fix(discord): distinguish 30032 cap error from generic sync failures

[itsXactlY]

Problem

On loaded installs, the inner _safe_sync_slash_commands can hit Discord's hard 100-global-application-commands cap (HTTP 400 / code 30032). Today this surfaces in the gateway log as:

[Discord] Slash command sync failed: 400 Bad Request (error code: 30032):
  Maximum number of application commands reached (100).

…followed by a full stack trace — identical to every other sync failure (timeouts, network errors, random HTTP 500s). The actual cause is buried under noise that looks like a real bug.

This is especially bad because:

• It triggers AFTER _record_command_sync_attempt already wrote a "we tried" entry, with no matching success / rate-limit record.
• The operator has no hint that the cap was the actual cause or how to fix it (drop plugins / trim COMMAND_REGISTRY).
• The outer except Exception swallows it so the gateway survives, but the log is misleading.

The pre-flight registration-time cap (_DISCORD_MAX_APP_COMMANDS = 100 at line 39) protects against hitting 30032 on registration, but doesn't make the sync resilient if 30032 surfaces for any reason (fetch/upsert drift, external command registration, plugin race conditions).

Fix

Detect the cap error in the inner except of _run_post_connect_initialization (where _safe_sync_slash_commands is awaited) BEFORE the rate-limit fallback. Emit a distinct, actionable warning naming the cause and the remediation. Never re-raises, so the outer defensive except never sees it.

Detection mirrors _is_discord_rate_limit:

• prefers exc.code == 30032 (set by discord.py's HTTPException)
• falls back to status == 400 + "Maximum number of application commands reached" in exc.text so older discord.py forks / mocks / exotic transports are still covered.

Diff

 plugins/platforms/discord/adapter.py  |  36 +++++++++++++
 tests/gateway/test_discord_connect.py |  90 +++++++++++++++++++++++++++++++++++
 2 files changed, 126 insertions(+)

126 lines added, 0 deleted, 1 commit. No churn.

Tests

Test	What it proves
test_post_connect_initialization_logs_cap_error_with_distinct_message	A 30032 mid-sync emits "cap reached" log line, does NOT re-raise, does NOT fall through to the outer "sync failed" branch
test_is_discord_command_cap_error_detects_30032	Helper discriminates 30032 from generic 400s (code 50035) and 429s; covers both code-based and text-based detection paths

$ bash scripts/run_tests.sh tests/gateway/test_discord_connect.py
[100.0% | 20/20] ✓ 2.5s

$ bash scripts/run_tests.sh tests/gateway/
314 files, 6858 tests passed, 0 failed (100% complete) in 58.4s

What this does NOT do

• Does NOT change the registration-time cap (_DISCORD_MAX_APP_COMMANDS = 100) — that already exists.
• Does NOT change _safe_sync_slash_commands — the existing diff-based sync logic is preserved.
• Does NOT introduce new state files, sentinels, or summary dict fields.

It's a pure logging-clarity change in the inner except of _run_post_connect_initialization, plus a focused unit test for the discriminator helper.

[teknium1]

Thanks for the focused diagnostic improvement. Current main still routes a non-rate-limit exception from _safe_sync_slash_commands() through the generic traceback logger (plugins/platforms/discord/adapter.py:1705-1708,1745), so the premise remains valid.

Problems

• plugins/platforms/discord/adapter.py:1315 labels 30032 as HTTP 30032. The PR's helper docstring at :1245 correctly describes this condition as HTTP 400 / error code 30032. The warning should preserve that distinction so the new diagnostic is accurate.

Suggested changes

• Log HTTP 400 / error code 30032 (or Discord error code 30032) and assert that wording in tests/gateway/test_discord_connect.py.

This is an automated hermes-sweeper review.

[teknium1]

30032 is Discord's application error code, not the HTTP status; the detector's docstring above correctly describes this as HTTP 400 / error code 30032. Please correct the warning so the new actionable diagnostic is itself accurate.

[itsXactlY]

Good catch on the wording — fixed, and the whole change is rebuilt on current main in #75696.

This branch is ~7.8k commits behind and conflicting, so rebasing it isn't meaningful; #75696 reimplements it against main, where the premise still holds (the non-rate-limit path still reaches the generic traceback logger at plugins/platforms/discord/adapter.py:1990).

• HTTP 30032 → HTTP 400 / error code 30032. 30032 is a Discord JSON error code, not an HTTP status. The code is now a named constant _DISCORD_COMMAND_CAP_ERROR_CODE beside the existing _DISCORD_MAX_APP_COMMANDS, and test_cap_error_logs_http_400_and_error_code asserts "HTTP 400" and "error code 30032" as separate fragments so the distinction can't regress.
• Detection now follows the house style of the neighbouring _is_discord_unknown_interaction() — it also reads the code from a JSON data payload and prefers exc.status over exc.response.status, matching how discord.py shapes HTTPException. Still narrow: a bare 400 proves nothing (50035 is also a 400), so it needs the code, or a 400 plus the cap message for older forks/mocks. Unrelated failures keep raising with their traceback — there's a test for that too.

7 tests in TestDiscordCommandCapDiagnostic, all passing; -k discord over tests/gateway diffed against untouched main shows the identical pre-existing 4-failure set plus the 7 new passes.

Happy to close this one in favour of #75696.

[GottZ]

This was generated by AI during triage.

Summary

Two PRs address the same diagnostic gap: #48087 introduces narrow handling for Discord error code 30032, while #75696 rebuilds that handling on current main and corrects the warning to distinguish HTTP 400 from Discord error code 30032.

Related pull requests

• #48087 related — (+126/-0) — close as duplicate of #75696: #48087 adds a cap-error discriminator, actionable warning, and focused tests, but its warning incorrectly labels 30032 as an HTTP status. Despite the keep_open review on #48087, #75696 carries the same core fix on current main, addresses that review finding, and adds broader regression coverage.
• #75696 duplicate — (+185/-0) — keep open with a salvage path: retain the narrow 30032 discriminator, the explicit “HTTP 400 / error code 30032” diagnostic, and tests proving unrelated failures still reach the generic handler. This agrees with the maintainer-bot keep_open verdict; unlike #48087, the diff also handles JSON data payloads and follows the current exception shape and surrounding implementation.

Duplicates

#48087 and #75696 implement substantially the same dedicated Discord command-cap diagnostic; #75696 is the current-main replacement that incorporates the review correction, so #48087 can be closed as a duplicate of #75696.

Suggested consolidation

Keep #75696 open with the salvage path identified by the maintainer-bot review: preserve its narrow error-code detection, actionable warning, and positive and negative regression tests. Close #48087 as a duplicate of #75696 because the latter reimplements the same fix on current main while correcting #48087’s HTTP-status wording and expanding compatibility and false-positive coverage.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    subgraph Dup48087 ["PRs duplicating each other"]
        P48087["PR #48087 (open)"]
        P75696["PR #75696 (open)"]
    end
    class P48087 open
    class P75696 open
    class P48087 target
    click P48087 "https://github.com/NousResearch/hermes-agent/pull/48087"
    click P75696 "https://github.com/NousResearch/hermes-agent/pull/75696"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 2 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 18 kB of PR diffs, 6 kB of issue/PR text, 3 kB of discussion (3 comments), 0 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.