gahow/agentic-kvc
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

replayer: restore optional max_inflight_sessions for backwards compat

Browse Source 
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Gahow Wang
2026-05-23 21:02:26 +08:00
parent a7df84bd3b
commit 2c7f7fdaae
1 changed files with 21 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
replayer/replay.py
View File

@@ -59,6 +59,7 @@ class ReplayConfig:
request_timeout_s: float = 600.0 request_timeout_s: float = 600.0
request_limit: int | None = None request_limit: int | None = None
model_name: str = "default" model_name: str = "default"
max_inflight_sessions: int | None = None # cap on concurrent sessions; None = unlimited
def _build_prompt_token_ids(req: TraceRequest) -> list[int]: def _build_prompt_token_ids(req: TraceRequest) -> list[int]:
@@ -215,21 +216,28 @@ async def _run_session(
earliest_ts: float, earliest_ts: float,
sweep_start: float, sweep_start: float,
sink: IncrementalMetricSink, sink: IncrementalMetricSink,
session_sem: asyncio.Semaphore | None = None,
) -> list[RequestMetrics]: ) -> list[RequestMetrics]:
for req in state.turns: if session_sem is not None:
# Wait until this request's trace timestamp await session_sem.acquire()
target_wall = (req.timestamp_s - earliest_ts) try:
elapsed = time.perf_counter() - sweep_start for req in state.turns:
if elapsed < target_wall: # Wait until this request's trace timestamp
await asyncio.sleep(target_wall - elapsed) target_wall = (req.timestamp_s - earliest_ts)
elapsed = time.perf_counter() - sweep_start
if elapsed < target_wall:
await asyncio.sleep(target_wall - elapsed)
token_ids = _build_prompt_token_ids(req) token_ids = _build_prompt_token_ids(req)
metric = await _dispatch_request( metric = await _dispatch_request(
client=client, config=config, req=req, client=client, config=config, req=req,
prompt_token_ids=token_ids, sem=request_sem, prompt_token_ids=token_ids, sem=request_sem,
) )
state.metrics.append(metric) state.metrics.append(metric)
await sink.append(metric) await sink.append(metric)
finally:
if session_sem is not None:
session_sem.release()
return state.metrics return state.metrics