buzz-acp: startup config summary omits no_mention_filter, so logs can't confirm the flag took effect
- Dominant language
- Rust
- Stars
- 32.7k
- Forks
- 4.3k
- Avg merge
- 1d 13h
- Merged PRs (30d)
- 253
Description
**Describe the bug**
The one-line startup summary that buzz-acp logs (relay, pubkey, subscribe mode, dedup, respond_to, …) does not include the `no_mention_filter` state. A unit running `--subscribe mentions --no-mention-filter` logs exactly `subscribe=Mentions` — byte-identical to a mention-gated deployment.
Since the flag materially changes both the relay subscription (the `#p` filter is dropped) and the operator's cost model (every allowed-author message becomes a potential model turn), the summary line is the natural place to attest it — it's the only at-a-glance config attestation an operator gets from the journal.
**Where** (verified on `main` @ `ac4fa13`, 2026-08-01)
`crates/buzz-acp/src/config.rs` — `summary()` at line 1111; the format string (line 1126) prints `subscribe={:?}` (the mode enum only). `no_mention_filter` appears nowhere in the line.
**Steps to reproduce**
1. Run buzz-acp with `--subscribe mentions --no-mention-filter`.
2. Read the startup line in the journal:
```
INFO buzz_acp: buzz-acp starting: relay=wss://… subscribe=Mentions dedup=Queue meh=Queue … respond_to=allowlist(1)
```
3. Compare with a unit running plain `--subscribe mentions`: the lines are indistinguishable.
**Expected behavior**
The summary includes the mention-filter state, e.g. `subscribe=Mentions no_mention_filter=true` (or `mention_filter=off`). To confirm the flag was active we currently have to run an end-to-end probe — publish an untagged event from an allowlisted author and watch for a turn — because there is no cheaper signal.
**Version and platform**
- Buzz version: buzz-acp from `ghcr.io/block/buzz:main` @ `sha256:13ddb4c12f4c0b8147cece717a57013a45fdb1610bd072d7e383ac79a070ddd9`; re-verified in source on `main` @ `ac4fa13` (2026-08-01)
- OS: Ubuntu 24.04, systemd unit
**Logs / additional context**
Related issues (searched open and closed): none found. I have a small PR ready for this (add the field to `summary()` plus a regression test) and will open it right after filing.
Contributor guide
Research direction
Start in crates/buzz-acp/src/config.rs at summary(), especially the format string around line 1126. Compare startup summaries for mentions mode with and without --no-mention-filter, then add a regression test as described in the issue. Done means the log clearly distinguishes the filter state without requiring an end-to-end probe.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- observability
- Issue type
- Bug
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Quiet
- Clarity
- Clearly specified
- Newbie friendliness
- 38/100