Broca 2 Hits 1,080 Tests Green — From 653 to Full Coverage

The Numbers

On October 13, Broca 2 had 653 passing tests and a long tail of failures across Telegram handlers, web chat, database operations, and the CLI. Three days later: 1,080 passed, 0 failed.

What It Took

The root causes were structural, not superficial:

• AsyncMock isolation: Singleton patterns in the Letta client and plugin manager leaked state between tests. A HybridExecuteMock pattern was introduced to handle both await and async with contexts.
• Database path handling: DB_PATH was made dynamic to respect TEST_DB_PATH, so tests don't clobber each other.
• Telegram bot plugin: Patch targets were corrected module by module — emit_event, register_event_handler, start, stop, validate_settings — until all 27 tests passed.
• CLI tests: 257 CLI test failures were fixed in one session by correcting how btool, qtool, ctool, and settings commands were mocked.
• Pydantic V2 migration: Deprecated V1-style validators were migrated to ConfigDict and V2 field validators, silencing a wave of deprecation warnings.

Why This Matters

Broca is the message middleware that sits between your agent and the outside world (Telegram, web chat, future platforms). A 100% green test suite means contributors can add new plugins — or refactor existing ones — with confidence that they haven't broken the message pipeline. That's the difference between a prototype and infrastructure.