Source Model¶

Stores immutable source content from which facts are extracted.

Class: `FactDb::Models::Source`¶

source = FactDb::Models::Source.new(
  content: "Document content...",
  kind: "email",
  captured_at: Time.current
)

Attributes¶

Attribute	Type	Description
`id`	Integer	Primary key
`content_hash`	String	SHA256 hash for deduplication
`kind`	String	Kind (email, document, etc.)
`content`	Text	Original unmodified text content
`title`	String	Optional title
`source_uri`	String	Original location
`metadata`	Hash	Additional metadata (JSONB)
`embedding`	Vector	Semantic search vector
`captured_at`	DateTime	When content was captured
`created_at`	DateTime	Record creation time

Associations¶

has_many :fact_sources
has_many :facts, through: :fact_sources

Callbacks¶

before_create :compute_hash
before_create :generate_embedding

Instance Methods¶

compute_hash¶

def compute_hash

Computes SHA256 hash of content for deduplication.

generate_embedding¶

def generate_embedding

Generates embedding vector using configured generator.

Class Methods¶

find_or_create_by_text¶

def self.find_or_create_by_text(text, **attributes)

Find existing source by hash or create new.

Example:

source = Source.find_or_create_by_text(
  "Document text",
  kind: "document",
  captured_at: Time.current
)

Scopes¶

by_kind¶

scope :by_kind, ->(kind) { where(kind: kind) }

Filter by content kind.

Source.by_kind('email')

captured_between¶

scope :captured_between, ->(from, to) {
  where(captured_at: from..to)
}

Filter by capture date range.

Source.captured_between(1.week.ago, Time.current)

search_text¶

scope :search_text, ->(query) {
  where("content @@ plainto_tsquery(?)", query)
}

Full-text search.

Source.search_text("quarterly earnings")

Usage Examples¶

Create Source¶

source = Source.create!(
  content: "Important document...",
  kind: "document",
  title: "Q4 Report",
  source_uri: "https://example.com/report.pdf",
  captured_at: Time.current,
  metadata: {
    author: "Jane Smith",
    department: "Finance"
  }
)

Find by Hash¶

hash = Digest::SHA256.hexdigest("Document text")
source = Source.find_by(content_hash: hash)

source.facts.each do |fact|
  puts fact.text
end

Semantic Search¶

# Requires embedding
similar = Source
  .where.not(embedding: nil)
  .order(Arel.sql("embedding <=> '#{query_embedding}'"))
  .limit(10)