agno 2.1.2__py3-none-any.whl → 2.3.13__py3-none-any.whl

agno/db/sqlite/utils.py

@@ -4,6 +4,9 @@ from datetime import date, datetime, timedelta, timezone
 from typing import Any, Dict, List, Optional
 from uuid import uuid4
 
+from sqlalchemy.ext.asyncio import AsyncEngine, AsyncSession
+
+from agno.db.schemas.culture import CulturalKnowledge
 from agno.db.sqlite.schemas import get_table_schema_definition
 from agno.utils.log import log_debug, log_error, log_warning
 
@@ -49,6 +52,7 @@ def is_table_available(session: Session, table_name: str, db_schema: Optional[st
     """
     Check if a table with the given name exists.
     Note: db_schema parameter is ignored in SQLite but kept for API compatibility.
+
     Returns:
         bool: True if the table exists, False otherwise.
     """
@@ -64,15 +68,34 @@ def is_table_available(session: Session, table_name: str, db_schema: Optional[st
         return False
 
 
-def is_valid_table(db_engine: Engine, table_name: str, table_type: str, db_schema: Optional[str] = None) -> bool:
+async def ais_table_available(session: AsyncSession, table_name: str, db_schema: Optional[str] = None) -> bool:
     """
-    Check if the existing table has the expected column names.
+    Check if a table with the given name exists.
     Note: db_schema parameter is ignored in SQLite but kept for API compatibility.
+
+    Returns:
+        bool: True if the table exists, False otherwise.
+    """
+    try:
+        exists_query = text("SELECT 1 FROM sqlite_master WHERE type = 'table' AND name = :table")
+        exists = (await session.execute(exists_query, {"table": table_name})).scalar() is not None
+        if not exists:
+            log_debug(f"Table {table_name} {'exists' if exists else 'does not exist'}")
+        return exists
+    except Exception as e:
+        log_error(f"Error checking if table exists: {e}")
+        return False
+
+
+def is_valid_table(db_engine: Engine, table_name: str, table_type: str) -> bool:
+    """
+    Check if the existing table has the expected column names.
+
     Args:
         db_engine (Engine): Database engine
         table_name (str): Name of the table to validate
         table_type (str): Type of table to get expected schema
-        db_schema (Optional[str]): Database schema name (ignored in SQLite)
+
     Returns:
         bool: True if table has all expected columns, False otherwise
     """
@@ -97,6 +120,45 @@ def is_valid_table(db_engine: Engine, table_name: str, table_type: str, db_schem
         return False
 
 
+async def ais_valid_table(db_engine: AsyncEngine, table_name: str, table_type: str) -> bool:
+    """
+    Check if the existing table has the expected column names.
+
+    Args:
+        db_engine (Engine): Database engine
+        table_name (str): Name of the table to validate
+        table_type (str): Type of table to get expected schema
+
+    Returns:
+        bool: True if table has all expected columns, False otherwise
+    """
+    try:
+        expected_table_schema = get_table_schema_definition(table_type)
+        expected_columns = {col_name for col_name in expected_table_schema.keys() if not col_name.startswith("_")}
+
+        # Get existing columns from the async engine
+        async with db_engine.connect() as conn:
+            existing_columns = await conn.run_sync(_get_table_columns, table_name)
+
+        missing_columns = expected_columns - existing_columns
+        if missing_columns:
+            log_warning(f"Missing columns {missing_columns} in table {table_name}")
+            return False
+
+        return True
+
+    except Exception as e:
+        log_error(f"Error validating table schema for {table_name}: {e}")
+        return False
+
+
+def _get_table_columns(conn, table_name: str) -> set[str]:
+    """Helper function to get table columns using sync inspector."""
+    inspector = inspect(conn)
+    columns_info = inspector.get_columns(table_name)
+    return {col["name"] for col in columns_info}
+
+
 # -- Metrics util methods --
 
 
@@ -133,6 +195,39 @@ def bulk_upsert_metrics(session: Session, table: Table, metrics_records: list[di
     return results  # type: ignore
 
