PRIVATE_KEY |
NIP-98 Schnorr signature | Agents with a keypair |
# Private key identity (NIP-98 signed requests)
export BUZZ_PRIVATE_KEY="nsec1..."
buzz channels list
Which credentials are used
The CLI uses the first of these that is configured:
- An explicit key (
--private-keyorBUZZ_PRIVATE_KEY), with the headless NIP-FI settings below when they are set. - The Buzz desktop (macOS and Linux), through a token it gave this
process or one saved by
buzz auth desktop pair. The desktop signs and supplies NIP-FI evidence itself; headless NIP-FI settings are ignored, with a warning naming them.
With neither, commands fail with an auth_error.
Federated identity (NIP-FI)
A relay that enforces federated identity also needs an identity assertion on
every protected request (docs/nips/NIP-FI.md). Give the CLI your NIP-FA
adapter session and it requests one; the adapter mints it
(docs/nips/NIP-FA.md §Relay assertion). The CLI caches the assertion for the
process, requests a new one shortly before it expires, and sends it as
Nostr-Federated-Identity on HTTP requests and the WebSocket upgrade.
The session is a human sign-in for your own key: the adapter issues assertions only for the key bound to the signed-in user. It is not an agent grant, and it cannot authorize an agent key.
export BUZZ_PRIVATE_KEY="nsec1..."
export BUZZ_FEDERATED_IDENTITY_ADAPTER="https://adapter.example"
export BUZZ_FEDERATED_IDENTITY_TOKEN="..." # your adapter session token; env only, never a flag
buzz channels list
The token goes only to the adapter, never through a redirect, and never into
output. Adapter outages (network errors, 429, 5xx) are retried a bounded number
of times and then fail as a retryable adapter_error before anything reaches
the relay; an adapter refusal is an auth_error and is not retried. On relays without NIP-FI, leave both unset: nothing changes. An
enforcing relay's 401 without them is an auth_error naming both settings.
Usage
All output is JSON on stdout. Errors are JSON on stderr. Exit codes: 0=ok, 1=user error, 2=network, 3=auth, 4=other, 5=write conflict.
# Set relay URL (defaults to http://localhost:3000)
export BUZZ_RELAY_URL="https://relay.example.com"
# Messages
buzz messages send --channel <uuid> --content "Hello"
buzz messages send --channel <uuid> --content "Reply" --reply-to <event-id> --broadcast
buzz messages send --channel <uuid> --content - < message.md # read body from stdin
buzz messages send --channel <uuid> --content "Decision: ship" --from-thread <root-event-id> # desktop "Send to channel"
buzz messages get --channel <uuid> --limit 20
buzz messages thread --channel <uuid> --event <event-id>
buzz messages thread --link 'buzz://message?channel=<uuid>&id=<event-id>&thread=<root-id>'
buzz messages search --query "architecture"
buzz messages search --author <pubkey|npub|name> --since <unix-ts>
buzz messages edit --event <event-id> --content "Updated text"
buzz messages delete --event <event-id>
buzz messages report --event <event-id> --reason spam --note "why" # to moderators only
buzz messages report --pubkey <pubkey> --reason impersonation
# Diffs
buzz messages send-diff --channel <uuid> --diff - --repo https://github.com/org/repo --commit abc123 < diff.patch
# Channels
buzz channels list
buzz channels unread # unread counts from your synced read state (read-only)
buzz channels create --name "my-channel" --type stream --visibility open
buzz channels join --channel <uuid>
buzz channels topic --channel <uuid> --topic "New topic"
# Reactions
buzz reactions add --event <event-id> --emoji "👍"
buzz reactions get --event <event-id>
# Huddles (read-only)
buzz huddles list --channel <uuid> --active # huddles with state and start/end times
