remdb 0.2.6__py3-none-any.whl → 0.3.103__py3-none-any.whl

rem/sql/migrations/003_seed_default_user.sql

@@ -0,0 +1,48 @@
+-- ============================================================================
+-- Migration: 003_seed_default_user.sql
+-- Description: Seed the default system user for CLI and API operations
+--
+-- The default user is derived from settings.test.user_email (test@rem.ai)
+-- using deterministic UUID v5 generation. This ensures consistent user ID
+-- across all environments and test runs.
+--
+-- Default user:
+--   email: test@rem.ai
+--   user_id: 9e7dc22b-13bb-5cea-8aee-f6b8e6dc962f (UUID v5 from DNS namespace)
+--
+-- This user is used when:
+--   - CLI commands run without --user-id flag
+--   - API requests come without X-User-Id header
+--   - Tests run without explicit user context
+-- ============================================================================
+
+-- Insert default user (idempotent - skip if exists)
+INSERT INTO users (
+    id,
+    user_id,
+    tenant_id,
+    name,
+    email,
+    role,
+    tags,
+    metadata,
+    created_at,
+    updated_at
+) VALUES (
+    '9e7dc22b-13bb-5cea-8aee-f6b8e6dc962f'::uuid,
+    '9e7dc22b-13bb-5cea-8aee-f6b8e6dc962f',
+    '9e7dc22b-13bb-5cea-8aee-f6b8e6dc962f',
+    'Default User',
+    'test@rem.ai',
+    'system',
+    ARRAY['system', 'default'],
+    '{"description": "Default system user for CLI and API operations without explicit user context"}'::jsonb,
+    NOW(),
+    NOW()
+) ON CONFLICT (id) DO NOTHING;
+
+-- Log migration
+DO $$
+BEGIN
+    RAISE NOTICE 'Seeded default user: test@rem.ai (id: 9e7dc22b-13bb-5cea-8aee-f6b8e6dc962f)';
+END $$;

rem/utils/constants.py

@@ -0,0 +1,97 @@
+"""
+Centralized constants for the REM system.
+
+All magic numbers and commonly-used values should be defined here
+to ensure consistency and make tuning easier.
+"""
+
+# =============================================================================
+# Embedding Model Constants
+# =============================================================================
+
+# OpenAI embedding dimensions by model
+OPENAI_EMBEDDING_DIMS_SMALL = 1536  # text-embedding-3-small
+OPENAI_EMBEDDING_DIMS_LARGE = 3072  # text-embedding-3-large
+OPENAI_EMBEDDING_DIMS_ADA = 1536  # text-embedding-ada-002
+
+# Default embedding dimension (text-embedding-3-small)
+DEFAULT_EMBEDDING_DIMS = 1536
+
+# Voyage AI embedding dimensions
+VOYAGE_EMBEDDING_DIMS = 1024  # voyage-2
+
+# =============================================================================
+# HTTP/API Timeouts (seconds)
+# =============================================================================
+
+HTTP_TIMEOUT_DEFAULT = 30.0  # Standard API calls
+HTTP_TIMEOUT_LONG = 60.0  # Vision/embedding APIs
+HTTP_TIMEOUT_VERY_LONG = 300.0  # Subprocess/batch operations
+
+# Request timeout for httpx AsyncClient
+ASYNC_CLIENT_TIMEOUT = 300.0
+
+# =============================================================================
+# Audio Processing Constants
+# =============================================================================
+
+# Minimum valid WAV file size (header only)
+WAV_HEADER_MIN_BYTES = 44
+
+# OpenAI Whisper API cost per minute (USD)
+WHISPER_COST_PER_MINUTE = 0.006
+
+# Audio chunking parameters
+AUDIO_CHUNK_TARGET_SECONDS = 60.0  # Target chunk duration
+AUDIO_CHUNK_WINDOW_SECONDS = 2.0  # Window for silence detection
+SILENCE_THRESHOLD_DB = -40.0  # Silence detection threshold
+MIN_SILENCE_MS = 500  # Minimum silence duration to split on
+
+# =============================================================================
+# File Processing Constants
+# =============================================================================
+
+# Subprocess timeout for document parsing
+SUBPROCESS_TIMEOUT_SECONDS = 300  # 5 minutes
+
+# Maximum file sizes
+MAX_AUDIO_FILE_SIZE_MB = 25  # Whisper API limit
+
+# =============================================================================
+# Database/Query Constants
+# =============================================================================
+
+# Default batch sizes
+DEFAULT_BATCH_SIZE = 100
+EMBEDDING_BATCH_SIZE = 50
+
+# Default pagination limits
+DEFAULT_PAGE_SIZE = 20
+MAX_PAGE_SIZE = 100
+
+# =============================================================================
+# Rate Limiting
+# =============================================================================
+
+# Default retry settings
+DEFAULT_MAX_RETRIES = 3
+RETRY_BACKOFF_MULTIPLIER = 1
+RETRY_BACKOFF_MIN = 1
+RETRY_BACKOFF_MAX = 60
+
+# =============================================================================
+# S3/Storage Constants
+# =============================================================================
+
+S3_URI_PREFIX = "s3://"
+FILE_URI_PREFIX = "file://"
+
+# =============================================================================
+# LLM Constants
+# =============================================================================
+
+# Default max tokens for vision analysis
+VISION_MAX_TOKENS = 2048
+
+# Default temperature
+DEFAULT_TEMPERATURE = 0.0

