fix(compression): notify gateway users when summary generation fails by iamagenius00 · Pull Request #16333 · NousResearch/hermes-agent

iamagenius00 · 2026-04-27T03:17:12Z

Problem

When auxiliary compression's summary LLM call fails (e.g. model 404, misconfigured auxiliary.compression.model), the compressor still drops the selected turns and inserts a static fallback placeholder. The dropped context is unrecoverable.

Previously the only signal of this failure was a WARNING in agent.log. Gateway users (Telegram / Discord / Slack / etc.) had no visible signal that context was lost. A user could come back to a session, find earlier work missing, and have no idea why.

Why the existing `_emit_warning` path didn't catch it

The gateway hygiene path uses a temporary _hyg_agent constructed with quiet_mode=True, _print_fn = no-op, and no status_callback. So _emit_warning (which depends on status_callback) silently no-ops in this code path. The CLI / normal-agent path was already covered.

Reproducible scenario: auxiliary.compression.model set to a model the main provider doesn't serve (e.g. a Google model name routed to Anthropic). Every hygiene-triggered compaction logs Failed to generate context summary: ... 404 ..., the placeholder gets inserted, and the user sees nothing in their chat.

Fix

agent/context_compressor.py — track per-compress state:
- _last_summary_fallback_used: bool
- _last_summary_dropped_count: int
Cleared at the top of compress() and on on_session_reset(). Set when summary generation fails and the static fallback is inserted.
gateway/run.py hygiene auto-compress — after _compress_context returns, inspect _hyg_agent.context_compressor. If _last_summary_fallback_used is True, send a visible ⚠️ warning to the user via the platform adapter, including the dropped count and the underlying error message.
gateway/run.py /compress manual command — append the same warning to the manual /compress reply so users running it explicitly also see the failure.

Acceptance

✅ Summary success → no user-visible warning (unchanged).
✅ Summary failure on gateway hygiene → user receives a TG/Discord message with dropped count, underlying error, and remediation hint (/reset or check auxiliary.compression).
✅ Summary failure on /compress → warning appended to the command reply.
✅ CLI status_callback / _emit_warning path is untouched (no behavioural regression).
✅ Two new tests in tests/agent/test_context_compressor.py verify tracking fields are set on failure and cleared on subsequent success.

Tests

tests/agent/test_context_compressor.py: 54 passed in 1.73s

New tests:

TestSummaryFailureTrackingForGatewayWarning::test_compress_records_fallback_and_dropped_count_on_summary_failure
TestSummaryFailureTrackingForGatewayWarning::test_compress_clears_fallback_flag_on_subsequent_success

Notes

The warning is only sent to the platform that triggered the hygiene compaction (source.platform/source.chat_id). It does not broadcast to other platforms or sessions. The gateway adapter .send call is wrapped in try/except so a warning-delivery failure can't break the compression result.

When auxiliary compression's summary LLM call fails (e.g. model 404, auxiliary model misconfigured), the compressor still drops the selected turns and inserts a static fallback placeholder — the dropped context is unrecoverable. Previously the only signal of this was a WARNING in agent.log. Gateway users (Telegram/Discord/etc.) had no way to know context was lost because the existing _emit_warning path requires a status_callback, and the gateway hygiene path uses a temporary _hyg_agent with quiet_mode=True and no callback wired up. Changes: - ContextCompressor: track _last_summary_fallback_used and _last_summary_dropped_count on each compress() call. Cleared at the start of compress() and on session reset. - gateway/run.py hygiene: after auto-compress, inspect the temp agent's compressor; if fallback was used, send a visible ⚠️ warning to the user via the platform adapter (TG/Discord/etc.) including dropped count and the underlying error. - gateway/run.py /compress: append the same warning to the manual compress reply so users running /compress see the failure too. Acceptance: - Summary success: no user-visible warning (unchanged). - Summary failure on gateway hygiene: user receives a TG/Discord message with dropped count + error + remediation hint. - Summary failure on /compress: warning appended to the command reply. - CLI status_callback / _emit_warning path is untouched. - Test coverage: two new tests verify the tracking fields are set on failure and cleared on subsequent success.

alt-glitch · 2026-04-27T03:20:02Z

Related to #10719 — this PR adds the user-visible notification that #10719 identified as missing when compression summary generation fails.