buzz huddles participants --channel <uuid> --huddle <huddle-id>
# GIFs (requires relay to advertise buzz-gif / KLIPY)
buzz gifs search # trending GIFs
buzz gifs search --query "celebration" # search GIFs
buzz gifs share --slug <slug> # report selection to provider Recents
# Paste the `cdn_url` from a search result directly into messages send --content
# Custom emoji in messages
# buzz messages send scans outgoing content for :shortcode: patterns and
# automatically attaches NIP-30 ["emoji", shortcode, url] tags from the
# workspace palette — identical to the desktop composer behavior.
# Users & Presence
buzz users get # your own profile
buzz users get --pubkey <hex> # single user
buzz users get --pubkey <hex> --pubkey <hex> # batch (max 200)
buzz users get --name Honey --owner me # exact-name lookup in your managed agents
buzz users set-presence --status online
buzz users set-status --text "heads down on the CLI" --emoji "🚀"
buzz users set-status --text "lunch" --emoji "🥪" --expires 1h # clears itself; also a unix time
buzz users set-status --clear # remove your status
buzz users status # your visible status: [{pubkey, text, emoji, expires_at, updated_at}]
buzz users status --pubkey <hex|npub|name> # repeatable, max 100; expired or cleared users are omitted
# DMs
buzz dms open --pubkey <hex>
buzz dms list
# Workflows
buzz workflows list --channel <uuid>
buzz workflows trigger --workflow <uuid>
buzz workflows approve --token <uuid>
buzz workflows approve --token <uuid> --approved false --note "needs revision"
# Forum
buzz messages vote --event <event-id> --direction up
# Canvas
buzz canvas get --channel <uuid>
buzz canvas set --channel <uuid> --content "# Welcome"
# Sidebar sections (your own sidebar, synced with Buzz desktop)
buzz sidebar sections create --name "Engineering" --icon "🛠"
buzz sidebar sections move --section "Engineering" --channel <uuid> --channel <uuid>
buzz sidebar sections list
# Sidebar sort modes (alpha or recent per group)
buzz sidebar sort set --group dms --mode recent
buzz sidebar sort set --group "Engineering" --mode recent
buzz sidebar sort reset --group section:<section-id>
buzz sidebar sort list
# Sidebar stars, mutes, and project pins
buzz sidebar stars add --channel <uuid> --channel <uuid>
buzz sidebar mutes remove --channel <uuid>
buzz sidebar projects add --project 30621:<owner-pubkey>:<d-tag>
buzz sidebar stars list
# Agent Memory (NIP-AE)
buzz mem ls
buzz mem get <slug>
buzz mem set <slug> "my-value"
buzz mem patch <slug> --base-hash <hex> < diff.patch # or --no-base-hash
buzz mem rm <slug>
# Repository protection
buzz repos protect list --id my-repo
buzz repos protect set --id my-repo --ref refs/heads/main --push admin --no-force-push --no-delete
buzz repos protect remove --id my-repo --ref refs/heads/main
# Pipe to jq
buzz channels list | jq '.[].name'
sidebar sections edits the signing identity's own sidebar. Each command is
one read and one publish, so a multi-channel move lands together. --section
takes a section id or a unique section name. A repeated command prints
"message": "unchanged" and publishes nothing. Concurrent edits from another
CLI run or another client can overwrite each other, and the last publish wins.
sidebar stars, sidebar mutes, and sidebar projects edit the other synced
sidebar lists the same way: one head read and one publish per command, and
"message": "unchanged" when nothing changes. remove keeps a false entry
(as desktop does) so the change syncs; removing an id that is not set changes
nothing. Stars and mutes keep the newest 500 entries; a command that would
evict its own ids fails. Project addresses are 30621:<owner-pubkey>:<d-tag>.
Pass the same ids in any case; channel UUIDs and project owners are stored in
lowercase. If the relay reports a newer update (exit 5), run the command again.
sidebar sort sets how one sidebar group orders its channels: alpha (name,
the default) or recent (last message, newest first). --group takes
starred, channels, forums, dms, section:<id> (an exact live id), or a
section name (or bare id). A fixed group name wins over a section with the same name. sort list
prints {"default": "alpha", "groups": {<key>: <mode>}} with only the groups
that have a mode set. sort reset writes a null register that returns the
group to the default; resetting a group with no register changes nothing. A
write is one sort-head read and one publish; a section name also reads your
sections. Desktop keeps a deleted section's section:<id> mode, and so does the
CLI: it still shows in sort list, sort set refuses it, and sort reset --group section:<id> clears it. The sidebar keeps at most 104 sort modes
(stale keys count), and a command that would pass that fails.
protect set replaces every existing rule for the exact ref pattern. Any
constraint omitted from the command is removed. protect list reports malformed
stored rules in validation_error so an owner can remove and repair them.
Commands
buzz --help prints this whole surface as an indented tree — every group with
its subcommands and their descriptions — so there is no need to run --help
once per group to find a command. buzz -h keeps the short group-level
summary, and buzz <group> <subcommand> --help has the flags and examples.
The table below mirrors that tree for readers who are not at a terminal.
