From ebcb154e3724f2b69c106524ed7db962825653ce Mon Sep 17 00:00:00 2001
From: jinliyl <6469360+jinliyl@users.noreply.github.com>
Date: Thu, 20 Aug 2026 16:15:30 +0800
Subject: [PATCH] fix(search): isolate range dedup state (#465)

---
 reme/steps/index/_dedup.py     | 17 +++++++--------
 tests/unit/test_search_step.py | 39 ++++++++++++++++++++++++++++++++++
 2 files changed, 47 insertions(+), 9 deletions(-)

diff --git a/reme/steps/index/_dedup.py b/reme/steps/index/_dedup.py
index e18d1c40..96929e7b 100644
--- a/reme/steps/index/_dedup.py
+++ b/reme/steps/index/_dedup.py
@@ -1,8 +1,8 @@
 """Shared tool_context-scoped chunk dedup with TTL.
 
-Used by ``search``/``vector_search``/``bm25_search`` to avoid returning the
+Used by ``search_v2``/``vector_search``/``bm25_search`` to avoid returning the
 same content twice within one agent tool_context. Per-context state lives at
-``app_context.metadata["tool_contexts"][tool_context_id]["search_seen_chunk_ids"]``
+``app_context.metadata["tool_contexts"][tool_context_id]["search_seen_chunk_ranges"]``
 as ``{path: [(start_line, end_line, timestamp), ...]}``; a chunk is skipped when
 its ``[start_line, end_line]`` is fully covered by the union of seen entries
 (merged overlapping/adjacent intervals) for the same ``path``. Entries older
@@ -30,7 +30,7 @@ class _ToolContextDedupMixin:
     """
 
     TOOL_CONTEXTS_KEY: Final[str] = "tool_contexts"
-    SEARCH_SEEN_KEY: Final[str] = "search_seen_chunk_ids"
+    SEARCH_SEEN_RANGES_KEY: Final[str] = "search_seen_chunk_ranges"
 
     def __init_subclass__(cls, **kwargs):
         super().__init_subclass__(**kwargs)
@@ -105,11 +105,10 @@ class _ToolContextDedupMixin:
         now = (clock or self._now_ts)()
         ttl = ttl_override if ttl_override is not None else float(self.seen_ttl_hours) * 60 * 60
         store = self._tool_context_store(tool_context_id)
-        seen = store.get(self.SEARCH_SEEN_KEY, {})
-        # Normalize legacy in-memory formats to {path: [(s, e, t), ...]}.
-        # Older shapes ({chunk_id: timestamp} or a plain list of ids) cannot
-        # be migrated because chunk_id is an opaque hash; seen is a transient
-        # per-Application cache, so dropping it is safe.
+        seen = store.get(self.SEARCH_SEEN_RANGES_KEY, {})
+        # Normalize unexpected in-memory formats to {path: [(s, e, t), ...]}.
+        # Other shapes cannot be migrated because chunk_id is an opaque hash;
+        # seen is a transient per-Application cache, so dropping it is safe.
         if not isinstance(seen, dict) or (seen and all(not isinstance(v, list) for v in seen.values())):
             seen = {}
 
@@ -117,7 +116,7 @@ class _ToolContextDedupMixin:
         # Expire stale tuples across all paths.
         seen = {path: [(s, e, t) for (s, e, t) in entries if now - t < ttl] for path, entries in seen.items()}
         seen = {path: entries for path, entries in seen.items() if entries}
-        store[self.SEARCH_SEEN_KEY] = seen
+        store[self.SEARCH_SEEN_RANGES_KEY] = seen
 
         seen_before = sum(len(v) for v in seen.values())
 
diff --git a/tests/unit/test_search_step.py b/tests/unit/test_search_step.py
index cfa9d575..67040f8b 100644
--- a/tests/unit/test_search_step.py
+++ b/tests/unit/test_search_step.py
@@ -667,6 +667,45 @@ def test_plain_search_steps_all_deduped_shows_all_returned_message():
     asyncio.run(run())
 
 
+def test_search_step_and_range_dedup_states_coexist_in_both_call_orders():
+    """Chunk-id and range dedup keep independent state in a shared tool context."""
+
+    async def run_order(search_first: bool):
+        chunk = _chunk("a", "daily/a.md", "first", "keyword", 5.0)
+        shared_tool_contexts: dict = {}
+        search = SearchStep(
+            file_store=FakeSearchStore(keyword_results=[chunk]),
+            tool_contexts=shared_tool_contexts,
+            vector_weight=0,
+            expand_links=False,
+        )
+        bm25 = Bm25SearchStep(
+            file_store=FakeSearchStore(keyword_results=[chunk]),
+            tool_contexts=shared_tool_contexts,
+            include_source=False,
+        )
+        ordered_steps = (search, bm25) if search_first else (bm25, search)
+
+        for step in ordered_steps:
+            first = await step(RuntimeContext(query="alpha", limit=5, tool_context_id="ctx-1"))
+            assert [result["id"] for result in first.metadata["results"]] == ["a"]
+
+        state = shared_tool_contexts["ctx-1"]
+        assert set(state) == {"search_seen_chunk_ids", "search_seen_chunk_ranges"}
+        assert set(state["search_seen_chunk_ids"]) == {"a"}
+        assert set(state["search_seen_chunk_ranges"]) == {"daily/a.md"}
+
+        for step in ordered_steps:
+            second = await step(RuntimeContext(query="alpha", limit=5, tool_context_id="ctx-1"))
+            assert second.metadata["results"] == []
+
+    async def run():
+        await run_order(search_first=True)
+        await run_order(search_first=False)
+
+    asyncio.run(run())
+
+
 def test_search_steps_no_results_shows_no_results_message():
     """When there are no results at all (before dedup), the answer explains that nothing was found."""