Fix split-field DSML promote and silent turn end after tools.

Parse DSML from combined content+reasoning when markers span fields; show a clear mismatch message when parse fails; run finalize synthesis when tools executed but the assistant body is empty.

Co-authored-by: Cursor <cursoragent@cursor.com>

runtime/direct_loop.py

@@ -27,7 +27,7 @@ from runtime.hooks_runtime import trigger_hook_event
 from runtime.dsml_tool_parse import (
     contains_dsml_tool_markers,
     dsml_text_tools_enabled,
-    try_parse_dsml_tool_calls_from_fields,
+    promote_dsml_tool_calls_in_response,
 )
 from runtime.tools.experts.network_ops.netx_tools import ops_netx_system_context_extension
 
@@ -843,6 +843,18 @@ def _prepare_llm_tools(
     return llm_tools
 
 
+def _dsml_mismatch_user_message(*, lang: str) -> str:
+    if str(lang or "").strip().lower().startswith("zh"):
+        return (
+            "模型返回了 DSML 工具标记，但未能解析为可执行的工具调用，本轮未实际执行工具。"
+            "请重试，或检查模型/网关是否应输出原生 tool_calls。"
+        )
+    return (
+        "The model returned DSML tool markup, but it could not be parsed into executable tool calls; "
+        "no tools were run this turn. Please retry or verify the model/gateway emits native tool_calls."
+    )
+
+
 def _promote_dsml_tool_calls(
     *,
     allow: bool,
@@ -853,12 +865,13 @@ def _promote_dsml_tool_calls(
     """Runtime fallback: promote DSML in content/reasoning to native tool calls."""
     if not allow or llm_tool_calls:
         return assistant_text, reasoning_text, llm_tool_calls
-    parsed, clean_content, clean_reasoning = try_parse_dsml_tool_calls_from_fields(
-        content=assistant_text,
-        reasoning_content=reasoning_text,
+    clean_content, clean_reasoning, promoted = promote_dsml_tool_calls_in_response(
+        assistant_text,
+        reasoning_text,
+        [],
     )
-    if parsed:
-        return clean_content, clean_reasoning, parsed
+    if promoted:
+        return clean_content, clean_reasoning, promoted
     return assistant_text, reasoning_text, llm_tool_calls
 
 
@@ -902,8 +915,8 @@ def _chat_with_empty_body_retry(
             _extract_textual_tool_intent_names(f"{content}\n{reasoning}")
         )
         if allow_dsml_text_tools:
-            parsed, _, _ = try_parse_dsml_tool_calls_from_fields(content=content, reasoning_content=reasoning)
-            if parsed:
+            _, _, promoted_probe = promote_dsml_tool_calls_in_response(content, reasoning, [])
+            if promoted_probe:
                 textual_tool_intent = False
         if (content.strip() or tool_calls) and not textual_tool_intent:
             return resp
@@ -1502,6 +1515,7 @@ def run_oclaw_direct_loop(
                 assistant_text=assistant_text,
                 invoke_names=textual_tool_intent_names,
             )
+            final_text = _dsml_mismatch_user_message(lang=lang)
         else:
             step = _persist_assistant_step(
                 store=store,
@@ -1511,7 +1525,7 @@ def run_oclaw_direct_loop(
                 reasoning_text=reasoning_text,
                 llm_tool_calls=llm_tool_calls,
             )
-        final_text = step.assistant_text
+            final_text = step.assistant_text
         if not step.llm_tool_calls:
             break
         if round_idx == (max_rounds - 1):
@@ -1574,7 +1588,8 @@ def run_oclaw_direct_loop(
         if on_progress:
             on_progress(f"oclaw: tools done ({elapsed_ms}ms)")
 
-    if hit_tool_round_limit:
+    need_finalize = hit_tool_round_limit or (bool(tool_traces) and not str(final_text or "").strip())
+    if need_finalize:
         _check_stop(should_stop)
         if on_progress:
             on_progress("oclaw: finalize…")

runtime/dsml_tool_parse.py

@@ -307,6 +307,44 @@ def try_parse_dsml_tool_calls_from_fields(
     return None, str(content or ""), str(reasoning_content or "")
 