rem/utils/date_utils.py

@@ -0,0 +1,228 @@
+"""
+Centralized datetime utilities for consistent UTC-naive datetime handling.
+
+IMPORTANT: REM uses UTC-naive datetimes throughout the codebase.
+PostgreSQL stores TIMESTAMP WITHOUT TIME ZONE, so all Python datetime
+operations should use UTC-naive datetimes to avoid comparison errors.
+
+Convention:
+- All timestamps are implicitly UTC
+- Use utc_now() instead of datetime.utcnow() or datetime.now(timezone.utc)
+- Use parse_iso() to parse ISO format strings (handles "Z" suffix)
+- Use to_iso() to format datetimes as ISO strings
+
+See CLAUDE.md Section 1 (Datetime Convention) for details.
+"""
+
+from datetime import datetime, timedelta
+from typing import Optional
+
+
+def utc_now() -> datetime:
+    """
+    Get current UTC time as a naive datetime.
+
+    Returns:
+        UTC-naive datetime representing current time.
+
+    Example:
+        >>> now = utc_now()
+        >>> now.tzinfo is None
+        True
+    """
+    return datetime.utcnow()
+
+
+def to_iso(dt: datetime) -> str:
+    """
+    Convert datetime to ISO 8601 format string.
+
+    Args:
+        dt: Datetime to format (should be UTC-naive)
+
+    Returns:
+        ISO format string (e.g., "2024-01-15T10:30:00")
+
+    Example:
+        >>> dt = datetime(2024, 1, 15, 10, 30, 0)
+        >>> to_iso(dt)
+        '2024-01-15T10:30:00'
+    """
+    return dt.isoformat()
+
+
+def to_iso_with_z(dt: datetime) -> str:
+    """
+    Convert datetime to ISO 8601 format with Z suffix.
+
+    Use this when interfacing with external APIs that expect
+    the Z suffix to indicate UTC.
+
+    Args:
+        dt: Datetime to format (should be UTC-naive)
+
+    Returns:
+        ISO format string with Z suffix (e.g., "2024-01-15T10:30:00Z")
+    """
+    return dt.isoformat() + "Z"
+
+
+def parse_iso(iso_string: str) -> datetime:
+    """
+    Parse ISO 8601 format string to UTC-naive datetime.
+
+    Handles:
+    - Standard ISO format: "2024-01-15T10:30:00"
+    - Z suffix: "2024-01-15T10:30:00Z"
+    - Timezone offset: "2024-01-15T10:30:00+00:00" (converts to naive)
+    - Microseconds: "2024-01-15T10:30:00.123456"
+
+    Args:
+        iso_string: ISO format datetime string
+
+    Returns:
+        UTC-naive datetime
+
+    Raises:
+        ValueError: If string cannot be parsed
+
+    Example:
+        >>> parse_iso("2024-01-15T10:30:00Z")
+        datetime.datetime(2024, 1, 15, 10, 30)
+        >>> parse_iso("2024-01-15T10:30:00+00:00")
+        datetime.datetime(2024, 1, 15, 10, 30)
+    """
+    # Handle Z suffix (replace with +00:00 for fromisoformat)
+    if iso_string.endswith("Z"):
+        iso_string = iso_string[:-1] + "+00:00"
+
+    # Parse the ISO string
+    dt = datetime.fromisoformat(iso_string)
+
+    # Convert to naive UTC if timezone-aware
+    if dt.tzinfo is not None:
+        # Convert to UTC and strip timezone
+        from datetime import timezone
+        dt = dt.astimezone(timezone.utc).replace(tzinfo=None)
+
+    return dt
+
+
+def parse_iso_safe(iso_string: Optional[str], default: Optional[datetime] = None) -> Optional[datetime]:
+    """
+    Safely parse ISO string, returning default on failure.
+
+    Args:
+        iso_string: ISO format string or None
+        default: Default value if parsing fails
+
+    Returns:
+        Parsed datetime or default value
+    """
+    if not iso_string:
+        return default
+    try:
+        return parse_iso(iso_string)
+    except (ValueError, TypeError):
+        return default
+
+
+def format_timestamp(dt: Optional[datetime] = None) -> str:
+    """
+    Format datetime for display/logging.
+
+    Args:
+        dt: Datetime to format (defaults to current UTC time)
+
+    Returns:
+        Formatted string like "2024-01-15 10:30:00 UTC"
+    """
+    if dt is None:
+        dt = utc_now()
+    return dt.strftime("%Y-%m-%d %H:%M:%S") + " UTC"
+
+
+def format_timestamp_compact(dt: Optional[datetime] = None) -> str:
+    """
+    Format datetime as compact string for filenames/IDs.
+
+    Args:
+        dt: Datetime to format (defaults to current UTC time)
+
+    Returns:
+        Formatted string like "20240115_103000"
+    """
+    if dt is None:
+        dt = utc_now()
+    return dt.strftime("%Y%m%d_%H%M%S")
+
+
+def format_timestamp_for_experiment(dt: Optional[datetime] = None) -> str:
+    """
+    Format datetime for experiment names.
+
+    Args:
+        dt: Datetime to format (defaults to current UTC time)
+
+    Returns:
+        Formatted string like "20240115-103000"
+    """
+    if dt is None:
+        dt = utc_now()
+    return dt.strftime("%Y%m%d-%H%M%S")
+
+
+def days_ago(days: int) -> datetime:
+    """
+    Get datetime N days ago from now.
+
+    Args:
+        days: Number of days ago
+
+    Returns:
+        UTC-naive datetime
+    """
+    return utc_now() - timedelta(days=days)
+
+
+def hours_ago(hours: int) -> datetime:
+    """
+    Get datetime N hours ago from now.
+
+    Args:
+        hours: Number of hours ago
+
+    Returns:
+        UTC-naive datetime
+    """
+    return utc_now() - timedelta(hours=hours)
+
+
+def is_within_hours(dt: datetime, hours: int) -> bool:
+    """
+    Check if datetime is within N hours of now.
+
+    Args:
+        dt: Datetime to check (should be UTC-naive)
+        hours: Number of hours
+
+    Returns:
+        True if dt is within the time window
+    """
+    cutoff = hours_ago(hours)
+    return dt >= cutoff
+
+
+def is_within_days(dt: datetime, days: int) -> bool:
+    """
+    Check if datetime is within N days of now.
+
+    Args:
+        dt: Datetime to check (should be UTC-naive)
+        days: Number of days
+
+    Returns:
+        True if dt is within the time window
+    """
+    cutoff = days_ago(days)
+    return dt >= cutoff

