config.yaml.example

# Agentic Memorizer Configuration
# Copy this file to ~/.config/memorizer/config.yaml and customize as needed.
# All values shown are defaults unless otherwise noted.
# Environment variables can override any setting using the MEMORIZER_ prefix.
# Example: MEMORIZER_DAEMON_HTTP_PORT=9000 overrides daemon.http_port

# ------------------------------------------------------------------------------
# Logging Configuration
# ------------------------------------------------------------------------------

# Log level controls verbosity of log output.
# Valid values: debug, info, warn, error
log_level: info

# Path to the log file. Supports ~ for home directory expansion.
log_file: ~/.config/memorizer/memorizer.log

# ------------------------------------------------------------------------------
# Daemon Configuration
# ------------------------------------------------------------------------------
# Settings for the background daemon process that monitors directories,
# processes files, and maintains the knowledge graph.

daemon:
  # Port for the HTTP health check server.
  # The daemon exposes /health and /status endpoints on this port.
  http_port: 7600

  # Bind address for the HTTP server.
  # Use 127.0.0.1 for localhost-only access, 0.0.0.0 for all interfaces.
  http_bind: 127.0.0.1

  # Graceful shutdown timeout in seconds.
  # The daemon waits this long for in-flight operations to complete before
  # forcefully terminating.
  shutdown_timeout: 30

  # Path to the PID file. Used to detect if daemon is already running
  # and for service management commands (stop, status).
  pid_file: ~/.config/memorizer/daemon.pid

  # Interval in seconds for periodic knowledge graph rebuilds.
  # Set to 0 to disable periodic rebuilds.
  rebuild_interval: 3600

  # Metrics collection settings
  metrics:
    # Interval in seconds between metrics collection cycles.
    # Metrics include component health, queue depths, and processing stats.
    collection_interval: 15

  # Event bus settings
  event_bus:
    # Buffer size for each subscriber's event channel.
    # Larger values reduce drops but use more memory per subscriber.
    buffer_size: 100

    # Maximum number of critical events retained in the queue.
    # Critical events are persisted to the consolidated storage database.
    critical_queue_capacity: 1000

# ------------------------------------------------------------------------------
# Storage Configuration
# ------------------------------------------------------------------------------
# Settings for the consolidated SQLite database that stores remembered paths,
# file state, critical events, and the persistence queue.

storage:
  # Path to the consolidated SQLite database file.
  # Supports ~ for home directory expansion.
  database_path: ~/.config/memorizer/memorizer.db

# ------------------------------------------------------------------------------
# Persistence Queue Configuration
# ------------------------------------------------------------------------------
# Settings for the durable persistence queue that stores analysis results
# when the graph database is unavailable. Results are automatically persisted
# to the graph when connectivity is restored.

persistence_queue:
  # Maximum number of persistence attempts before marking an item as failed.
  # Failed items are retained for inspection but not automatically retried.
  max_retries: 3

  # Base delay in milliseconds between retry attempts.
  # Actual delay may include exponential backoff.
  retry_backoff_ms: 1000

  # Number of items to process in each drain batch.
  # Larger values improve throughput but use more memory.
  drain_batch_size: 10

  # How long to keep completed items (in minutes) before purging.
  # Completed items are purged after each drain cycle.
  completed_retention_min: 60

  # How long to keep failed items (in days) before purging.
  # Failed items are retained longer for debugging purposes.
  failed_retention_days: 7

# ------------------------------------------------------------------------------
# Graph Database Configuration
# ------------------------------------------------------------------------------
# Connection settings for FalkorDB (Redis Graph), which stores the
# knowledge graph containing files, chunks, metadata, and relationships.

graph:
  # FalkorDB server hostname or IP address.
  host: localhost

  # FalkorDB server port.
  port: 6379

  # Name of the graph within FalkorDB.
  # Multiple memorizer instances can use different graph names on the same server.
  name: memorizer

  # Environment variable name containing the FalkorDB password.
  # Leave the env var unset for passwordless connections.
  # The password is read from this env var at runtime, not stored in config.
  password_env: MEMORIZER_GRAPH_PASSWORD

  # Maximum retry attempts for failed graph operations.
  max_retries: 3

  # Delay in milliseconds between retry attempts.
  retry_delay_ms: 1000

  # Size of the write queue for batching graph updates.
  # Larger values improve throughput but use more memory.
  write_queue_size: 1000

# ------------------------------------------------------------------------------
# Semantic Analysis Provider Configuration
# ------------------------------------------------------------------------------
# Settings for the AI provider used to analyze file content, extract topics,
# identify entities, and generate summaries.

semantic:
  # Enable or disable semantic analysis.
  # When disabled, only metadata and embeddings (if enabled) are generated.
  # Env override: MEMORIZER_SEMANTIC_ENABLED
  enabled: true

