synth-ai 0.2.4.dev4__py3-none-any.whl → 0.2.4.dev5__py3-none-any.whl

synth_ai/environments/examples/enron/art_helpers/email_search_tools.py

@@ -0,0 +1,156 @@
+import sqlite3
+import logging
+import textwrap
+from typing import List, Optional
+from dataclasses import dataclass
+
+from synth_ai.environments.environment.db.sqlite import SQLiteManager
+from synth_ai.environments.examples.enron.art_helpers.types_enron import Email
+
+# Configure logger for this module
+logger = logging.getLogger(__name__)
+if not logger.handlers:  # avoid duplicate handlers in pytest -x
+    h = logging.StreamHandler()
+    h.setFormatter(logging.Formatter("%(asctime)s  %(levelname)s  %(message)s"))
+    logger.addHandler(h)
+logger.setLevel(logging.DEBUG)  # DEBUG so we see the raw SQL
+
+
+@dataclass
+class SearchResult:
+    message_id: str
+    snippet: str
+    score: float
+
+
+def search_emails(
+    sqlite_manager: SQLiteManager,
+    inbox: str,
+    keywords: List[str],
+    from_addr: Optional[str] = None,
+    to_addr: Optional[str] = None,
+    sent_after: Optional[str] = None,
+    sent_before: Optional[str] = None,
+    max_results: int = 10,
+) -> List[SearchResult]:
+    """
+    Searches the email database based on keywords, inbox, sender, recipient, and date range.
+
+    Args:
+        sqlite_manager: The SQLiteManager instance for database operations.
+        inbox: The email address of the user performing the search.
+               Results include emails sent from or to (inc. cc/bcc) this address.
+        keywords: A list of keywords that must all appear in the subject or body.
+        from_addr: Optional email address to filter emails sent *from*.
+        to_addr: Optional email address to filter emails sent *to* (inc. cc/bcc).
+        sent_after: Optional date string 'YYYY-MM-DD'. Filters for emails sent on or after this date.
+        sent_before: Optional date string 'YYYY-MM-DD'. Filters for emails sent before this date.
+        max_results: The maximum number of results to return. Cannot exceed 10.
+
+    Returns:
+        A list of SearchResult objects, each containing 'message_id' and 'snippet'.
+        Returns an empty list if no results are found or an error occurs.
+    """
+
+    if not keywords:
+        raise ValueError("No keywords provided for search.")
+    if max_results > 10:
+        # The user snippet implies max_results isn't part of the simplified SQL here.
+        # Keeping the check, but the new SQL query below does not use all filters.
+        # This might need reconciliation if all filters are intended to be used with the new SQL.
+        logger.warning(
+            "max_results > 10, but the provided SQL snippet for logging might not respect all filters."
+        )
+
+    safe_keywords = [k.replace("'", "''") for k in keywords]
+    fts_match_query = " ".join(f'"{k}"' for k in safe_keywords)
+
+    sql_query = textwrap.dedent("""
+        SELECT DISTINCT
+               e.message_id,
+               snippet(emails_fts, -1, '⟪', '⟫', ' … ', 15) AS snip
+          FROM emails e
+          JOIN emails_fts ON e.id = emails_fts.rowid
+         WHERE emails_fts MATCH ?
+         LIMIT ?
+    """).strip()
+
+    params = (fts_match_query, max_results)
+
+    try:
+        with sqlite_manager.connection() as db_conn:
+            rows = db_conn.execute(sql_query, params).fetchall()
+        return [SearchResult(message_id=row[0], snippet=row[1], score=0.0) for row in rows]
+    except sqlite3.Error as e:
+        logger.error(f"Database error during search: {e}\nSQL: {sql_query}\nParams: {params}")
+        return []
+
+
+def read_email(sqlite_manager: SQLiteManager, message_id: str) -> Optional[Email]:
+    """
+    Retrieves a single email by its message_id from the database.
+
+    Args:
+        sqlite_manager: The SQLiteManager instance for database operations.
+        message_id: The unique identifier of the email to retrieve.
+
+    Returns:
+        An Email object containing the details of the found email,
+        or None if the email is not found or an error occurs.
+    """
+
+    email_sql = """
+        SELECT id, message_id, date, subject, from_address, body, file_name
+        FROM emails
+        WHERE message_id = ?;
+    """
+
+    recipients_sql = """
+        SELECT recipient_address, recipient_type
+        FROM recipients
+        WHERE email_id = ?;
+    """
+
+    try:
+        with sqlite_manager.connection() as db_conn:
+            cursor = db_conn.cursor()
+            cursor.execute(email_sql, (message_id,))
+            email_row = cursor.fetchone()
+
+            if not email_row:
+                logging.warning(f"Email with message_id '{message_id}' not found.")
