Database Configuration

PyBA can log all actions to a database, enabling code generation, auditing, and replay capabilities.

Overview 

The database stores:

Session information: Unique IDs for each automation run
Actions: Every action taken during automation
Page URLs: Where each action was performed
Status: Success/failure of each action
Extraction results: Data extracted during automation

Supported Databases 

PyBA supports three database engines:

Engine	Type	Best For
SQLite	File-based	Local development
PostgreSQL	Server	Production, teams
MySQL	Server	Production, teams

SQLite Setup 

The simplest option—no server required:

from pyba import Engine, Database

database = Database(
    engine="sqlite",
    name="/tmp/pyba.db"  # Path to the database file
)

engine = Engine(
    openai_api_key="sk-...",
    database=database
)

The database file is created automatically if it doesn’t exist.

PostgreSQL Setup 

For production environments:

from pyba import Engine, Database

database = Database(
    engine="postgres",
    name="pyba_db",           # Database name
    host="localhost",         # Or your server IP
    port=5432,                # Default PostgreSQL port
    username="your_user",
    password="your_password",
    ssl_mode="disable"        # Or "require" for encrypted connections
)

engine = Engine(
    openai_api_key="sk-...",
    database=database
)

SSL Mode options:

"disable" - No encryption (local development)
"require" - Encrypted connection (production)

MySQL Setup 

from pyba import Engine, Database

database = Database(
    engine="mysql",
    name="pyba_db",           # Database name
    host="localhost",         # Or your server IP
    port=3306,                # Default MySQL port
    username="your_user",
    password="your_password"
)

engine = Engine(
    openai_api_key="sk-...",
    database=database
)

Database Schema 

PyBA creates an EpisodicMemory table to store automation logs:

CREATE TABLE EpisodicMemory (
    id INTEGER PRIMARY KEY,
    session_id VARCHAR(64),
    actions TEXT,              -- JSON array of actions
    urls TEXT,                 -- JSON array of URLs
    action_status BOOLEAN,     -- Success/failure
    fail_reason TEXT           -- Error message if failed
);

For BFS mode, there’s an additional BFSEpisodicMemory table with a context_id field to track parallel browser sessions.

Querying the Database 

SQLite Example 

sqlite3 /tmp/pyba.db

# List tables
.tables

# View all sessions
SELECT DISTINCT session_id FROM EpisodicMemory;

# View actions for a session
SELECT * FROM EpisodicMemory WHERE session_id = 'your-session-id';

Python Example 

from pyba.database import DatabaseFunctions

# After running automation with a database
db_funcs = DatabaseFunctions(database)

# Get logs for a session
logs = db_funcs.get_episodic_memory_by_session_id(session_id="your-session-id")

print(logs.actions)  # JSON string of actions
print(logs.urls)     # JSON string of URLs

Code Generation 

The main reason to use a database: generating reusable scripts.

from pyba import Engine, Database

db = Database(engine="sqlite", name="/tmp/pyba.db")
engine = Engine(openai_api_key="sk-...", database=db)

# Run the automation
engine.sync_run(prompt="Go to example.com and fill the contact form")

# Generate a standalone script
engine.generate_code(output_path="/tmp/contact_form.py")

The generated script:

Uses playwright.sync_api directly
Replays the exact actions from the database
Has no AI dependencies—runs forever without API costs

See Code Generation for more details.

Configuration via YAML 

You can also configure database defaults in pyba/config.yaml:

database:
  engine: "sqlite"
  name: "/tmp/pyba.db"
  host: "localhost"
  port: 5432
  username: ""
  password: ""
  ssl_mode: "disable"

These serve as fallbacks when parameters aren’t provided to the Database constructor.

Best Practices 

Security 

Never put passwords directly in code:

import os

database = Database(
    engine="postgres",
    name="pyba_db",
    host=os.environ["DB_HOST"],
    username=os.environ["DB_USER"],
    password=os.environ["DB_PASSWORD"],
)

File Paths 

For SQLite, use absolute paths to avoid confusion:

# Good
database = Database(engine="sqlite", name="/tmp/pyba/runs.db")

# Risky - depends on working directory
database = Database(engine="sqlite", name="runs.db")

Session Management 

Each Engine run gets a unique session_id. You can access it:

engine.sync_run(prompt="Your task")
print(f"Session ID: {engine.session_id}")

Use this ID to query logs or generate code for specific runs.

Cleanup 

Database files grow over time. Periodically clean up old sessions:

-- SQLite: Delete sessions older than 7 days
DELETE FROM EpisodicMemory
WHERE session_id IN (
    SELECT DISTINCT session_id
    FROM EpisodicMemory
    GROUP BY session_id
    HAVING MAX(id) < (SELECT MAX(id) - 1000 FROM EpisodicMemory)
);