  # Semantic analysis provider name.
  # Valid values: anthropic, openai, google
  provider: anthropic

  # Model identifier for the semantic provider.
  # Default varies by provider:
  #   anthropic: claude-sonnet-4-5-20250929, claude-opus-4-5-20251101, claude-haiku-4-5-20251015
  #   openai: gpt-5.2, gpt-5.2-pro, gpt-5-mini
  #   google: gemini-3-pro-preview, gemini-3-flash-preview, gemini-2.5-flash
  model: claude-sonnet-4-5-20250929

  # Rate limit in requests per minute.
  # Prevents exceeding provider API limits. The daemon uses token bucket
  # rate limiting with burst capacity equal to this value.
  rate_limit: 10

  # API key for the semantic provider.
  # Prefer using api_key_env instead of storing keys in config files.
  # If both are set, api_key takes precedence.
  # api_key: sk-...

  # Environment variable name containing the API key.
  # Default varies by provider:
  #   anthropic: ANTHROPIC_API_KEY
  #   openai: OPENAI_API_KEY
  #   google: GOOGLE_API_KEY
  api_key_env: ANTHROPIC_API_KEY

# ------------------------------------------------------------------------------
# Embeddings Provider Configuration
# ------------------------------------------------------------------------------
# Settings for the AI provider used to generate vector embeddings for
# semantic similarity search across chunks.

embeddings:
  # Enable or disable embeddings generation.
  # When disabled, semantic similarity search will not be available.
  enabled: true

  # Embeddings provider name.
  # Valid values: openai, voyage, google
  provider: openai

  # Model identifier for the embeddings provider.
  # Default varies by provider:
  #   openai: text-embedding-3-large, text-embedding-3-small
  #   voyage: voyage-3-large, voyage-3.5, voyage-code-3
  #   google: gemini-embedding-001, text-embedding-004
  model: text-embedding-3-large

  # Vector dimensions for the embeddings model.
  # Must match the model's output dimensions:
  #   openai text-embedding-3-large: 3072, text-embedding-3-small: 1536
  #   voyage voyage-3-large/voyage-3.5/voyage-code-3: 1024
  #   google gemini-embedding-001: 3072, text-embedding-004: 768
  dimensions: 3072

  # API key for the embeddings provider.
  # Prefer using api_key_env instead of storing keys in config files.
  # If both are set, api_key takes precedence.
  # api_key: sk-...

  # Environment variable name containing the API key.
  # Default varies by provider:
  #   openai: OPENAI_API_KEY
  #   voyage: VOYAGE_API_KEY
  #   google: GOOGLE_API_KEY
  api_key_env: OPENAI_API_KEY

# ------------------------------------------------------------------------------
# Default Skip/Include Patterns
# ------------------------------------------------------------------------------
# These patterns are applied by default when remembering new directories.
# They can be overridden per-directory using flags on the remember command.
# Include patterns override corresponding skip patterns for the same items.

defaults:
  skip:
    # File extensions to skip (without content analysis).
    # These files are not chunked, analyzed, or added to the knowledge graph.
    extensions: [".exe", ".dll", ".so", ".dylib", ".bin", ".o", ".a", ".lib", ".obj", ".pyc", ".pyo", ".class", ".zip", ".tar", ".gz", ".tgz", ".rar", ".7z", ".jar", ".war", ".map", ".tmp", ".temp", ".bak", ".swp", ".swo", ".log"]

    # Directory names to skip entirely (not traversed).
    # The daemon will not descend into directories with these names.
    directories: [".git", ".svn", ".hg", "node_modules", "bower_components", "__pycache__", ".pytest_cache", ".mypy_cache", ".tox", "venv", ".venv", "dist", "build", "target", ".idea", ".vscode", "coverage", ".nyc_output", "htmlcov"]

    # Specific file names or glob patterns to skip.
    # Supports glob patterns like *.min.js for pattern matching.
    files: [".DS_Store", "Thumbs.db", "desktop.ini", "package-lock.json", "yarn.lock", "pnpm-lock.yaml", "Cargo.lock", "go.sum", "Gemfile.lock", "poetry.lock", "composer.lock", "*.min.js", "*.min.css", "*.bundle.js", "4913", "#*", "*~"]

    # Skip hidden files and directories (names starting with .).
    # When true, dotfiles and dotfolders are ignored unless explicitly included.
    hidden: true

  include:
    # File extensions to explicitly include (overrides skip.extensions).
    # Use this to include specific extensions that would otherwise be skipped.
    extensions: []

    # Directory names to explicitly include (overrides skip.directories).
    # Use this to include specific directories that would otherwise be skipped.
    directories: []

    # Specific file names to explicitly include (overrides skip.files).
    # Use this to include specific files that would otherwise be skipped.
    files: []