+                return None
+
+            email_id, msg_id, date, subject, from_addr, body, file_name = email_row
+            # Fetch recipients for this email primary key
+            cursor.execute(recipients_sql, (email_id,))
+            recipient_rows = cursor.fetchall()
+    except sqlite3.Error as e:
+        logger.error(f"Database error reading email {message_id}: {e}")
+        return None
+
+    to_addresses: List[str] = []
+    cc_addresses: List[str] = []
+    bcc_addresses: List[str] = []
+
+    for addr, type_val in recipient_rows:
+        type_lower = type_val.lower()
+        if type_lower == "to":
+            to_addresses.append(addr)
+        elif type_lower == "cc":
+            cc_addresses.append(addr)
+        elif type_lower == "bcc":
+            bcc_addresses.append(addr)
+
+    email_obj = Email(
+        message_id=msg_id,
+        date=date,
+        subject=subject,
+        from_address=from_addr,
+        to_addresses=to_addresses,
+        cc_addresses=cc_addresses,
+        bcc_addresses=bcc_addresses,
+        body=body,
+        file_name=file_name,
+    )
+    return email_obj

synth_ai/environments/examples/enron/art_helpers/local_email_db.py

@@ -0,0 +1,280 @@
+import sqlite3
+import os
+import logging
+from datasets import load_dataset, Dataset, Features, Value, Sequence
+from tqdm import tqdm
+from datetime import datetime
+
+# Resolve paths relative to this file so it works regardless of the current working directory
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+# Database will live in "../data/enron_emails.db" relative to project root
+DEFAULT_DB_PATH = os.path.join(BASE_DIR, "..", "..", "data", "enron_emails.db")
+
+DEFAULT_REPO_ID = "corbt/enron-emails"
+
+logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
+
+
+# --- Database Schema ---
+SQL_CREATE_TABLES = """
+DROP TABLE IF EXISTS recipients;
+DROP TABLE IF EXISTS emails_fts;
+DROP TABLE IF EXISTS emails;
+
+CREATE TABLE emails (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    message_id TEXT UNIQUE,
+    subject TEXT,
+    from_address TEXT,
+    date TEXT, -- Store as ISO 8601 string 'YYYY-MM-DD HH:MM:SS'
+    body TEXT,
+    file_name TEXT
+);
+
+CREATE TABLE recipients (
+    email_id INTEGER,
+    recipient_address TEXT,
+    recipient_type TEXT, -- 'to', 'cc', 'bcc'
+    FOREIGN KEY(email_id) REFERENCES emails(id) ON DELETE CASCADE
+);
+"""
+
+SQL_CREATE_INDEXES_TRIGGERS = """
+CREATE INDEX idx_emails_from ON emails(from_address);
+CREATE INDEX idx_emails_date ON emails(date);
+CREATE INDEX idx_emails_message_id ON emails(message_id);
+CREATE INDEX idx_recipients_address ON recipients(recipient_address);
+CREATE INDEX idx_recipients_type ON recipients(recipient_type);
+CREATE INDEX idx_recipients_email_id ON recipients(email_id);
+CREATE INDEX idx_recipients_address_email ON recipients(recipient_address, email_id);
+
+CREATE VIRTUAL TABLE emails_fts USING fts5(
+    subject,
+    body,
+    content='emails',
+    content_rowid='id'
+);
+
+CREATE TRIGGER emails_ai AFTER INSERT ON emails BEGIN
+    INSERT INTO emails_fts (rowid, subject, body)
+    VALUES (new.id, new.subject, new.body);
+END;
+
+CREATE TRIGGER emails_ad AFTER DELETE ON emails BEGIN
+    DELETE FROM emails_fts WHERE rowid=old.id;
+END;
+
+CREATE TRIGGER emails_au AFTER UPDATE ON emails BEGIN
+    UPDATE emails_fts SET subject=new.subject, body=new.body WHERE rowid=old.id;
+END;
+
+INSERT INTO emails_fts (rowid, subject, body) SELECT id, subject, body FROM emails;
+"""
+
+
+# --- Functions ---
+
+
+def download_dataset(repo_id: str) -> Dataset:
+    """Downloads the dataset from Hugging Face Hub."""