rem/utils/embeddings.py

@@ -20,7 +20,6 @@ Usage:
     embeddings = generate_embeddings("openai:text-embedding-3-small", texts)
 """
 
-import os
 from typing import Any, cast
 
 import requests
@@ -31,6 +30,16 @@ from tenacity import (
     wait_exponential,
 )
 
+from rem.utils.constants import (
+    HTTP_TIMEOUT_LONG,
+    OPENAI_EMBEDDING_DIMS_SMALL,
+    OPENAI_EMBEDDING_DIMS_LARGE,
+    VOYAGE_EMBEDDING_DIMS,
+    RETRY_BACKOFF_MULTIPLIER,
+    RETRY_BACKOFF_MIN,
+    RETRY_BACKOFF_MAX,
+)
+
 
 class EmbeddingError(Exception):
     """Base exception for embedding generation errors."""
@@ -166,7 +175,11 @@ def _create_retry_decorator(max_retries: int):
     return retry(
         retry=retry_if_exception_type(RateLimitError),
         stop=stop_after_attempt(max_retries),
-        wait=wait_exponential(multiplier=1, min=1, max=60),
+        wait=wait_exponential(
+            multiplier=RETRY_BACKOFF_MULTIPLIER,
+            min=RETRY_BACKOFF_MIN,
+            max=RETRY_BACKOFF_MAX,
+        ),
         reraise=True,
     )
 
@@ -234,7 +247,7 @@ def _generate_openai_embeddings(
     }
 
     try:
-        response = requests.post(url, json=payload, headers=headers, timeout=60)
+        response = requests.post(url, json=payload, headers=headers, timeout=HTTP_TIMEOUT_LONG)
 
         # Handle rate limits
         if response.status_code == 429:
@@ -334,7 +347,7 @@ def _generate_voyage_embeddings(
     }
 
     try:
-        response = requests.post(url, json=payload, headers=headers, timeout=60)
+        response = requests.post(url, json=payload, headers=headers, timeout=HTTP_TIMEOUT_LONG)
 
