fix(ai): creation dossier+fichier en mode agent (BUG-050)

backend/agent/loop.py

@@ -109,6 +109,30 @@ def _assistant_tool_message(content: str | None, tool_calls: list[Any]) -> dict[
     }
 
 
+def _deferred_tool_message(call: Any) -> dict[str, Any]:
+    """Answer a tool call that was not reached because the run paused.
+
+    A single LLM response may carry several tool calls. When one of them is
+    mutating and pauses the run for confirmation, the assistant message already
+    lists *all* of them, so every ``tool_call_id`` must get a tool result before
+    the next LLM call (the OpenAI tool protocol rejects dangling ids). The calls
+    that were not reached get a synthetic ``deferred`` result; the model
+    re-issues them once the confirmed call has been applied (BUG-050).
+    """
+    return {
+        "role": "tool",
+        "tool_call_id": call.id,
+        "name": call.name,
+        "content": json.dumps({
+            "status": "deferred",
+            "reason": (
+                "Not executed: the run paused to confirm an earlier tool call. "
+                "Re-issue this call if it is still needed."
+            ),
+        }, ensure_ascii=False),
+    }
+
+
 def _execute_confirmed(
     ctx: ToolContext,
     confirm_pending: dict[str, Any],
@@ -248,7 +272,7 @@ async def run_agent(
         _emit_note(response.content or "")
         convo.append(_assistant_tool_message(response.content, response.tool_calls))
 
-        for call in response.tool_calls:
+        for index, call in enumerate(response.tool_calls):
             if quota is not None and len(executed) >= quota:
                 logger.warning(f"Agent reached the tool-call quota ({quota})")
                 return AgentResult(
@@ -268,6 +292,11 @@ async def run_agent(
                 pending = e.to_dict()
                 # Include the tool-call id so the client can echo it back.
                 pending["error"]["id"] = call.id
+                # BUG-050: the assistant message lists every tool call of this
+                # batch, so answer the ones we did not reach to keep the
+                # conversation valid for the resumed turn.
+                for skipped in response.tool_calls[index + 1:]:
+                    convo.append(_deferred_tool_message(skipped))
                 return AgentResult(
                     content=response.content or "",
                     messages=convo,