Accept JSON-string parallel tool calls

Restore parallel normalization tolerance for provider/model outputs that stringify the tool_calls array inside tool_args.

The wrapper still expects normal tool-call objects after decoding, so nested parallel and document_query guards continue to run through the same validation path.

helpers/parallel_tools.py

@@ -78,6 +78,13 @@ def extract_tool_calls(args: dict[str, Any]) -> Any:
 
 
 def normalize_parallel_tool_calls(raw_calls: Any) -> list[NormalizedToolCall]:
+    if isinstance(raw_calls, str):
+        try:
+            raw_calls = json.loads(raw_calls)
+        except json.JSONDecodeError as exc:
+            raise ValueError(
+                "`tool_calls` must be an array of normal tool-call objects."
+            ) from exc
     if not isinstance(raw_calls, list):
         raise ValueError("`tool_calls` must be an array of normal tool-call objects.")
     if not raw_calls:

helpers/parallel_tools.py.dox.md

@@ -24,6 +24,7 @@
 - Helper modules own reusable framework APIs and must preserve public callers unless all callers, tests, and docs are updated together.
 - Wrapped tool-call items must use the same shape as normal tool calls: a tool name plus arguments.
 - Normalization accepts full agent-reply-shaped objects when `tool_name` and `tool_args` are present; non-contract planning fields such as `thoughts` or `headline` are ignored.
+- `tool_calls` should be an array, but normalization also accepts a valid JSON string encoding of that array to recover provider/model stringification.
 - Normalization rejects `document_query` inside `parallel` because document parsing and Q&A fan out into heavier worker/model paths that must run sequentially.
 - `call_subordinate` jobs run in isolated child chat contexts tagged with parent-chat metadata; they must not be added to the scheduler task list and may use normal child-chat tools, including `parallel`.
 - Direct tool jobs run in isolated background contexts and are blocked from recursively invoking `parallel`.

tests/test_parallel_tool.py

@@ -103,6 +103,41 @@ def test_normalize_parallel_tool_calls_accepts_normal_tool_request_shapes() -> N
     assert calls[1].tool_args == {"method": "list_tasks", "action": "list_tasks"}
 
 
+def test_normalize_parallel_tool_calls_accepts_json_string_array() -> None:
+    calls = parallel_tools.normalize_parallel_tool_calls(
+        json.dumps(
+            [
+                {
+                    "tool_name": "call_subordinate",
+                    "tool_args": {
+                        "profile": "researcher",
+                        "reset": True,
+                        "message": "Research nuclear fusion news in French.",
+                    },
+                    "headline": "Researching nuclear fusion news in French",
+                },
+                {
+                    "tool_name": "call_subordinate",
+                    "tool_args": {
+                        "profile": "researcher",
+                        "reset": True,
+                        "message": "Research nuclear fusion news in Italian.",
+                    },
+                    "headline": "Researching nuclear fusion news in Italian",
+                },
+            ]
+        )
+    )
+
+    assert [call.tool_name for call in calls] == [
+        "call_subordinate",
+        "call_subordinate",
+    ]
+    assert calls[0].tool_args["profile"] == "researcher"
+    assert calls[0].tool_args["reset"] is True
+    assert calls[1].tool_args["message"] == "Research nuclear fusion news in Italian."
+
+
 def test_normalize_parallel_tool_calls_rejects_nested_parallel() -> None:
     with pytest.raises(ValueError, match="cannot be nested"):
         parallel_tools.normalize_parallel_tool_calls(

tools/parallel.py.dox.md

@@ -32,7 +32,7 @@
 
 ## Key Concepts
 
-- `tool_calls`, `calls`, and `items` are accepted aliases for the wrapped call list.
+- `tool_calls`, `calls`, and `items` are accepted aliases for the wrapped call list; a valid JSON string encoding of the list is tolerated for provider/model recovery.
 - Wrapped call items can use the same `tool_name`/`tool_args` shape as top-level agent replies; extra planning fields are ignored by normalization.
 - `job_ids` can be supplied as a string or list when awaiting, collecting, or canceling existing jobs.
 - The response is compact JSON intended for the model to read and continue with.