 
+async def abulk_upsert_metrics(session: AsyncSession, table: Table, metrics_records: list[dict]) -> list[dict]:
+    """Bulk upsert metrics into the database.
+
+    Args:
+        table (Table): The table to upsert into.
+        metrics_records (list[dict]): The metrics records to upsert.
+
+    Returns:
+        list[dict]: The upserted metrics records.
+    """
+    if not metrics_records:
+        return []
+
+    results = []
+    stmt = sqlite.insert(table)
+
+    # Columns to update in case of conflict
+    update_columns = {
+        col.name: stmt.excluded[col.name]
+        for col in table.columns
+        if col.name not in ["id", "date", "created_at", "aggregation_period"]
+    }
+
+    stmt = stmt.on_conflict_do_update(index_elements=["date", "aggregation_period"], set_=update_columns).returning(  # type: ignore
+        table
+    )
+    result = await session.execute(stmt, metrics_records)
+    results = [dict(row._mapping) for row in result.fetchall()]
+    await session.commit()
+
+    return results  # type: ignore
+
+
 def calculate_date_metrics(date_to_process: date, sessions_data: dict) -> dict:
     """Calculate metrics for the given single date.
 
@@ -173,15 +268,17 @@ def calculate_date_metrics(date_to_process: date, sessions_data: dict) -> dict:
     all_user_ids = set()
 
     for session_type, sessions_count_key, runs_count_key in session_types:
-        sessions = sessions_data.get(session_type, [])
+        sessions = sessions_data.get(session_type, []) or []
         metrics[sessions_count_key] = len(sessions)
 
         for session in sessions:
             if session.get("user_id"):
                 all_user_ids.add(session["user_id"])
-            metrics[runs_count_key] += len(session.get("runs", []))
+
+            # Parse runs from JSON string
             if runs := session.get("runs", []):
-                runs = json.loads(runs)
+                runs = json.loads(runs) if isinstance(runs, str) else runs
+                metrics[runs_count_key] += len(runs)
                 for run in runs:
                     if model_id := run.get("model"):
                         model_provider = run.get("model_provider", "")
@@ -189,14 +286,17 @@ def calculate_date_metrics(date_to_process: date, sessions_data: dict) -> dict:
                             model_counts.get(f"{model_id}:{model_provider}", 0) + 1
                         )
 
-            session_data = json.loads(session.get("session_data", {}))
+            # Parse session_data from JSON string
+            session_data = session.get("session_data", {})
+            if isinstance(session_data, str):
+                session_data = json.loads(session_data)
             session_metrics = session_data.get("session_metrics", {})
             for field in token_metrics:
                 token_metrics[field] += session_metrics.get(field, 0)
 
     model_metrics = []
     for model, count in model_counts.items():
-        model_id, model_provider = model.split(":")
+        model_id, model_provider = model.rsplit(":", 1)
         model_metrics.append({"model_id": model_id, "model_provider": model_provider, "count": count})
 