+    logging.info(f"Attempting to download dataset from Hugging Face Hub: {repo_id}")
+    expected_features = Features(
+        {
+            "message_id": Value("string"),
+            "subject": Value("string"),
+            "from": Value("string"),
+            "to": Sequence(Value("string")),
+            "cc": Sequence(Value("string")),
+            "bcc": Sequence(Value("string")),
+            "date": Value("timestamp[us]"),
+            "body": Value("string"),
+            "file_name": Value("string"),
+        }
+    )
+    dataset_obj = load_dataset(repo_id, features=expected_features, split="train")
+    # Basic type check remains useful
+    if not isinstance(dataset_obj, Dataset):
+        raise TypeError(f"Expected Dataset, got {type(dataset_obj)}")
+    logging.info(f"Successfully loaded dataset '{repo_id}' with {len(dataset_obj)} records.")
+    return dataset_obj
+
+
+def create_database(db_path: str):
+    """Creates the SQLite database and tables."""
+    logging.info(f"Creating SQLite database and tables at: {db_path}")
+    conn = sqlite3.connect(db_path)
+    cursor = conn.cursor()
+    cursor.executescript(SQL_CREATE_TABLES)
+    conn.commit()
+    conn.close()
+    logging.info("Database tables created successfully.")
+
+
+def populate_database(db_path: str, dataset: Dataset):
+    """Populates the database with data from the Hugging Face dataset."""
+    logging.info(f"Populating database {db_path}...")
+    conn = sqlite3.connect(db_path)
+    cursor = conn.cursor()
+
+    # --- Performance Pragmas ---
+    conn.execute("PRAGMA synchronous = OFF;")
+    conn.execute("PRAGMA journal_mode = MEMORY;")
+
+    record_count = 0
+    skipped_count = 0  # Keep track of skipped emails due to filters
+    duplicate_count = 0  # Keep track of skipped duplicate emails
+    processed_emails = set()  # Track (subject, body, from) tuples to dedupe
+
+    conn.execute("BEGIN TRANSACTION;")  # Single transaction for bulk insert
+
+    for email_data in tqdm(dataset, desc="Inserting emails"):
+        assert isinstance(email_data, dict)
+        message_id = email_data["message_id"]
+        subject = email_data["subject"]
+        from_address = email_data["from"]
+        date_obj: datetime = email_data["date"]
+        body = email_data["body"]
+        file_name = email_data["file_name"]
+        to_list_raw = email_data["to"]
+        cc_list_raw = email_data["cc"]
+        bcc_list_raw = email_data["bcc"]
+
+        # --- Data Cleaning and Filtering ---
+        date_str = date_obj.strftime("%Y-%m-%d %H:%M:%S")
+        to_list = [str(addr) for addr in to_list_raw if addr]
+        cc_list = [str(addr) for addr in cc_list_raw if addr]
+        bcc_list = [str(addr) for addr in bcc_list_raw if addr]
+
+        # Check body length
+        if len(body) > 5000:
+            logging.debug(f"Skipping email {message_id}: Body length > 5000 characters.")
+            skipped_count += 1
+            continue
+
+        # Check total recipients
+        total_recipients = len(to_list) + len(cc_list) + len(bcc_list)
+        if total_recipients > 30:
+            logging.debug(
+                f"Skipping email {message_id}: Total recipients ({total_recipients}) > 30."
+            )
+            skipped_count += 1
+            continue
+        # --- End Filtering ---
+
+        # --- Deduplication Check ---
+        email_key = (subject, body, from_address)
+        if email_key in processed_emails:
+            logging.debug(
+                f"Skipping duplicate email (Subject: {subject[:50]}..., From: {from_address})"
+            )
+            duplicate_count += 1
+            continue
+        else:
+            processed_emails.add(email_key)
+        # --- End Deduplication ---
+
+        cursor.execute(
+            """
+            INSERT INTO emails (message_id, subject, from_address, date, body, file_name)
+            VALUES (?, ?, ?, ?, ?, ?)
+            """,
+            (message_id, subject, from_address, date_str, body, file_name),
+        )
+        email_pk_id = cursor.lastrowid
+
+        recipient_data = []
+        for addr in to_list:
+            recipient_data.append((email_pk_id, addr, "to"))
+        for addr in cc_list:
+            recipient_data.append((email_pk_id, addr, "cc"))
+        for addr in bcc_list:
+            recipient_data.append((email_pk_id, addr, "bcc"))
+
+        if recipient_data:
+            cursor.executemany(
+                """
+                INSERT INTO recipients (email_id, recipient_address, recipient_type)
+                VALUES (?, ?, ?)
