fix(vector_stores): pass namespace as top-level kwarg to UpstashVector query_many

[gaurav0107]

Linked Issue

Closes #4207

Description

UpstashVector.search() was passing namespace inside each per-query
dict to upstash_vector.Index.query_many(). The Upstash client expects
namespace as a top-level keyword argument; the QueryRequest
TypedDict has no namespace field, so the per-query value was silently
dropped and every query routed to the default namespace.

When len(queries) == 1, query_many internally calls
self.query(**query, namespace=namespace) — which raises
TypeError: got multiple values for keyword argument 'namespace' because
namespace ends up bound twice.

This PR moves namespace from each query dict onto query_many's
top-level kwarg, matching the Upstash client signature
(upstash_vector.core.index_operations.IndexOperations.query_many,
v0.8.0).

-            queries = [
-                {
-                    "vector": v,
-                    "top_k": top_k,
-                    "filter": filters_str or "",
-                    "include_metadata": True,
-                    "namespace": self.collection_name,
-                }
-                for v in vectors
-            ]
-            responses = self.client.query_many(queries=queries)
+            queries = [
+                {
+                    "vector": v,
+                    "top_k": top_k,
+                    "filter": filters_str or "",
+                    "include_metadata": True,
+                }
+                for v in vectors
+            ]
+            responses = self.client.query_many(queries=queries, namespace=self.collection_name)

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Refactor (no functional changes)
• Documentation update

Breaking Changes

N/A — this fixes the search path so enable_embeddings=False users
actually receive results from their configured namespace. Users who never
set a custom collection_name saw no visible effect; users who did were
silently returned data from the wrong namespace, or hit a TypeError on
single-query search.

Test Coverage

• I added/updated unit tests
• I added/updated integration tests
• I tested manually (describe below)
• No tests needed (explain why)

Updated test_search_vectors and test_search_vectors_empty_filters to
assert the new (correct) call shape, and added
test_search_vectors_multi_query_namespace_at_top_level as a regression
test that locks in the fix on the multi-query path. The new test was
verified to fail on the buggy code and pass on the fix.

tests/vector_stores/test_upstash_vector.py ............. 19 passed

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my code
• I have added tests that prove my fix/feature works
• New and existing tests pass locally
• I have updated documentation if needed (not applicable — internal
  bug fix in vector store call shape, no public API change)

[CLAassistant]

All committers have signed the CLA.

[gaurav0107]

Friendly check-in on this one — the fix is small (move namespace to the query_many top-level kwarg, matching the Upstash client signature), CLA is signed, tests pass locally, and the only failing check is the Vercel preview-deploy (auth, not a code issue). Happy to rebase onto current main or split the test into a separate file if that helps reviewers — just let me know what would make this easier to land.

Thanks for taking a look when you get a chance!

[kartik-mem0]

LGTM

Linked issue: #4207 (P1-high) — validated
Bug reproduction: Confirmed via code tracing — query_many takes namespace as a top-level kwarg, not per-query. Passing it per-query caused TypeError on single-vector search and silent wrong-namespace routing on multi-vector search.
Pre-push gate: ruff lint clean, isort clean, 19/19 tests pass
Independent review: approved (separate unbiased reviewer)
Upstash API verification: Checked against current Upstash Vector Python SDK docs — namespace is indeed a top-level query_many parameter, not part of the per-query QueryRequest dict.

The fix is minimal and correct: moves namespace from each per-query dict to the query_many(namespace=...) top-level kwarg, matching how every other method in the same file (upsert, delete, update, fetch, resumable_query, reset) already passes it. The new test_search_vectors_multi_query_namespace_at_top_level regression test is thorough — it verifies both the call shape and the result flattening across multiple query response lists.

Optional (pre-existing, not introduced by this PR):

• tests/vector_stores/test_upstash_vector.py:134 has a stray print("filters", filters) debug statement
• mem0/vector_stores/upstash_vector.py has a minor ruff format drift (pre-existing)

Ready for human review and merge.