Python: Fix Redis FT.CREATE PREFIX sending one prefix per character of collec…

[daric93]

Fixes #13894.

The Python Redis connector was passing a plain string to redis-py's IndexDefinition(prefix=...). redis-py treats that argument as an iterable of prefixes and iterates the string character-by-character. As a result, FT.CREATE was registering one prefix per character of the collection name instead of one prefix equal to "<collection_name>:".

Observed on the wire for collection redis_json_list_data_model:

FT.CREATE redis_json_list_data_model ON JSON
  PREFIX 27 r e d i s _ j s o n _ l i s t _ d a t a _ m o d e l :
  SCORE 1.0 SCHEMA ...

Expected:

FT.CREATE redis_json_list_data_model ON JSON
  PREFIX 1 redis_json_list_data_model:
  SCORE 1.0 SCHEMA ...

RediSearch on Redis Stack silently accepts the malformed command, so the bug has been latent for single-collection users.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the SK Contribution Guidelines and the pre-submission formatting script raises no violations
• All unit tests pass, and I have added new tests where possible
• I didn't break anyone 😄

[github-actions]

Automated Code Review

Reviewers: 4 | Confidence: 94%

✓ Correctness

The change from prefix=f"{self.collection_name}:" (a string) to prefix=[f"{self.collection_name}:"] (a list) is a correct bug fix. The upstream IndexDefinition._append_prefix method (redis-py) calls len(prefix) to emit the PREFIX count argument and then iterates for p in prefix. When a bare string is passed, len() returns the character count and iteration yields individual characters, producing a malformed Redis command. Wrapping it in a list ensures the count is 1 and the single prefix string is emitted correctly.

✓ Security Reliability

This is a correct and important bug fix. The IndexDefinition._append_prefix() method (in redis-py) expects an iterable of prefix strings (typically a list), not a bare string. When a bare string is passed, len() returns the character count and iteration yields individual characters, producing a malformed Redis FT.CREATE command (e.g., PREFIX 7 m y n a m e : instead of PREFIX 1 myname:). Changing to [f"{self.collection_name}:"] fixes this. The legacy memory store at redis_memory_store.py:132 has the same latent bug with prefix=f"{collection_name}:" but is outside this diff's scope. No security or reliability concerns with the change itself.

✓ Test Coverage

The change from a string to a list for the prefix parameter in IndexDefinition is a bug fix (redis-py expects a list for prefix), but the existing unit test test_create_index does not verify the arguments passed to IndexDefinition or create_index. Since create_index is fully mocked, the test only checks that no exception is raised — it never asserts that prefix is a list. This means the fix has no regression coverage; reverting it would not cause any test to fail.

✓ Design Approach

The change correctly fixes a bug where a bare string was passed to IndexDefinition(prefix=...). The redis-py IndexDefinition._append_prefix method calls len(prefix) (expecting a list length) and then iterates over prefix element-by-element. Passing a string like "test:" caused the count to be the string length (e.g. 5) and iterated character-by-character, producing a malformed Redis FT.CREATE command. Wrapping the prefix in a list [f"{self.collection_name}:"] is the correct fix per the library's API contract.

Suggestions

• The legacy redis_memory_store.py (line 132) has the same bug: IndexDefinition(prefix=f"{collection_name}:", ...) should also be wrapped in a list. Consider fixing it in this PR for consistency.
• The test_create_index test (line 294) should assert that create_index was called with an IndexDefinition whose prefix is a list (e.g., ["test_collection:"]). Currently the mock makes zero assertions on its arguments, so this fix has no regression protection. Additionally, test_create_index_manual (line 298) passes prefix="test:" (a bare string); consider updating it to a list to match the production code and document the expected format.

---

Automated review by daric93's agents

[daric93]

@microsoft-github-policy-service agree

@microsoft-github-policy-service agree

[Copilot]

Pull request overview

Fixes a Redis Search index creation bug in the Python Redis vector store connector where FT.CREATE received one PREFIX entry per character of the collection name (due to passing a plain string into redis-py’s IndexDefinition(prefix=...)).

Changes:

• Update RedisCollection.ensure_collection_exists() to pass prefix as a list ([f"{collection_name}:"]) when constructing IndexDefinition.
• Add a unit test asserting the generated FT.CREATE ... PREFIX arguments represent a single prefix, not a character-split list.
• Update the manual IndexDefinition(prefix=...) usage in unit tests to use a list.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
python/semantic_kernel/connectors/redis.py	Fixes index definition creation to pass prefix as a list to redis-py.
python/tests/unit/connectors/memory/test_redis_store.py	Adds/updates unit tests to validate correct PREFIX argument formatting and manual IndexDefinition usage.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

There is at least one other Redis connector in this repo that still constructs IndexDefinition(prefix=f"{collection_name}:") (e.g., python/semantic_kernel/connectors/memory_stores/redis/redis_memory_store.py:132). That code path will exhibit the same character-by-character PREFIX bug against redis-py. Consider updating that call to pass a list (and/or adding coverage) so the fix is consistent across Redis connectors.

[daric93]

Closing — this fix has been combined into #13905 along with the other Redis connector bug fixes (#13896, #13904) so the integration tests can validate everything end-to-end.
Sorry for confusion