config

type AppConfig struct {
	DefaultUserAgent        string                 `yaml:"default_user_agent"`
	DefaultDelayPerHost     time.Duration          `yaml:"default_delay_per_host"`
	NumWorkers              int                    `yaml:"num_workers"`
	NumImageWorkers         int                    `yaml:"num_image_workers,omitempty"`
	MaxRequests             int                    `yaml:"max_requests"`
	MaxRequestsPerHost      int                    `yaml:"max_requests_per_host"`
	OutputBaseDir           string                 `yaml:"output_base_dir"`
	StateDir                string                 `yaml:"state_dir"`
	MaxRetries              int                    `yaml:"max_retries,omitempty"`
	InitialRetryDelay       time.Duration          `yaml:"initial_retry_delay,omitempty"`
	MaxRetryDelay           time.Duration          `yaml:"max_retry_delay,omitempty"`
	SemaphoreAcquireTimeout time.Duration          `yaml:"semaphore_acquire_timeout,omitempty"`
	GlobalCrawlTimeout      time.Duration          `yaml:"global_crawl_timeout,omitempty"`
	PerPageTimeout          time.Duration          `yaml:"per_page_timeout,omitempty"` // Timeout for processing a single page (0 = no timeout)
	SkipImages              bool                   `yaml:"skip_images,omitempty"`
	MaxImageSizeBytes       int64                  `yaml:"max_image_size_bytes,omitempty"`
	MaxPageSizeBytes        int64                  `yaml:"max_page_size_bytes,omitempty"` // Max HTML page body size in bytes (0 = 50MB default)
	HTTPClientSettings      HTTPClientConfig       `yaml:"http_client_settings,omitempty"`
	Sites                   map[string]*SiteConfig `yaml:"sites"`
	EnableOutputMapping     bool                   `yaml:"enable_output_mapping,omitempty"`
	OutputMappingFilename   string                 `yaml:"output_mapping_filename,omitempty"`
	EnableMetadataYAML      bool                   `yaml:"enable_metadata_yaml,omitempty"`
	MetadataYAMLFilename    string                 `yaml:"metadata_yaml_filename,omitempty"`
	EnableJSONLOutput       bool                   `yaml:"enable_jsonl_output,omitempty"`
	JSONLOutputFilename     string                 `yaml:"jsonl_output_filename,omitempty"`
	EnableTokenCounting     bool                   `yaml:"enable_token_counting,omitempty"`
	TokenizerEncoding       string                 `yaml:"tokenizer_encoding,omitempty"` // e.g., "cl100k_base" (GPT-4 default, approximate for Claude)
	DBGCInterval            time.Duration          `yaml:"db_gc_interval,omitempty"`     // Interval for BadgerDB value log GC (default: 10m)
	EnableIncremental       bool                   `yaml:"enable_incremental,omitempty"` // Enable incremental crawling (skip unchanged pages)
	Chunking                ChunkingConfig         `yaml:"chunking,omitempty"`
}

type ChunkingConfig struct {
	Enabled        bool   `yaml:"enabled,omitempty"`         // Enable chunking output
	MaxChunkSize   int    `yaml:"max_chunk_size,omitempty"`  // Max chunk size in tokens (default: 512)
	ChunkOverlap   int    `yaml:"chunk_overlap,omitempty"`   // Overlap between chunks in tokens (default: 50)
	OutputFilename string `yaml:"output_filename,omitempty"` // Output filename (default: chunks.jsonl)
}

type HTTPClientConfig struct {
	Timeout               time.Duration `yaml:"timeout,omitempty"`                 // Overall request timeout
	MaxIdleConns          int           `yaml:"max_idle_conns,omitempty"`          // Max total idle connections
	MaxIdleConnsPerHost   int           `yaml:"max_idle_conns_per_host,omitempty"` // Max idle connections per host
	IdleConnTimeout       time.Duration `yaml:"idle_conn_timeout,omitempty"`       // Timeout for idle connections
	TLSHandshakeTimeout   time.Duration `yaml:"tls_handshake_timeout,omitempty"`   // Timeout for TLS handshake
	ExpectContinueTimeout time.Duration `yaml:"expect_continue_timeout,omitempty"` // Timeout for 100-continue
	ForceAttemptHTTP2     *bool         `yaml:"force_attempt_http2,omitempty"`     // Explicitly enable/disable HTTP/2 attempt (use pointer for tri-state: nil=default, true=force, false=disable)
	DialerTimeout         time.Duration `yaml:"dialer_timeout,omitempty"`          // Connection dial timeout
	DialerKeepAlive       time.Duration `yaml:"dialer_keep_alive,omitempty"`       // TCP keep-alive interval
}

type ResolvedSiteConfig struct {
	UserAgent              string
	OutputMappingFilename  string
	MetadataYAMLFilename   string
	JSONLOutputFilename    string
	ChunkingOutputFilename string
	DelayPerHost           time.Duration
	MaxPageSizeBytes       int64
	MaxImageSizeBytes      int64
	SkipImages             bool
	EnableOutputMapping    bool
	EnableMetadataYAML     bool
	EnableJSONLOutput      bool
	ChunkingEnabled        bool
	ChunkingMaxSize        int
	ChunkingOverlap        int
}

type SiteChunkingConfig struct {
	Enabled        *bool  `yaml:"enabled,omitempty"`
	MaxChunkSize   *int   `yaml:"max_chunk_size,omitempty"`
	ChunkOverlap   *int   `yaml:"chunk_overlap,omitempty"`
	OutputFilename string `yaml:"output_filename,omitempty"`
}

type SiteConfig struct {
	StartURLs               []string           `yaml:"start_urls"`
	AllowedDomain           string             `yaml:"allowed_domain"`
	AllowedPathPrefix       string             `yaml:"allowed_path_prefix"`
	ContentSelector         string             `yaml:"content_selector"`
	LinkExtractionSelectors []string           `yaml:"link_extraction_selectors,omitempty"`
	DisallowedPathPatterns  []string           `yaml:"disallowed_path_patterns,omitempty"` // Regex patterns for paths to exclude
	RespectNofollow         bool               `yaml:"respect_nofollow,omitempty"`
	UserAgent               string             `yaml:"user_agent,omitempty"`
	DelayPerHost            time.Duration      `yaml:"delay_per_host,omitempty"`
	MaxDepth                int                `yaml:"max_depth"`
	SkipImages              *bool              `yaml:"skip_images,omitempty"`
	MaxImageSizeBytes       *int64             `yaml:"max_image_size_bytes,omitempty"`
	AllowedImageDomains     []string           `yaml:"allowed_image_domains,omitempty"`
	DisallowedImageDomains  []string           `yaml:"disallowed_image_domains,omitempty"`
	EnableOutputMapping     *bool              `yaml:"enable_output_mapping,omitempty"`
	OutputMappingFilename   string             `yaml:"output_mapping_filename,omitempty"`
	EnableMetadataYAML      *bool              `yaml:"enable_metadata_yaml,omitempty"`
	MetadataYAMLFilename    string             `yaml:"metadata_yaml_filename,omitempty"`
	EnableJSONLOutput       *bool              `yaml:"enable_jsonl_output,omitempty"`
	JSONLOutputFilename     string             `yaml:"jsonl_output_filename,omitempty"`
	Chunking                SiteChunkingConfig `yaml:"chunking,omitempty"`
}