…ing delivery Address review feedback on PR NousResearch#16333: 1. The hygiene-path warning send was missing metadata=_hyg_meta. On Telegram topics / Slack threads / Discord threads the warning would land in the main channel instead of the originating thread. Now reuses the same _hyg_meta dict already computed for the hygiene compaction itself. 2. New gateway-level test test_session_hygiene_warns_user_when_summary_generation_fails verifies end-to-end: - When the compressor's _last_summary_fallback_used flag is True, the gateway invokes adapter.send() exactly once. - The warning message includes the dropped count and the underlying error string. - metadata={'thread_id': ...} is propagated so the warning lands in the originating topic/thread. Tests: 20 gateway hygiene + 54 context_compressor — all pass.

…ning The fallback placeholder said "N conversation turns were removed" while the gateway warning said "N historical message(s) were removed". Use "messages" in both so users don't wonder if the two counters refer to different things.

The per-call reset block at the top of compress() cleared _last_summary_dropped_count and _last_summary_fallback_used but not _last_summary_error. Functionally this didn't break the gateway warning path (callers gate on _last_summary_fallback_used first, and _last_summary_error is overwritten on the next failure), but it left the three tracking fields inconsistent — anyone reading _last_summary_error standalone after a successful compress would see a stale value from a previous failed compress. Reset all three together so the per-call contract is uniform.

PR NousResearch#16333 added a warning to the manual /compress reply when the auxiliary summariser fails and the static fallback placeholder is used, but only the gateway-hygiene path had a test (test_session_hygiene_warns_user_when_summary_generation_fails). The /compress branch in _handle_compress_command was uncovered. New test test_compress_command_appends_warning_when_summary_generation_fails mocks the compressor's _last_summary_fallback_used / _last_summary_dropped_count / _last_summary_error fields and verifies the /compress reply contains the ⚠️ marker, the underlying error string, the dropped message count, and the 'historical message(s) were removed' wording — i.e. the same contract the hygiene-path test enforces.

…ing delivery Address review feedback on PR #16333: 1. The hygiene-path warning send was missing metadata=_hyg_meta. On Telegram topics / Slack threads / Discord threads the warning would land in the main channel instead of the originating thread. Now reuses the same _hyg_meta dict already computed for the hygiene compaction itself. 2. New gateway-level test test_session_hygiene_warns_user_when_summary_generation_fails verifies end-to-end: - When the compressor's _last_summary_fallback_used flag is True, the gateway invokes adapter.send() exactly once. - The warning message includes the dropped count and the underlying error string. - metadata={'thread_id': ...} is propagated so the warning lands in the originating topic/thread. Tests: 20 gateway hygiene + 54 context_compressor — all pass.

PR #16333 added a warning to the manual /compress reply when the auxiliary summariser fails and the static fallback placeholder is used, but only the gateway-hygiene path had a test (test_session_hygiene_warns_user_when_summary_generation_fails). The /compress branch in _handle_compress_command was uncovered. New test test_compress_command_appends_warning_when_summary_generation_fails mocks the compressor's _last_summary_fallback_used / _last_summary_dropped_count / _last_summary_error fields and verifies the /compress reply contains the ⚠️ marker, the underlying error string, the dropped message count, and the 'historical message(s) were removed' wording — i.e. the same contract the hygiene-path test enforces.

teknium1 · 2026-04-28T02:18:32Z

Merged via PR #16771 — #16771

Your 5 commits were cherry-picked onto current main and rebase-merged, so your authorship is preserved in git log. Thanks for catching this — silent context loss on misconfigured auxiliary.compression.model is nasty, and the two-hour debug session writeup made the premise easy to verify.

xinbenlv · 2026-04-28T14:09:38Z

That's fast merging, @teknium1 , let's go let's go!