mward4/deepseek-harness
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 32 Wiki Activity

refactor(web): use empty retry user messages without loop changes

Browse Source
This commit is contained in:
Yichen Jiang committed 2026-11-02 15:58:09 +08:00
1 parent 22db5af493
commit 818e8eced2
65 files changed
+412 -736
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

scripts/smoke-python-runtime.py
+4 -18
View File
@@ -113,6 +113,7 @@ RESTART_SECOND_SESSION_ID = "process-two"
RESUME_PROMPT = "Continue the first Python SDK session after restarting the runtime."
RESUME_TEXT = "RESUMED_PROCESS_OK"
RECOVERY_PROMPT = "Run three todo updates; the scheduler will reject the second call."
RECOVERY_CONTINUE_PROMPT = "Continue after the scheduler failure without repeating any tools."
RECOVERY_TEXT = "SCHEDULER_RECOVERY_OK"
RECOVERY_SESSION_ID = "scheduler-recovery"
RECOVERY_CALL_IDS = ("scheduler-complete", "scheduler-fail", "scheduler-unstarted")
@@ -347,7 +348,7 @@ def completion_chunks(body: dict[str, object]) -> list[dict[str, object]]:
block.get("text")
for block in latest.get("content", [])
if isinstance(block, dict) and block.get("type") == "text"
and block.get("text") == RECOVERY_PROMPT
and block.get("text") in {RECOVERY_PROMPT, RECOVERY_CONTINUE_PROMPT}
), None)
if recovery_prompt == RECOVERY_PROMPT:
assert_advertised_tool(body, "todo_write")
@@ -357,9 +358,7 @@ def completion_chunks(body: dict[str, object]) -> list[dict[str, object]]:
chunks.extend({**chunk, "index": index} for chunk in call[1:-2])
chunks.extend(call[-2:])
return chunks
if any(block.get("tool_use_id") == RECOVERY_CALL_IDS[-1]
for message in messages for block in message.get("content", [])
if isinstance(block, dict) and block.get("type") == "tool_result"):
if recovery_prompt == RECOVERY_CONTINUE_PROMPT:
results = [
block for message in messages
for block in message.get("content", [])
@@ -1910,7 +1909,7 @@ def smoke_sdk_scheduler_recovery(base_url: str, executable: Path, update_snapsho
with connect() as harness:
failed = harness.run(RECOVERY_PROMPT, session_id=RECOVERY_SESSION_ID)
continued = harness.run([], session_id=RECOVERY_SESSION_ID)
continued = harness.run(RECOVERY_CONTINUE_PROMPT, session_id=RECOVERY_SESSION_ID)
results = [failed, continued]
if [result.finish_reason for result in results] != ["error", "completed"]:
outcomes = [
@@ -1925,18 +1924,6 @@ def smoke_sdk_scheduler_recovery(base_url: str, executable: Path, update_snapsho
raise AssertionError(f"scheduler recovery expected two serialized model requests: {requests}")
logs = read_session_logs(sessions)
records = logs[RECOVERY_SESSION_ID]
prompts = [record for record in records if record.get("type") == "user/message"
and record["data"]["source"].get("kind") == "user"]
retries = [record for record in records if record.get("type") == "agent/retry"]
if len(prompts) != 1 or len(retries) != 1:
raise AssertionError("scheduler recovery must record one original prompt and one retry control")
retry_id = retries[0]["data"]["inputId"]
queued_retry_ids = [message.get("id") for record in records
if record.get("type") == "agent/inbox/spliced"
for message in record["data"]["inserted"]
if message.get("source", {}).get("kind") == "retry"]
if not isinstance(retry_id, str) or queued_retry_ids != [retry_id]:
raise AssertionError("scheduler recovery lost the retry control's durable inbox identity")
calls = [record["data"]["callId"] for record in records if record.get("type") == "tool/call"]
if calls != list(RECOVERY_CALL_IDS[:2]):
raise AssertionError(f"scheduler recovery dispatched the unstarted call: {calls}")
@@ -1953,7 +1940,6 @@ def smoke_sdk_scheduler_recovery(base_url: str, executable: Path, update_snapsho
replacements = [
*snapshot_directory_replacements(root, "{{cwd}}"),
(RECOVERY_SESSION_ID, "{{parent}}"),
(retry_id, "{{messageId}}"),
]
result_value = [{
"session_id": result.session_id,