     metrics["users_count"] = len(all_user_ids)
@@ -266,3 +366,64 @@ def get_dates_to_calculate_metrics_for(starting_date: date) -> list[date]:
     if days_diff <= 0:
         return []
     return [starting_date + timedelta(days=x) for x in range(days_diff)]
+
+
+# -- Cultural Knowledge util methods --
+def serialize_cultural_knowledge_for_db(cultural_knowledge: CulturalKnowledge) -> str:
+    """Serialize a CulturalKnowledge object for database storage.
+
+    Converts the model's separate content, categories, and notes fields
+    into a single JSON string for the database content column.
+    SQLite requires JSON to be stored as strings.
+
+    Args:
+        cultural_knowledge (CulturalKnowledge): The cultural knowledge object to serialize.
+
+    Returns:
+        str: A JSON string containing content, categories, and notes.
+    """
+    content_dict: Dict[str, Any] = {}
+    if cultural_knowledge.content is not None:
+        content_dict["content"] = cultural_knowledge.content
+    if cultural_knowledge.categories is not None:
+        content_dict["categories"] = cultural_knowledge.categories
+    if cultural_knowledge.notes is not None:
+        content_dict["notes"] = cultural_knowledge.notes
+
+    return json.dumps(content_dict) if content_dict else None  # type: ignore
+
+
+def deserialize_cultural_knowledge_from_db(db_row: Dict[str, Any]) -> CulturalKnowledge:
+    """Deserialize a database row to a CulturalKnowledge object.
+
+    The database stores content as a JSON dict containing content, categories, and notes.
+    This method extracts those fields and converts them back to the model format.
+
+    Args:
+        db_row (Dict[str, Any]): The database row as a dictionary.
+
+    Returns:
+        CulturalKnowledge: The cultural knowledge object.
+    """
+    # Extract content, categories, and notes from the JSON content field
+    content_json = db_row.get("content", {}) or {}
+
+    if isinstance(content_json, str):
+        content_json = json.loads(content_json) if content_json else {}
+
+    return CulturalKnowledge.from_dict(
+        {
+            "id": db_row.get("id"),
+            "name": db_row.get("name"),
+            "summary": db_row.get("summary"),
+            "content": content_json.get("content"),
+            "categories": content_json.get("categories"),
+            "notes": content_json.get("notes"),
+            "metadata": db_row.get("metadata"),
+            "input": db_row.get("input"),
+            "created_at": db_row.get("created_at"),
+            "updated_at": db_row.get("updated_at"),
+            "agent_id": db_row.get("agent_id"),
+            "team_id": db_row.get("team_id"),
+        }
+    )

agno/db/surrealdb/__init__.py

@@ -0,0 +1,3 @@
+from agno.db.surrealdb.surrealdb import SurrealDb
+
+__all__ = ["SurrealDb"]

