fix(neo4j): route database_ to connection kwarg (#1481)

[gauravch-code]

Summary

Fixes #1481.

Neo4jDriver.execute_query was placing database_ inside parameters_ (the Cypher variables dict) rather than passing it as a driver-level kwarg to Neo4j's execute_query. As a result, database_ was treated as an unused Cypher parameter and silently ignored — every query routed to the connection's home database regardless of the database value passed to Neo4jDriver(...). Only writes that went through session(database=...) explicitly (e.g. transaction()) hit the intended database.

Fix

Route database_ to the driver-level kwarg on AsyncDriver.execute_query. Accept it from either params or kwargs for backwards compatibility so callers that historically stuck it in params keep working.

database = (
    params.pop('database_', None)
    or kwargs.pop('database_', None)
    or self._database
)
result = await self.client.execute_query(
    cypher_query_,
    parameters_=params,
    database_=database,
    **kwargs,
)

Tests

New tests/driver/test_neo4j_driver.py — 7 tests, mock-only (no live Neo4j required):

• test_configured_database_is_passed_as_connection_kwarg — the core regression for [BUG] Database Parameter Not Honored in Neo4jDriver.execute_query() #1481
• test_default_database_is_used_when_none_configured
• test_explicit_database_kwarg_overrides_configured_default
• test_database_nested_in_params_is_promoted_to_connection_kwarg — backwards compat
• test_routing_kwarg_is_forwarded_at_connection_level
• test_extra_kwargs_are_forwarded
• test_provider

Added the new file to the database-integration-tests job in .github/workflows/unit_tests.yml, mirroring the existing test_falkordb_driver.py entry.

Verified locally

• ruff format + ruff check — clean
• pyright graphiti_core/driver/neo4j_driver.py — 0 errors, 0 warnings
• pytest tests/driver/test_falkordb_driver.py tests/driver/test_neo4j_driver.py — 32 passed, 1 skipped
• Full unit suite with CI's ignore list — 366 passed, 11 skipped, no regressions

Test plan

• Regression test fails on unmodified main (asserts database_ is a top-level kwarg; observed {'parameters_': {'database_': 'mydb'}} on main)
• Regression test passes after fix
• Full unit suite passes
• ruff + pyright clean

[zep-cla-assistant]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. For privacy information, see our Privacy Notice. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA behalf on myself, e-mail: example@example.com

or

I have read the CLA Document and I hereby sign the CLA behalf of my company, e-mail: example@example.com

Signature is valid for 6 months.

---

_{This bot will be retriggered when the Contributor License Agreement comment has been provided. Posted by the CLA Assistant Lite bot.}

[gauravch-code]

I have read the CLA Document and I hereby sign the CLA behalf on myself, e-mail: gaurav.pvt25@gmail.com

[mehulp93]

Thanks for contributing to Graphiti and for diagnosing this.

We’ve shipped the fix in graphiti-core v0.30.1 and MCP v1.1.0: database_ is now passed as a Neo4j connection argument, and the MCP server honors NEO4J_DATABASE. We’re closing this PR in favor of that release.

Please note:

• If your server’s home database is not named neo4j and you don’t pass database to the driver, queries now target neo4j instead of your home database — pass your database name explicitly to keep the old behavior.
• If you configured a custom database, reads now correctly target it. Data previously written via execute_query may reside in your home database and may need to be migrated.