chore: import upstream snapshot with attribution

2026-07-13 13:12:33 +08:00
commit aacb60a4af
3387 changed files with 981307 additions and 0 deletions
@@ -0,0 +1,165 @@
+from __future__ import annotations
+
+import csv
+import importlib.util
+import json
+import sys
+from pathlib import Path
+
+
+def _load_script():
+    path = Path(__file__).parents[2] / "scripts" / "experiments" / (
+        "analyze_dashscope_payload_parity.py"
+    )
+    spec = importlib.util.spec_from_file_location("analyze_dashscope_payload_parity", path)
+    assert spec and spec.loader
+    module = importlib.util.module_from_spec(spec)
+    sys.modules[spec.name] = module
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_qwen_payload_parity_reports_hard_failures(tmp_path: Path) -> None:
+    mod = _load_script()
+    trace = tmp_path / "artifacts" / "case-1" / "llm_calls.jsonl"
+    trace.parent.mkdir(parents=True)
+    payload = {
+        "model": "qwen3.6-flash",
+        "extra_body": {"enable_thinking": True, "preserve_thinking": True},
+        "max_tokens": 2048,
+        "stream": True,
+        "tool_choice": {
+            "type": "function",
+            "function": {"name": "read_source"},
+        },
+        "stream_options": {"include_usage": False},
+        "messages": [
+            {
+                "role": "assistant",
+                "reasoning_content": "historical reasoning should not replay",
+                "tool_calls": [
+                    {
+                        "id": "call-1",
+                        "type": "function",
+                        "function": {"name": "read_source", "arguments": "{}"},
+                    }
+                ],
+            }
+        ],
+        "tools": [
+            {
+                "type": "function",
+                "function": {
+                    "name": "read_source",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {"path": False},
+                    },
+                },
+            }
+        ],
+    }
+    trace.write_text(
+        json.dumps({"event": "llm.request", "call_index": 1, "payload": payload}) + "\n",
+        encoding="utf-8",
+    )
+
+    summary, rows = mod.analyze_paths([tmp_path / "artifacts"])
+    json_path = tmp_path / "summary.json"
+    csv_path = tmp_path / "rows.csv"
+    mod.write_outputs(summary, rows, json_path=json_path, csv_path=csv_path)
+
+    assert summary["checked_payloads"] == 1
+    assert summary["failed_checks"] >= 6
+    failed = {row["check"] for row in rows if row["status"] == "fail"}
+    assert {
+        "dashscope_max_completion_tokens",
+        "qwen_flash_no_reasoning_replay",
+        "qwen_flash_no_preserve_thinking",
+        "dashscope_thinking_no_forced_tool_choice",
+        "stream_include_usage",
+        "tool_schema_no_boolean_values",
+        "tool_call_pairing",
+    } <= failed
+    csv_rows = list(csv.DictReader(csv_path.open()))
+    assert csv_rows[0]["instance_id"] == "case-1"
+    assert json.loads(json_path.read_text())["failed_checks_by_name"]["stream_include_usage"] == 1
+
+
+def test_qwen_payload_parity_accepts_live_dashscope_request_shape(tmp_path: Path) -> None:
+    mod = _load_script()
+    trace = tmp_path / "artifacts" / "case-2" / "llm_calls.jsonl"
+    trace.parent.mkdir(parents=True)
+    payload = {
+        "model": "qwen3.6-flash",
+        "enable_thinking": True,
+        "max_completion_tokens": 32768,
+        "stream_options": {"include_usage": True},
+        "messages": [
+            {"role": "system", "content": "tools available"},
+            {
+                "role": "assistant",
+                "tool_calls": [
+                    {
+                        "id": "call-1",
+                        "type": "function",
+                        "function": {"name": "read_source", "arguments": "{}"},
+                    }
+                ],
+            },
+            {"role": "tool", "tool_call_id": "call-1", "content": "ok"},
+        ],
+        "tools": [
+            {
+                "type": "function",
+                "function": {
+                    "name": "read_source",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {"path": {"type": "string"}},
+                        "required": ["path"],
+                        "additionalProperties": False,
+                    },
+                },
+            }
+        ],
+    }
+    trace.write_text(
+        json.dumps({"event": "llm.request", "call_index": 1, "payload": payload}) + "\n",
+        encoding="utf-8",
+    )
+
+    summary, rows = mod.analyze_paths([tmp_path / "artifacts"])
+
+    assert summary["checked_payloads"] == 1
+    assert summary["failed_checks"] == 0
+    checks = {row["check"]: row for row in rows}
+    assert checks["dashscope_enable_thinking"]["status"] == "pass"
+    assert checks["tool_schema_no_boolean_values"]["status"] == "pass"
+
+
+def test_qwen_payload_parity_skips_stream_usage_check_for_non_stream_fallback(
+    tmp_path: Path,
+) -> None:
+    mod = _load_script()
+    trace = tmp_path / "artifacts" / "case-3" / "llm_calls.jsonl"
+    trace.parent.mkdir(parents=True)
+    payload = {
+        "model": "qwen3.6-flash",
+        "enable_thinking": True,
+        "max_completion_tokens": 32768,
+        "stream": False,
+        "messages": [{"role": "user", "content": "fix this"}],
+        "tools": [],
+    }
+    trace.write_text(
+        json.dumps({"event": "llm.request", "call_index": 1, "payload": payload}) + "\n",
+        encoding="utf-8",
+    )
+
+    summary, rows = mod.analyze_paths([tmp_path / "artifacts"])
+
+    assert summary["checked_payloads"] == 1
+    assert summary["failed_checks"] == 0
+    checks = {row["check"]: row for row in rows}
+    assert checks["stream_include_usage"]["status"] == "skip"