         # Handle rate limits
         if response.status_code == 429:

rem/utils/files.py

@@ -0,0 +1,167 @@
+"""
+File utilities for consistent file handling throughout REM.
+
+Provides context managers and helpers for temporary file operations,
+ensuring proper cleanup and consistent patterns.
+"""
+
+import tempfile
+from contextlib import contextmanager
+from pathlib import Path
+from typing import Generator, Optional
+
+from loguru import logger
+
+
+@contextmanager
+def temp_file_from_bytes(
+    content: bytes,
+    suffix: str = "",
+    prefix: str = "rem_",
+    dir: Optional[str] = None,
+) -> Generator[Path, None, None]:
+    """
+    Create a temporary file from bytes, yield path, cleanup automatically.
+
+    This context manager ensures proper cleanup of temporary files even
+    if an exception occurs during processing.
+
+    Args:
+        content: Bytes to write to the temporary file
+        suffix: File extension (e.g., ".pdf", ".wav")
+        prefix: Prefix for the temp file name
+        dir: Directory for temp file (uses system temp if None)
+
+    Yields:
+        Path to the temporary file
+
+    Example:
+        >>> with temp_file_from_bytes(pdf_bytes, suffix=".pdf") as tmp_path:
+        ...     result = process_pdf(tmp_path)
+        # File is automatically cleaned up after the block
+
+    Note:
+        The file is created with delete=False so we control cleanup.
+        This allows the file to be read by external processes.
+    """
+    tmp_path: Optional[Path] = None
+    try:
+        with tempfile.NamedTemporaryFile(
+            suffix=suffix,
+            prefix=prefix,
+            dir=dir,
+            delete=False,
+        ) as tmp:
+            tmp.write(content)
+            tmp_path = Path(tmp.name)
+
+        yield tmp_path
+
+    finally:
+        if tmp_path is not None:
+            try:
+                tmp_path.unlink(missing_ok=True)
+            except Exception as e:
+                logger.warning(f"Failed to cleanup temp file {tmp_path}: {e}")
+
+
+@contextmanager
+def temp_file_empty(
+    suffix: str = "",
+    prefix: str = "rem_",
+    dir: Optional[str] = None,
+) -> Generator[Path, None, None]:
+    """
+    Create an empty temporary file, yield path, cleanup automatically.
+
+    Useful when you need to write to a file after creation or when
+    an external process will write to the file.
+
+    Args:
+        suffix: File extension
+        prefix: Prefix for the temp file name
+        dir: Directory for temp file
+
+    Yields:
+        Path to the empty temporary file
+    """
+    tmp_path: Optional[Path] = None
+    try:
+        with tempfile.NamedTemporaryFile(
+            suffix=suffix,
+            prefix=prefix,
+            dir=dir,
+            delete=False,
+        ) as tmp:
+            tmp_path = Path(tmp.name)
+
+        yield tmp_path
+
+    finally:
+        if tmp_path is not None:
+            try:
+                tmp_path.unlink(missing_ok=True)
+            except Exception as e:
+                logger.warning(f"Failed to cleanup temp file {tmp_path}: {e}")
+
+
+@contextmanager
+def temp_directory(
+    prefix: str = "rem_",
+    dir: Optional[str] = None,
+) -> Generator[Path, None, None]:
+    """
+    Create a temporary directory, yield path, cleanup automatically.
+
+    Args:
+        prefix: Prefix for the temp directory name
+        dir: Parent directory for temp directory
+
+    Yields:
+        Path to the temporary directory
+    """
+    import shutil
+
+    tmp_dir: Optional[Path] = None
+    try:
+        tmp_dir = Path(tempfile.mkdtemp(prefix=prefix, dir=dir))
+        yield tmp_dir
+
+    finally:
+        if tmp_dir is not None:
+            try:
+                shutil.rmtree(tmp_dir, ignore_errors=True)
+            except Exception as e:
+                logger.warning(f"Failed to cleanup temp directory {tmp_dir}: {e}")
+
+
+def ensure_parent_exists(path: Path) -> Path:
+    """
+    Ensure parent directory exists, creating if necessary.
+
+    Args:
+        path: File path whose parent should exist
+
+    Returns:
+        The original path (for chaining)
+    """
+    path.parent.mkdir(parents=True, exist_ok=True)
+    return path
+
+
+def safe_delete(path: Path) -> bool:
+    """
+    Safely delete a file, returning success status.
+
+    Args:
+        path: Path to delete
+
+    Returns:
+        True if deleted or didn't exist, False on error
+    """
+    try:
+        path.unlink(missing_ok=True)
+        return True
+    except Exception as e:
+        logger.warning(f"Failed to delete {path}: {e}")
+        return False