mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-07-18 14:52:04 +00:00
fix(langfuse): bound _TRACE_STATE growth from non-finalizing turns
Scoping the trace key by turn_id (the prior commit) fixed cross-turn collisions but introduced a slow leak: _finish_trace only pops a key when a turn ends cleanly (final response has content and no tool calls), so any turn that is interrupted, ends on a tool call, or has empty final content now leaves its uniquely-keyed entry in _TRACE_STATE forever. Previously the constant per-session key was overwritten by the next turn, capping growth at ~1 entry per session. Add an LRU cap (_MAX_TRACE_STATE) enforced by _evict_stale_locked, called under _STATE_LOCK immediately before each insert. It evicts the least-recently-updated entries (using the previously-dead last_updated_at field) and ends their root span so nothing dangles. Regression test drives 50 non-finalizing turns against a cap of 8 and asserts the dict stays bounded with the most-recent turns surviving.
This commit is contained in:
parent
e1d10ec1ed
commit
f4fbaa6cda
2 changed files with 56 additions and 0 deletions
|
|
@ -373,6 +373,27 @@ class TestTurnTraceIsolation:
|
|||
|
||||
assert k_pre_api == k_post_api == k_post_turn
|
||||
|
||||
def test_non_finalizing_turns_do_not_grow_state_unboundedly(self, monkeypatch):
|
||||
"""Per-turn keys mean a turn that never finalizes leaves a lingering
|
||||
entry. Without a cap that grows once per non-finalizing turn forever;
|
||||
the LRU eviction must bound _TRACE_STATE at _MAX_TRACE_STATE.
|
||||
"""
|
||||
mod = self._fresh_plugin()
|
||||
started: list = []
|
||||
monkeypatch.setattr(mod, "_get_langfuse", lambda: self._fake_client(started))
|
||||
monkeypatch.setattr(mod, "_end_observation", lambda *a, **k: None)
|
||||
monkeypatch.setattr(mod, "_MAX_TRACE_STATE", 8)
|
||||
mod._TRACE_STATE.clear()
|
||||
|
||||
# Far more non-finalizing turns than the cap.
|
||||
for n in range(50):
|
||||
self._run_turn(mod, session="sess-leak", turn_n=n, finalize=False)
|
||||
|
||||
assert len(mod._TRACE_STATE) <= 8
|
||||
# The survivors are the most-recently-updated turns (LRU eviction).
|
||||
surviving = sorted(int(k.rsplit("turn", 1)[1]) for k in mod._TRACE_STATE)
|
||||
assert surviving == list(range(42, 50))
|
||||
|
||||
def test_trace_key_strings_unchanged_by_refactor(self):
|
||||
"""Pin the exact key strings across all task/session/turn/api
|
||||
combinations so the _scope_prefix extraction can never silently change
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue