From 048460a3f056c99ea6e3f7713c099cd76510ad77 Mon Sep 17 00:00:00 2001
From: danielhanchen <unslothai@gmail.com>
Date: Wed, 22 Jul 2026 08:13:30 +0000
Subject: [PATCH] Studio: do not bind a research run to a populated assistant
 reply

create_run adopted any assistant message under the user turn whose
researchRunId was unset, including a prior answer reused by a retry. On
completion _update_assistant drops the untagged text and source parts, so
that answer was silently overwritten. Only bind to an empty placeholder or
this run's own message, and reject a reply that already carries content.
---
 studio/backend/storage/research_runs_db.py    | 14 +++++++
 .../tests/test_research_runs_storage.py       | 37 +++++++++++++++++++
 2 files changed, 51 insertions(+)

diff --git a/studio/backend/storage/research_runs_db.py b/studio/backend/storage/research_runs_db.py
index 564510da7e..72d8971bdd 100644
--- a/studio/backend/storage/research_runs_db.py
+++ b/studio/backend/storage/research_runs_db.py
@@ -192,11 +192,25 @@ def create_run(
                     if isinstance(existing_metadata, dict)
                     else None
                 )
+                # Only bind to an empty placeholder or this run's own message. An
+                # untagged reply carries text/source parts that _update_assistant
+                # drops on completion, so binding one silently overwrites an
+                # existing answer (for example a retry reusing a prior answer id).
+                existing_answer = any(
+                    isinstance(part, dict)
+                    and (
+                        (part.get("type") == "text" and (part.get("text") or "").strip())
+                        or part.get("type") == "source"
+                    )
+                    and part.get("researchRunId") is None
+                    for part in _loads(message["content_json"], [])
+                )
                 if (
                     message["thread_id"] != thread_id
                     or message["role"] != "assistant"
                     or message["parent_id"] != user_message_id
                     or existing_run_id not in (None, run_id)
+                    or (existing_run_id is None and existing_answer)
                 ):
                     raise ResearchConflictError(
                         "Assistant message does not match this research run"
diff --git a/studio/backend/tests/test_research_runs_storage.py b/studio/backend/tests/test_research_runs_storage.py
index c336ad8629..0e9bf3b91b 100644
--- a/studio/backend/tests/test_research_runs_storage.py
+++ b/studio/backend/tests/test_research_runs_storage.py
@@ -2208,6 +2208,43 @@ def test_create_run_conflict_rolls_back_placeholder_and_run(research_home):
     assert studio_db.get_chat_message("thread-1", "conflict")["parentId"] is None
 
 
+def test_create_run_rejects_binding_to_populated_reply(research_home):
+    # A prior answer under the same user turn (untagged, no researchRunId) must
+    # not be adopted as the placeholder: _update_assistant would drop its
+    # text/source parts on completion and silently overwrite that answer.
+    studio_db.upsert_chat_message(
+        {
+            "id": "prior-answer",
+            "threadId": "thread-1",
+            "parentId": "user-1",
+            "role": "assistant",
+            "content": [
+                {"type": "text", "text": "existing answer"},
+                {"type": "source", "sourceType": "url", "url": "https://kept.example"},
+            ],
+            "createdAt": 4,
+        }
+    )
+    with pytest.raises(research_db.ResearchConflictError):
+        _create(assistant_message_id = "prior-answer")
+    assert research_db.get_run("run-1") is None
+    preserved = studio_db.get_chat_message("thread-1", "prior-answer")
+    assert preserved["content"][0]["text"] == "existing answer"
+    # An empty placeholder under the same turn is still accepted.
+    studio_db.upsert_chat_message(
+        {
+            "id": "empty-placeholder",
+            "threadId": "thread-1",
+            "parentId": "user-1",
+            "role": "assistant",
+            "content": [],
+            "createdAt": 5,
+        }
+    )
+    run = _create(assistant_message_id = "empty-placeholder")
+    assert run["assistantMessageId"] == "empty-placeholder"
+
+
 def test_update_assistant_replaces_report_parts_without_duplication(research_home):
     from core.research_runs import _update_assistant