agno/db/surrealdb/metrics.py

@@ -0,0 +1,292 @@
+from datetime import date, datetime, timedelta, timezone
+from textwrap import dedent
+from typing import Any, Callable, Dict, List, Optional, Union
+
+from surrealdb import BlockingHttpSurrealConnection, BlockingWsSurrealConnection, RecordID
+
+from agno.db.base import SessionType
+from agno.db.surrealdb import utils
+from agno.db.surrealdb.models import desurrealize_session, surrealize_dates
+from agno.db.surrealdb.queries import WhereClause
+from agno.utils.log import log_error
+
+
+def get_all_sessions_for_metrics_calculation(
+    client: Union[BlockingWsSurrealConnection, BlockingHttpSurrealConnection],
+    table: str,
+    start_timestamp: Optional[datetime] = None,
+    end_timestamp: Optional[datetime] = None,
+) -> List[Dict[str, Any]]:
+    """
+    Get all sessions of all types (agent, team, workflow) as raw dictionaries.
+
+    Args:
+        start_timestamp (Optional[int]): The start timestamp to filter by. Defaults to None.
+        end_timestamp (Optional[int]): The end timestamp to filter by. Defaults to None.
+
+    Returns:
+        List[Dict[str, Any]]: List of session dictionaries with session_type field.
+
+    Raises:
+        Exception: If an error occurs during retrieval.
+    """
+    where = WhereClause()
+
+    # starting_date
+    if start_timestamp is not None:
+        where = where.and_("created_at", start_timestamp, ">=")
+
+    # ending_date
+    if end_timestamp is not None:
+        where = where.and_("created_at", end_timestamp, "<=")
+
+    where_clause, where_vars = where.build()
+
+    # Query
+    query = dedent(f"""
+        SELECT *
+        FROM {table}
+        {where_clause}
+    """)
+
+    results = utils.query(client, query, where_vars, dict)
+    return [desurrealize_session(x) for x in results]
+
+
+def get_metrics_calculation_starting_date(
+    client: Union[BlockingWsSurrealConnection, BlockingHttpSurrealConnection], table: str, get_sessions: Callable
+) -> Optional[date]:
+    """Get the first date for which metrics calculation is needed:
+
+    1. If there are metrics records, return the date of the first day without a complete metrics record.
+    2. If there are no metrics records, return the date of the first recorded session.
+    3. If there are no metrics records and no sessions records, return None.
+
+    Args:
+        table (Table): The table to get the starting date for.
+
+    Returns:
+        Optional[date]: The starting date for which metrics calculation is needed.
+    """
+    query = dedent(f"""
+        SELECT * FROM ONLY {table}
+        ORDER BY date DESC
+        LIMIT 1
+    """)
+    result = utils.query_one(client, query, {}, dict)
+    if result:
+        # 1. Return the date of the first day without a complete metrics record
+        result_date = result["date"]
+        assert isinstance(result_date, datetime)
+        result_date = result_date.date()
+
+        if result.get("completed"):
+            return result_date + timedelta(days=1)
+        else:
+            return result_date
+
+    # 2. No metrics records. Return the date of the first recorded session
+    first_session, _ = get_sessions(
+        session_type=SessionType.AGENT,  # this is ignored because of component_id=None and deserialize=False
+        sort_by="created_at",
+        sort_order="asc",
+        limit=1,
+        component_id=None,
+        deserialize=False,
+    )
+    assert isinstance(first_session, list)
+
+    first_session_date = first_session[0]["created_at"] if first_session else None
+
+    # 3. No metrics records and no sessions records. Return None
+    if first_session_date is None:
+        return None
+
+    # Handle different types for created_at
+    if isinstance(first_session_date, datetime):
+        return first_session_date.date()
+    elif isinstance(first_session_date, int):
+        # Assume it's a Unix timestamp
+        return datetime.fromtimestamp(first_session_date, tz=timezone.utc).date()
+    elif isinstance(first_session_date, str):
+        # Try parsing as ISO format
+        return datetime.fromisoformat(first_session_date.replace("Z", "+00:00")).date()
+    else:
+        # If it's already a date object
+        if isinstance(first_session_date, date):
+            return first_session_date
+        raise ValueError(f"Unexpected type for created_at: {type(first_session_date)}")
+
+
+def bulk_upsert_metrics(
+    client: Union[BlockingWsSurrealConnection, BlockingHttpSurrealConnection],
+    table: str,
+    metrics_records: List[Dict[str, Any]],
+) -> List[Dict[str, Any]]:
+    """Bulk upsert metrics into the database.
+
+    Args:
+        table (Table): The table to upsert into.
+        metrics_records (List[Dict[str, Any]]): The list of metrics records to upsert.
+
+    Returns:
+        list[dict]: The upserted metrics records.
+    """
+    if not metrics_records:
+        return []
+
+    metrics_records = [surrealize_dates(x) for x in metrics_records]
+
+    try:
+        results = []
+        from agno.utils.log import log_debug
+
+        for metric in metrics_records:
+            log_debug(f"Upserting metric: {metric}")  # Add this
+            result = utils.query_one(
+                client,
+                "UPSERT $record CONTENT $content",
