fix(arcadedb): wire up fulltext/similarity search (BM25 currently crashes)

[agc-63]

Hey — while stress-testing the driver end-to-end for #1, ran into another one: ArcadeDBSearchOperations already has fulltext/similarity search implemented correctly (CONTAINS-based, sidesteps the Neo4j-only db.index.fulltext.query* procedures ArcadeDB doesn't support) — it just was never connected via search_interface. Same story for the other 4 providers' equivalent search_ops classes, per the driver-operations-architecture-redesign spec — this migration is still in progress project-wide, not an ArcadeDB-specific gap. In the meantime, search_utils.py checks driver.search_interface before falling back to the generic Neo4j-assuming path, and ArcadeDBDriver never set it, so every fulltext search crashed today.

Fix: a thin ArcadeDBSearchInterface adapter delegating to the existing (correct!) ArcadeDBSearchOperations, wired in ArcadeDBDriver.__init__. Wired exactly the 6 methods that search_utils.py calls without a try/except NotImplementedError safety net — the rest (6 more) keep using their already-working generic path, so this can't regress anything.

Also found and fixed 2 pre-existing bugs while finally exercising this end-to-end for the first time:

1. node_similarity_search / edge_similarity_search both built an invalid double-WHERE when a search filter was present.
2. lucene_sanitize() was used to build the CONTAINS search terms — it Lucene-escapes special characters with a literal backslash ('Alice' → '\Alice'), which broke almost every fulltext query silently (returned empty instead of matching). Verified live with real queries after the fix ('Alice' now correctly finds 'Alice Smith').

Known, pre-existing follow-ups not fixed here (neither used by us today): list_has_all() isn't implemented by ArcadeDB's Cypher engine (crashes if SearchFilters.node_labels is used against ArcadeDB); edge_bfs_search's generic fallback silently returns [] for one specific MATCH pattern shape.

Verified against a real 26.8.1-SNAPSHOT instance: full add_episode → retrieve_episodes → search cycle now returns correct BM25 results instead of crashing.

[gemini-code-assist]

Code Review

This pull request integrates ArcadeDB-specific search operations by implementing the ArcadeDBSearchInterface, resolving crashes caused by unsupported Neo4j fulltext procedures. It also fixes fulltext search matching by using raw query terms instead of Lucene-sanitized strings, and prevents engine-level datetime comparison errors in ArcadeDB by serializing datetime parameters to ISO strings. The review feedback points out a critical issue where ArcadeDBSearchInterface overrides the constructor of a Pydantic BaseModel without calling super().__init__(), which can bypass Pydantic's internal initialization and should be corrected.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Overriding __init__ on a Pydantic BaseModel (which SearchInterface inherits from) without calling super().__init__() bypasses Pydantic's internal initialization process. This can cause runtime errors or unexpected behavior when Pydantic's internal features (such as validation, serialization, or field tracking) are utilized.

To fix this, ensure super().__init__() is called within the constructor, accepting and forwarding any keyword arguments (**data).

Suggested change

	def __init__(self):
	self._ops = ArcadeDBSearchOperations()
	def __init__(self, **data: Any):
	super().__init__(**data)
	self._ops = ArcadeDBSearchOperations()