SqliteMemoryService with FTS5 finds nothing for a Chinese or Japanese keyword inside a sentence #7959

🔴 Required Information

Describe the Bug:
SqliteMemoryService cannot find a stored Chinese or Japanese memory by a keyword when FTS5 is used, which is the default (fts="auto").

The memory 我想买这辆车，价格是二十万。 ("I want to buy this car, the price is 200k.") is not found by searching 价格 ("price") or 二十万 ("200k"). The same search finds it with fts="off" and with InMemoryMemoryService. English keywords work in all three.

Steps to Reproduce:

1. Install ADK from main (reproduced at 3565ad3dd)
2. Save the script below as repro.py
3. Run python repro.py

Expected Behavior:
价格 and 二十万 each return 1 memory in every service and mode, like price does.

Observed Behavior:

InMemoryMemoryService              query='price'    -> 1 memories
InMemoryMemoryService              query='价格'       -> 1 memories
InMemoryMemoryService              query='二十万'      -> 1 memories
SqliteMemoryService(fts='auto')    query='price'    -> 1 memories
SqliteMemoryService(fts='auto')    query='价格'       -> 0 memories
SqliteMemoryService(fts='auto')    query='二十万'      -> 0 memories
SqliteMemoryService(fts='off')     query='price'    -> 1 memories
SqliteMemoryService(fts='off')     query='价格'       -> 1 memories
SqliteMemoryService(fts='off')     query='二十万'      -> 1 memories

Environment Details:

• ADK Library Version (pip show google-adk): main at 3565ad3dd (version string 2.13.0)
• Desktop OS: macOS 26.6.2
• Python Version (python -V): 3.12.7

Model Information:

• Are you using LiteLLM: No
• Which model is being used: N/A (no model call)

🟡 Optional Information

Additional Context:
The FTS5 table uses tokenize='unicode61'. Chinese and Japanese are written without spaces, so unicode61 stores a whole run of characters between punctuation (e.g. 价格是二十万) as one token, and a keyword inside it never matches. The LIKE path (fts="off") does substring matching, so the two modes give different results for the same data.

The existing test test_search_cjk_mixed_latin_query only searches for the Latin word Python inside 私はPythonを使う; searching 使う in the same text also returns nothing with FTS. InMemoryMemoryService already covers this case (('我喜欢机器学习', '机器学习', 1) in its tests).

I found this while building a Chinese-language voice agent. I'd be happy to send a PR.

Minimal Reproduction Code:

import asyncio

from google.adk.events.event import Event
from google.adk.memory import InMemoryMemoryService, SqliteMemoryService
from google.adk.sessions.session import Session
from google.genai import types

session = Session(
    id="s1",
    app_name="app",
    user_id="u1",
    events=[
        Event(
            id="e1",
            author="user",
            timestamp=1.0,
            content=types.Content(
                role="user",
                parts=[types.Part(text="我想买这辆车，价格是二十万。")],
            ),
        ),
        Event(
            id="e2",
            author="user",
            timestamp=2.0,
            content=types.Content(
                role="user", parts=[types.Part(text="The price is 200k.")]
            ),
        ),
    ],
)


async def main():
  services = {
      "InMemoryMemoryService": InMemoryMemoryService(),
      "SqliteMemoryService(fts='auto')": SqliteMemoryService(":memory:"),
      "SqliteMemoryService(fts='off')": SqliteMemoryService(
          ":memory:", fts="off"
      ),
  }
  for name, service in services.items():
    await service.add_session_to_memory(session)
    for query in ["price", "价格", "二十万"]:
      response = await service.search_memory(
          app_name="app", user_id="u1", query=query
      )
      print(f"{name:34} query={query!r:10} -> {len(response.memories)} memories")
    if hasattr(service, "close"):
      await service.close()


asyncio.run(main())

How often has this issue occurred?:

• Always (100%)