mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-07-31 19:16:29 +00:00
6b81590c55
Systematic prune per AGENTS.md test policy, one pass over every major test tree (gateway, hermes_cli, tools, agent, run_agent, plugins, cli, cron, tui_gateway, honcho/openviking, root-level): - DELETE: source-reading tests (read_text/getsource on prod files), change-detector tests (exact catalog counts, model-name snapshots, config version literals), mock-echo tests (assert a mock returns what it was told), assertion-free/trivial tests, near-duplicate parametrizations (boundaries + one representative kept), async/sync twin duplicates, cosmetic within-file variations. - KEEP (mandatory): security/redaction/approval guards, message-role alternation invariants, prompt-caching/deterministic-call-id invariants, issue-number regression tests (deduped), E2E tests. - 6 test files deleted outright (script-style/no-assert or fully redundant); conftest.py, fakes/, fixtures/ untouched. - tests/acp/conftest.py added: autouse fixture stubs the live models.dev/GitHub/Copilot/Anthropic inventory fetches that ACP server tests performed on every session create — test_server.py 147s → 3.4s, and the tests are now genuinely hermetic. - Sleep-based slowness shrunk where safe (codex_ttfb_watchdog, compression_concurrent_fork, etc.); no wall-clock assertion tightened. Verification: full hermetic suite via scripts/run_tests.sh — 2439 files, 31,130 tests passed, 0 failed, 0 flaky retries, 315s wall (baseline: 583s wall, 13,564s subprocess CPU).
67 lines
2.3 KiB
Python
67 lines
2.3 KiB
Python
"""Regression tests for issue #27405.
|
|
|
|
The preflight compression gate must trigger when *either* the message
|
|
count exceeds the protected ranges OR the cheap char-based token
|
|
estimate already crosses the configured threshold. Pre-fix, only the
|
|
message-count condition was checked, so a session with a small number
|
|
of huge messages would silently skip compression and eventually hit a
|
|
hard context-overflow error.
|
|
"""
|
|
|
|
from agent.turn_context import _should_run_preflight_estimate
|
|
|
|
|
|
# Protected-range counts mirror the compressor defaults. THRESHOLD_TOKENS is an
|
|
# arbitrary test threshold passed explicitly into the helper — it is NOT the
|
|
# live runtime threshold (which is max(0.5*window, MINIMUM_CONTEXT_LENGTH) per
|
|
# model); the helper takes the threshold as a parameter so the tests are
|
|
# self-contained and independent of model metadata.
|
|
PROTECT_FIRST_N = 3
|
|
PROTECT_LAST_N = 20
|
|
THRESHOLD_TOKENS = 64_000
|
|
|
|
|
|
def _msg(content: str) -> dict:
|
|
return {"role": "user", "content": content}
|
|
|
|
|
|
def test_few_messages_huge_content_triggers_gate():
|
|
"""The bug from #27405: 8 messages with one massive content blob."""
|
|
# ~280K chars in one message ~= 70K tokens at 4 chars/token.
|
|
big = "x" * 280_000
|
|
messages = [_msg("hi")] * 7 + [_msg(big)]
|
|
assert len(messages) <= PROTECT_FIRST_N + PROTECT_LAST_N + 1 # would fail old gate
|
|
assert _should_run_preflight_estimate(
|
|
messages, PROTECT_FIRST_N, PROTECT_LAST_N, THRESHOLD_TOKENS
|
|
) is True
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def test_content_above_threshold_triggers():
|
|
"""A single message comfortably above the threshold trips branch (b)."""
|
|
# ~threshold*4 chars => ~threshold tokens; +1000 tokens of margin so the
|
|
# test doesn't depend on per-message dict-wrapping overhead in the
|
|
# shared estimator's (chars+3)//4 rounding.
|
|
messages = [_msg("x" * ((THRESHOLD_TOKENS + 1000) * 4))]
|
|
assert _should_run_preflight_estimate(
|
|
messages, PROTECT_FIRST_N, PROTECT_LAST_N, THRESHOLD_TOKENS
|
|
) is True
|
|
|
|
|
|
def test_content_below_threshold_does_not_trigger():
|
|
"""A single message comfortably below the threshold (and few messages)
|
|
must not trigger — the estimator stays under and the count gate is not
|
|
tripped."""
|
|
messages = [_msg("x" * ((THRESHOLD_TOKENS - 1000) * 4))]
|
|
assert _should_run_preflight_estimate(
|
|
messages, PROTECT_FIRST_N, PROTECT_LAST_N, THRESHOLD_TOKENS
|
|
) is False
|
|
|
|
|
|
|
|
|
|
|
|
|