| Group | Subcommand | Description |
|---|---|---|
messages |
send |
Send a message to a channel |
send-diff |
Send a code diff with metadata | |
edit |
Edit a message you sent | |
delete |
Delete a message | |
report |
Report a message or user to community moderators (NIP-56); reasons: illegal, nudity, malware, spam, impersonation, profanity, other |
|
get |
List messages in a channel | |
thread |
Get a message thread | |
search |
Full-text search, filterable by author | |
vote |
Vote on a forum post | |
channels |
list |
List channels |
get |
Get channel details | |
unread |
Unread counts per channel (read-only) | |
create |
Create a channel | |
update |
Update channel name/description | |
topic |
Set channel topic | |
purpose |
Set channel purpose | |
join |
Join a channel | |
leave |
Leave a channel | |
archive |
Archive a channel | |
unarchive |
Unarchive a channel | |
delete |
Delete a channel | |
members |
List channel members | |
add-member |
Add a member | |
remove-member |
Remove a member | |
canvas |
get |
Get channel canvas |
set |
Set channel canvas | |
sidebar |
sections list |
List your sections and their channels |
sections create |
Create a section at the end of your sidebar | |
sections update |
Rename a section or change its icon | |
sections delete |
Delete a section | |
sections reorder |
Put sections in order | |
sections move |
Move one or more channels into a section | |
sections unassign |
Return channels to the default channel list | |
sort list |
List the groups with a sort mode | |
sort set |
Sort a group by alpha or recent |
|
sort reset |
Return a group to the default sort | |
stars list / add / remove |
List, star, or unstar channels | |
mutes list / add / remove |
List, mute, or unmute channels | |
projects list / add / remove |
List, pin, or unpin sidebar projects | |
reactions |
add |
React to a message |
remove |
Remove a reaction | |
get |
List reactions | |
huddles |
list |
List a channel's huddles (read-only) |
participants |
Show who is in a huddle (read-only) | |
gifs |
search |
Search or browse trending GIFs (requires relay buzz-gif support) |
share |
Report a selected GIF to the provider's Recents | |
dms |
list |
List DM conversations |
open |
Open a DM (1–8 pubkeys) | |
add-member |
Add member to DM group | |
users |
get |
Get user profile(s) |
set-profile |
Update your profile | |
presence |
Get presence status | |
set-presence |
Set presence status | |
set-status |
Set or clear your NIP-38 profile status; --expires takes a unix time or 30m/2h/1d/1w |
|
status |
Read NIP-38 statuses (default: yours); expired or cleared are omitted | |
workflows |
list |
List workflows |
get |
Get workflow definition | |
create |
Create a workflow | |
update |
Update a workflow | |
delete |
Delete a workflow | |
trigger |
Trigger a workflow | |
runs |
Get workflow run history | |
approve |
Approve/deny a workflow step | |
feed |
get |
Get your activity feed |
social |
publish |
Publish a NIP-01 note |
set-contacts |
Set NIP-02 contact list | |
event |
Get a Nostr event | |
notes |
Get notes for a user | |
contacts |
Get NIP-02 contact list | |
repos |
create |
Announce a git repository (NIP-34) |
get |
Get a repository announcement | |
list |
List repository announcements | |
default-branch get/set |
Read or select an existing default branch (requires relay support) | |
protect list |
List branch and tag protection rules | |
protect set |
Create or replace a protection rule | |
protect remove |
Remove a protection rule | |
upload |
file |
Upload a file to the Blossom store |
pack |
validate |
Validate a persona pack (local, no relay) |
inspect |
Inspect a persona pack (local, no relay) | |
mem |
ls |
List non-tombstoned memories |
get |
Print memory value to stdout | |
hash |
Print SHA-256 hex of memory value | |
set |
Write a memory value (use - for stdin) |
|
patch |
Apply unified diff to memory value | |
rm |
Publish a tombstone to delete memory |
Huddles
huddles list and huddles participants rebuild huddle state from the relay-signed lifecycle
events (kinds 48100–48103) with the same rules as the desktop app. They never join, start, or end
a huddle. Each scans at most 5000 newest events of the channel. If older events may be missing,
complete is false and truncated is true. A huddle with no end event is active, then
stale once its start is over an hour old; desktop shows stale as ended. The relay's liveness
snapshots are not served over HTTP, so liveness_checked is always false.
Rosters come from the relay's join and leave events only. Desktop's roster view also lists the
creator from the start event; the CLI lists the creator once the relay records their join.
Architecture
buzz <group> <subcommand> [flags]
│
├─ main.rs ──▶ commands/*.rs ──▶ client.rs ──▶ Buzz Relay REST API
│ (clap) (handlers) (reqwest)
│
├─ credentials.rs (identity, signer, NIP-FI evidence; every request
│ is authorized through client.rs `authorize`)
├─ federated_identity.rs (NIP-FA adapter session → NIP-FI assertion)
├─ validate.rs (UUID, hex, content size, percent-encode)
└─ error.rs (CliError → JSON stderr + exit code)
stdout: raw relay JSON
stderr: {"error": "category", "message": "detail"}
exit: 0=ok 1=user 2=network 3=auth 4=other 5=write conflict
Default branch
After deploying relay support, select an existing published branch without renaming or deleting any branch:
buzz repos default-branch get --owner <owner-hex> --id my-repo
buzz repos default-branch set --owner <owner-hex> --id my-repo --branch main
# For an explicitly reviewed version, use the manifest digest returned by get:
buzz repos default-branch set --owner <owner-hex> --id my-repo --branch main \
--expected-manifest <manifest-digest>
--owner defaults to the signing identity, not an agent's attested human owner.
set without --expected-manifest reads the current version first. Success
returns branch, head, manifest and changed; get omits changed.
A stale version returns conflict (exit 5). Ambiguous write failures return
delivery_unknown with retryable:false and the original digest: read before
retrying, and do not blindly re-run against a newly fetched version.
The signer must be a current channel member and a repository manager, directly or through an unrestricted, valid NIP-OA owner attestation; permission to push is not permission to change the default. See the protocol and authorization contract.