+                {"record": RecordID(table, metric["id"]), "content": metric},
+                dict,
+            )
+            if result:
+                results.append(result)
+        return results
+
+    except Exception as e:
+        import traceback
+
+        log_error(traceback.format_exc())
+        log_error(f"Error upserting metrics: {e}")
+
+    return []
+
+
+def fetch_all_sessions_data(
+    sessions: List[Dict[str, Any]], dates_to_process: list[date], start_timestamp: int
+) -> Optional[dict]:
+    """Return all session data for the given dates, for all session types.
+
+    Args:
+        sessions (List[Dict[str, Any]]): The sessions to process.
+        dates_to_process (list[date]): The dates to fetch session data for.
+        start_timestamp (int): The start timestamp (fallback if created_at is missing).
+
+    Returns:
+        dict: A dictionary with dates as keys and session data as values, for all session types.
+
+    Example:
+    {
+        "2000-01-01": {
+            "agent": [<session1>, <session2>, ...],
+            "team": [...],
+            "workflow": [...],
+        }
+    }
+    """
+    if not dates_to_process:
+        return None
+
+    all_sessions_data: Dict[str, Dict[str, List[Dict[str, Any]]]] = {
+        date_to_process.isoformat(): {"agent": [], "team": [], "workflow": []} for date_to_process in dates_to_process
+    }
+
+    for session in sessions:
+        created_at = session.get("created_at", start_timestamp)
+
+        # Handle different types for created_at
+        if isinstance(created_at, datetime):
+            session_date = created_at.date().isoformat()
+        elif isinstance(created_at, int):
+            session_date = datetime.fromtimestamp(created_at, tz=timezone.utc).date().isoformat()
+        elif isinstance(created_at, date):
+            session_date = created_at.isoformat()
+        else:
+            # Fallback to start_timestamp if type is unexpected
+            session_date = datetime.fromtimestamp(start_timestamp, tz=timezone.utc).date().isoformat()
+
+        if session_date in all_sessions_data:
+            session_type = session.get("session_type", "agent")  # Default to agent if missing
+            all_sessions_data[session_date][session_type].append(session)
+
+    return all_sessions_data
+
+
+def calculate_date_metrics(date_to_process: date, sessions_data: dict) -> dict:
+    """Calculate metrics for the given single date.
+
+    Args:
+        date_to_process (date): The date to calculate metrics for.
+        sessions_data (dict): The sessions data to calculate metrics for.
+
+    Returns:
+        dict: The calculated metrics.
+    """
+    metrics = {
+        "users_count": 0,
+        "agent_sessions_count": 0,
+        "team_sessions_count": 0,
+        "workflow_sessions_count": 0,
+        "agent_runs_count": 0,
+        "team_runs_count": 0,
+        "workflow_runs_count": 0,
+    }
+    token_metrics = {
+        "input_tokens": 0,
+        "output_tokens": 0,
+        "total_tokens": 0,
+        "audio_total_tokens": 0,
+        "audio_input_tokens": 0,
+        "audio_output_tokens": 0,
+        "cache_read_tokens": 0,
+        "cache_write_tokens": 0,
+        "reasoning_tokens": 0,
+    }
+    model_counts: Dict[str, int] = {}
+
+    session_types = [
+        ("agent", "agent_sessions_count", "agent_runs_count"),
+        ("team", "team_sessions_count", "team_runs_count"),
+        ("workflow", "workflow_sessions_count", "workflow_runs_count"),
+    ]
+    all_user_ids = set()
+
+    for session_type, sessions_count_key, runs_count_key in session_types:
+        sessions = sessions_data.get(session_type, [])
+        metrics[sessions_count_key] = len(sessions)
+
+        for session in sessions:
+            if session.get("user_id"):
+                all_user_ids.add(session["user_id"])
+            metrics[runs_count_key] += len(session.get("runs", []))
+            if runs := session.get("runs", []):
+                for run in runs:
+                    if model_id := run.get("model"):
+                        model_provider = run.get("model_provider", "")
+                        model_counts[f"{model_id}:{model_provider}"] = (
+                            model_counts.get(f"{model_id}:{model_provider}", 0) + 1
+                        )
+
+            session_metrics = session.get("session_data", {}).get("session_metrics", {})
+            for field in token_metrics:
+                token_metrics[field] += session_metrics.get(field, 0)
+
+    model_metrics = []
+    for model, count in model_counts.items():
+        model_id, model_provider = model.split(":")
+        model_metrics.append({"model_id": model_id, "model_provider": model_provider, "count": count})
+
+    metrics["users_count"] = len(all_user_ids)
+    current_time = datetime.now(timezone.utc)
+
+    return {
+        "id": date_to_process.isoformat(),  # Changed: Use date as ID (e.g., "2025-10-16")
+        "date": current_time.replace(hour=0, minute=0, second=0, microsecond=0),  # Date at midnight UTC
+        "completed": date_to_process < datetime.now(timezone.utc).date(),
+        "token_metrics": token_metrics,
+        "model_metrics": model_metrics,
+        "created_at": current_time,
+        "updated_at": current_time,
+        "aggregation_period": "daily",
+        **metrics,
+    }