mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-07-24 16:54:43 +00:00
eebc2286fc
Gateway half of the #49874 salvage: pass compression_deferred through both _run_agent_inner result dicts and guard the compression-exhausted auto-reset block with it — a lock-contended defer keeps the session intact (the concurrent compressor is actively shrinking it) instead of wiping it via reset_session. Regression tests: - tests/run_agent/test_compression_lock_defer.py — provider-mock 413 and 400-overflow turns whose compression pass lost the lock end as compression_deferred (failed=False, no compression_exhausted); flag unset keeps the terminal exhaustion path byte-identical; type-pin tests vs MagicMock agents and junk flag values; cap=1 e2e proving the refunded pre-API defer leaves the budget for the provider-proven 413 retry. - tests/agent/test_preflight_lock_defer.py — a lock-skipped preflight pass stops the loop WITHOUT arming preflight_compression_blocked; plain no-op still arms it; MagicMock junk does not defer. - tests/gateway/test_compression_deferred_soft_result.py — AST pin that the deferred branch guards the auto-reset chain and performs no session mutation (mirrors test_35809_auto_reset_clean_context.py).
98 lines
3.9 KiB
Python
98 lines
3.9 KiB
Python
"""Gateway must treat ``compression_deferred`` as a soft result (#49874).
|
|
|
|
A lock-contended compression defer means a CONCURRENT compressor is actively
|
|
shrinking the session — the opposite of ``compression_exhausted`` (session
|
|
permanently too large). The gateway's auto-reset (#9893/#35809) must never
|
|
fire for a deferred turn: the session stays intact and the next message
|
|
retries normally.
|
|
|
|
AST invariants on ``gateway/run.py`` (mirrors
|
|
``test_35809_auto_reset_clean_context.py``'s load-bearing pin style):
|
|
|
|
* the ``compression_deferred`` branch guards the auto-reset block — a
|
|
deferred result can never reach ``reset_session``;
|
|
* the deferred branch itself performs NO session mutation (no
|
|
``reset_session``, no ``_evict_cached_agent``, no
|
|
``_clear_conversation_scope``).
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import ast
|
|
import inspect
|
|
|
|
from gateway import run as gateway_run
|
|
|
|
|
|
def _calls(node: ast.AST) -> set[str]:
|
|
return {
|
|
n.func.attr
|
|
for n in ast.walk(node)
|
|
if isinstance(n, ast.Call) and isinstance(n.func, ast.Attribute)
|
|
}
|
|
|
|
|
|
def _find_deferred_guarded_reset_chain() -> ast.If:
|
|
"""Return the ``if agent_result.get('compression_deferred') ... elif
|
|
agent_result.get('compression_exhausted') ... reset_session`` chain."""
|
|
tree = ast.parse(inspect.getsource(gateway_run))
|
|
|
|
for node in ast.walk(tree):
|
|
if not isinstance(node, ast.If):
|
|
continue
|
|
test_consts = [
|
|
n.value
|
|
for n in ast.walk(node.test)
|
|
if isinstance(n, ast.Constant) and isinstance(n.value, str)
|
|
]
|
|
if "compression_deferred" not in test_consts:
|
|
continue
|
|
# The reset must live in the orelse (elif compression_exhausted ...),
|
|
# never in the deferred body.
|
|
orelse_calls = set()
|
|
for sub in node.orelse:
|
|
orelse_calls |= _calls(sub)
|
|
if "reset_session" in orelse_calls:
|
|
return node
|
|
raise AssertionError(
|
|
"Could not locate the compression_deferred guard in front of the "
|
|
"compression-exhausted auto-reset block in gateway/run.py. The "
|
|
"soft-defer contract (#49874: lock-contended defer must never "
|
|
"auto-reset the session) is no longer structurally guaranteed."
|
|
)
|
|
|
|
|
|
class TestCompressionDeferredIsSoft:
|
|
def test_deferred_branch_guards_the_auto_reset(self):
|
|
"""The auto-reset (``reset_session``) must be unreachable when
|
|
``compression_deferred`` is set: the deferred check comes FIRST and
|
|
the reset lives only in its elif chain."""
|
|
node = _find_deferred_guarded_reset_chain()
|
|
# The exhaustion reset is in the orelse — verified by the finder.
|
|
# The deferred body must not mutate the session in any way.
|
|
body_calls = set()
|
|
for sub in node.body:
|
|
body_calls |= _calls(sub)
|
|
forbidden = {
|
|
"reset_session",
|
|
"_evict_cached_agent",
|
|
"_clear_conversation_scope",
|
|
}
|
|
assert not (body_calls & forbidden), (
|
|
f"The compression_deferred branch in gateway/run.py performs "
|
|
f"session mutation ({body_calls & forbidden}). A lock-contended "
|
|
f"defer is transient — the session must stay intact so the next "
|
|
f"message retries against the freshly compressed context "
|
|
f"(#49874, #69870)."
|
|
)
|
|
|
|
def test_deferred_result_key_is_passed_through_run_agent_inner(self):
|
|
"""``_run_agent_inner``'s result dicts must carry the
|
|
``compression_deferred`` key so the persistence block can see it —
|
|
the exact gap that made the exhaustion misclassification possible
|
|
(the flag existed but nothing consumed it)."""
|
|
src = inspect.getsource(gateway_run)
|
|
assert src.count('"compression_deferred"') >= 3, (
|
|
"gateway/run.py must read AND pass through compression_deferred "
|
|
"(persistence-block guard + both _run_agent_inner result dicts)."
|
|
)
|