Database

Database and Repository Layer

SQLite is the central persistence mechanism. Runtime repositories provide CRUD-style access for users, conversations, memories, events, logs, faces, system state, local commands, personality metadata, and database metrics.

# cosmo/data/database/database.py
db = Database()

# Connection parameters:
database_path = "cosmo/data/database/cosmo.db"
row_type = sqlite3.Row
foreign_keys = ON
journal_mode = WAL
synchronous = NORMAL

Schema Overview

erDiagram
    users ||--o{ memories : has
    users ||--o{ conversations : has
    users ||--o{ faces : has
    events {
        integer id PK
        text type
        text payload
        timestamp created_at
    }
    system_state {
        text key PK
        text value
    }
    local_commands {
        integer id PK
        text intent
        text phrase
        text language
        integer active
        timestamp created_at
    }
    logs {
        integer id PK
        text level
        text logger
        text message
        text module
        text function
        integer line
        text exception
        timestamp created_at
    }
    faces {
        integer id PK
        integer user_id FK
        text encoding_path
    }

Repository Types

Logging System

Cosmo logs to console, file, and SQLite simultaneously through LoggerManager. SQLite log rows are richer than the older logger_name-only shape.

CREATE TABLE logs (
    id INTEGER PRIMARY KEY,
    level TEXT,
    logger TEXT,
    message TEXT,
    module TEXT,
    function TEXT,
    line INTEGER,
    exception TEXT,
    created_at TIMESTAMP
);

Imported historical logs may infer module and function from message patterns and leave line as null.

Log Search

LogRepository.get_recent_logs() applies filters at SQL level before ORDER BY id DESC LIMIT ?. Search currently spans level, logger, message, module, function, line, exception, and created_at. Dedicated filters exist for level, logger_name, module, and function.

Database Metrics

The WebUI compact snapshot exposes database metrics through DatabaseMetricsRepository.get_metrics().

• database_size_bytes and database_size_human.
• users_count, memories_count, conversations_count, events_count, logs_count, and faces_count.
• last_event_type and last_log_level.

Face Persistence

FaceRepository is present as a persistence layer for face records and encoding paths. It should be treated as groundwork for enrollment/recognition; the current vision pipeline detects faces but does not enroll users, generate identity embeddings, or match identities.

Async Event Bus Tracing