[yashs33244]

Linked Issue

Closes #6562

Description

_convert_filters in the Turbopuffer vector store only handled gte and lte for operator dicts. Any other operator mem0 supports (gt, lt, ne, eq, in, nin) matched neither branch, so a filter like {"age": {"gt": 18}} was silently dropped and the query ran unfiltered, returning too-broad results with no error.

This maps the full mem0 operator set to the Turbopuffer filter tokens (verified against the SDKs Filter type in turbopuffer/types/custom.py: eq->Eq, ne->NotEq, gt->Gt, gte->Gte, lt->Lt, lte->Lte, in->In, nin->NotIn) and raises a clear ValueError on an unknown operator instead of dropping it. Existing scalar-equality and gte/lte behavior is unchanged.

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

Test Coverage

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

Added regression tests in tests/vector_stores/test_turbopuffer.py: every operator maps to the right token, {"gt": 18} is no longer dropped, multiple operators on one field combine with And, and an unknown operator raises ValueError. Verified the new tests fail on the unpatched source.

Checklist

• My code follows the projects style guidelines
• I have performed a self-review of my code
• I have added tests that prove my fix works
• New and existing tests pass locally
• I have updated documentation if needed

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[yashs33244]

Open ~19 days, green CI, unreviewed 🙏 Turbopuffer _convert_filters now applies all operators (gt/lt/ne/in/nin) instead of silently dropping them; unknown ops raise. 4 tests. @kartik-mem0 @deshraj would appreciate a review.

[blazingretinol]

Changes requested

Correctly fixes the reported operator-drop bug (now merged), but the same _convert_filters still mistranslates "*" wildcard and bare-array filters into literal Eq matches that return nothing.

Worth a fast-follow:

1. mem0/vector_stores/turbopuffer.py:167 - guard value == "*" (match-any) and bare-list-as-in, same as every other Python backend and the merged TS sibling #6578

Optional: contains/icontains now raise instead of silently dropping (line 135); also, open PR #6878 edits this same method and will need a small rebase against this merge.

How I verified

• Problem is real: mem0/vector_stores/turbopuffer.py pre-PR (lines ~146-149) dropped any operator besides gte/lte; db._convert_filters({"age": {"gt": 18}}) returned None, matching issue #6562
• Tests / red-green: PYTHONPATH=. pytest tests/vector_stores/test_turbopuffer.py -v: 75 passed on head; reverting turbopuffer.py to pre-PR main failed the 9 new operator tests, 10 pre-existing passed; restoring returned to 75 passed
• CI: real checks (tests, lint, build, license/cla) green; Vercel preview-deploy red, not merge-blocking
• Security: cleared, filter values pass through as opaque tuples to the typed SDK client, no injection surface
• Bloat (ponytail): lean already, table-driven operator map is the right shape
• Not verified: live Turbopuffer API (hosted-only, no sandbox lane); unit-level mocked-client tests only

Solid, well-tested fix for the issue it names. Since this already merged, a short follow-up PR closing the wildcard/array gap in the same function would finish the job.