 
+def try_promote_dsml_from_fields(
+    *,
+    content: str = "",
+    reasoning_content: str = "",
+) -> tuple[list[LLMToolCall] | None, str, str]:
+    """Parse DSML per-field, then fall back to combined ``content+reasoning`` text."""
+    parsed, clean_content, clean_reasoning = try_parse_dsml_tool_calls_from_fields(
+        content=content,
+        reasoning_content=reasoning_content,
+    )
+    if parsed:
+        return parsed, clean_content, clean_reasoning
+
+    combined = "\n".join(p for p in (str(content or ""), str(reasoning_content or "")) if str(p or "").strip()).strip()
+    if not combined:
+        return None, str(content or ""), str(reasoning_content or "")
+
+    parsed2 = try_parse_deepseek_v4_dsml_tool_calls(combined)
+    if not parsed2:
+        return None, str(content or ""), str(reasoning_content or "")
+
+    clean_c = str(content or "")
+    clean_r = str(reasoning_content or "")
+    if contains_dsml_tool_markers(clean_c):
+        stripped_c = strip_first_dsml_tool_calls_block(clean_c)
+        if stripped_c is not None:
+            clean_c = stripped_c
+    if contains_dsml_tool_markers(clean_r):
+        stripped_r = strip_first_dsml_tool_calls_block(clean_r)
+        if stripped_r is not None:
+            clean_r = stripped_r
+    if contains_dsml_tool_markers(clean_c) or contains_dsml_tool_markers(clean_r):
+        stripped = strip_first_dsml_tool_calls_block(combined)
+        prefix = stripped if stripped is not None else ""
+        return parsed2, prefix, ""
+    return parsed2, clean_c, clean_r
+
+
 def dsml_text_tools_enabled(*, base_url: str = "", model_id: str = "") -> bool:
     """Whether DSML-in-text should be promoted to native tool calls."""
     env = str(os.getenv("AIA_DSML_TEXT_TOOLS") or "").strip().lower()
@@ -355,7 +393,7 @@ def promote_dsml_tool_calls_in_response(
 ) -> tuple[str, str, list[LLMToolCall]]:
     if tool_calls:
         return content, reasoning, tool_calls
-    parsed, clean_content, clean_reasoning = try_parse_dsml_tool_calls_from_fields(
+    parsed, clean_content, clean_reasoning = try_promote_dsml_from_fields(
         content=content,
         reasoning_content=reasoning,
     )
@@ -475,4 +513,5 @@ __all__ = [
     "strip_first_dsml_tool_calls_block",
     "try_parse_deepseek_v4_dsml_tool_calls",
     "try_parse_dsml_tool_calls_from_fields",
+    "try_promote_dsml_from_fields",
 ]

tests/test_direct_loop_empty_assistant_response.py

@@ -161,7 +161,7 @@ def test_direct_loop_dsml_text_persisted_as_failed_tool_pair(tmp_path, monkeypat
         persist_user_message=True,
         max_tool_rounds=1,
     )
-    assert out.final_text == ""
+    assert "DSML" in out.final_text
     rows = store.get_messages(session_id=sess.id, limit=20)
     tool_rows = [r for r in rows if getattr(r, "role", "") == "tool"]
     assert tool_rows
@@ -212,7 +212,7 @@ def test_direct_loop_mixed_text_with_tool_intent_is_blocked(tmp_path, monkeypatc
         persist_user_message=True,
         max_tool_rounds=1,
     )
-    assert out.final_text == ""
+    assert "DSML" in out.final_text
     rows = store.get_messages(session_id=sess.id, limit=20)
     tool_rows = [r for r in rows if getattr(r, "role", "") == "tool"]
     assert tool_rows

tests/test_dsml_tool_parse.py

@@ -67,6 +67,24 @@ def test_malformed_returns_none() -> None:
     assert try_parse_deepseek_v4_dsml_tool_calls("<||DSML||tool_calls>broken") is None
 
 
+def test_promote_combined_when_dsml_only_in_reasoning() -> None:
+    from runtime.dsml_tool_parse import try_promote_dsml_from_fields
+
+    p = "\uFF5C"
+    reasoning = (
+        f"<{p}{p}DSML{p}{p}tool_calls>\n"
+        f"<{p}{p}DSML{p}{p}invoke name=\"run_command\">\n"
+        f"<{p}{p}DSML{p}{p}parameter name=\"command\" string=\"true\">echo</{p}{p}DSML{p}{p}parameter>\n"
+        f"</{p}{p}DSML{p}{p}invoke>\n"
+        f"</{p}{p}DSML{p}{p}tool_calls>"
+    )
+    parsed, clean_c, clean_r = try_promote_dsml_from_fields(content="让我检查一下。", reasoning_content=reasoning)
+    assert parsed is not None and len(parsed) == 1
+    assert parsed[0].name == "run_command"
+    assert "检查一下" in clean_c
+    assert "DSML" not in clean_r
+
+
 def test_parse_double_fullwidth_pipe_variant() -> None:
     p = "\uFF5C"
     text = (