+                """,
+                recipient_data,
+            )
+        record_count += 1
+
+    conn.commit()
+    conn.close()
+    logging.info(f"Successfully inserted {record_count} email records.")
+    if skipped_count > 0:
+        logging.info(f"Skipped {skipped_count} email records due to length or recipient limits.")
+    if duplicate_count > 0:
+        logging.info(
+            f"Skipped {duplicate_count} duplicate email records (based on subject, body, from)."
+        )
+
+
+def create_indexes_and_triggers(db_path: str):
+    """Creates indexes and triggers on the populated database."""
+    logging.info(f"Creating indexes and triggers for database: {db_path}...")
+    conn = sqlite3.connect(db_path)
+    cursor = conn.cursor()
+    cursor.executescript(SQL_CREATE_INDEXES_TRIGGERS)
+    conn.commit()
+    conn.close()
+    logging.info("Indexes and triggers created successfully.")
+
+
+def generate_database(overwrite: bool = False):
+    """
+    Generates the SQLite database from the specified Hugging Face dataset.
+    Simplified version without extensive error handling.
+
+    Args:
+        repo_id: The Hugging Face repository ID for the dataset.
+        db_path: The path where the SQLite database file should be created.
+        overwrite: If True, any existing database file at db_path will be removed.
+    """
+    logging.info(
+        f"Starting database generation for repo '{DEFAULT_REPO_ID}' at '{DEFAULT_DB_PATH}'"
+    )
+    logging.info(f"Overwrite existing database: {overwrite}")
+
+    db_dir = os.path.dirname(DEFAULT_DB_PATH)
+    if db_dir and not os.path.exists(db_dir):
+        logging.info(f"Creating data directory: {db_dir}")
+        os.makedirs(db_dir)
+
+    if overwrite and os.path.exists(DEFAULT_DB_PATH):
+        logging.info(f"Removing existing database file: {DEFAULT_DB_PATH}")
+        os.remove(DEFAULT_DB_PATH)
+
+    if not os.path.exists(DEFAULT_DB_PATH):
+        dataset = download_dataset(DEFAULT_REPO_ID)
+        create_database(DEFAULT_DB_PATH)
+        populate_database(DEFAULT_DB_PATH, dataset)
+        create_indexes_and_triggers(DEFAULT_DB_PATH)
+
+        # ---- new: add unique index post-creation if not already handled by SQL_CREATE_INDEXES_TRIGGERS ---
+        # This ensures the index exists even if SQL_CREATE_INDEXES_TRIGGERS was modified
+        # or if we want to be absolutely certain this specific index is applied.
+        conn = sqlite3.connect(DEFAULT_DB_PATH)
+        cur = conn.cursor()
+        logging.info("Ensuring UNIQUE index on emails.message_id exists...")
+        cur.executescript(
+            """
+            CREATE UNIQUE INDEX IF NOT EXISTS idx_emails_message_id
+                ON emails(message_id);
+            """
+        )
+        conn.commit()
+        conn.close()
+        logging.info("UNIQUE index on emails.message_id verified/created.")
+        # ---- end new section ----
+    else:
+        logging.info(
+            f"Database already exists at {DEFAULT_DB_PATH}. Set overwrite=True to regenerate."
+        )
+
+    logging.info("Database generation process complete.")
+
+
+if __name__ == "__main__":
+    generate_database(overwrite=True)

synth_ai/environments/examples/enron/art_helpers/types_enron.py

@@ -0,0 +1,24 @@
+from pydantic import BaseModel
+from typing import List, Optional
+
+
+class SyntheticQuery(BaseModel):
+    id: int
+    question: str
+    answer: str
+    message_ids: List[str]  # message_ids (strings) of referenced emails
+    how_realistic: float
+    inbox_address: str
+    query_date: str
+
+
+class Email(BaseModel):
+    message_id: str
+    date: str  # ISO 8601 string 'YYYY-MM-DD HH:MM:SS'
+    subject: Optional[str] = None
+    from_address: Optional[str] = None
+    to_addresses: List[str] = []  # Populated from recipients table
+    cc_addresses: List[str] = []  # Populated from recipients table
+    bcc_addresses: List[str] = []  # Populated from recipients table
+    body: Optional[str] = None
+    file_name: Optional[str] = None