Class: ClaudeMemory::Distill::NullDistiller

Inherits:

Distiller

• Object
• Distiller
• ClaudeMemory::Distill::NullDistiller

Defined in:

lib/claude_memory/distill/null_distiller.rb

Constant Summary

DECISION_PATTERNS =

[
  /\b(?:we\s+)?decided\s+to\s+(.+)/i,
  /\b(?:we\s+)?agreed\s+(?:to\s+|on\s+)(.+)/i,
  /\blet'?s\s+(?:go\s+with|use)\s+(.+)/i,
  /\bgoing\s+(?:forward|ahead)\s+with\s+(.+)/i
].freeze

CONVENTION_PATTERNS =

[
  /\balways\s+(.+)/i,
  /\bnever\s+(.+)/i,
  /\bconvention[:\s]+(.+)/i,
  /\bstandard[:\s]+(.+)/i,
  /\bwe\s+use\s+(.+)/i
].freeze

ENTITY_PATTERNS =

{
  "database" => /\b(postgresql|postgres|mysql|sqlite|mongodb|redis)\b/i,
  "framework" => /\b(rails|sinatra|django|express|next\.?js|react|vue)\b/i,
  # `Go` is matched case-sensitively (via the inline (?-i:) flag) so the
  # English verb "go" / "go-to" doesn't masquerade as the language; the
  # other languages stay case-insensitive. `golang` normalizes to `go`.
  "language" => /\b(ruby|python|javascript|typescript|rust|(?-i:Go)|golang)\b/i,
  "platform" => /\b(aws|gcp|azure|heroku|vercel|netlify|docker|kubernetes)\b/i
}.freeze

GLOBAL_SCOPE_PATTERNS =

[
  /\bi\s+always\b/i,
  /\bin\s+all\s+(?:my\s+)?projects\b/i,
  /\beverywhere\b/i,
  /\bacross\s+all\s+(?:my\s+)?(?:projects|repos|codebases)\b/i,
  /\bmy\s+(?:personal\s+)?(?:preference|convention|standard)\b/i,
  /\bglobally\b/i,
  /\buniversally\b/i
].freeze

OBSERVATION_CONVENTION_PATTERNS =

Observation-specific convention patterns: stricter than the shared CONVENTION_PATTERNS. Bare ‘always (.+)` / `never (.+)` / `we use (.+)` match code, prose, and instruction text (“never answer from memory”, “never nil. def …”), so observations require explicit convention framing or a first-person “we always/never”.

[
  /\bconvention[:\s]+(.+)/i,
  /\bstandard[:\s]+(.+)/i,
  /\bwe\s+(?:should\s+)?(?:always|never)\s+(.+)/i
].freeze

NOISE_BODY_SIGNATURE =

Bodies that look like code / JSON / shell / markup / transcript rather than a prose statement. High-precision gate: the Layer-1 observer scrapes raw transcript spans, which on a code-heavy project are dominated by source, specs, docs, and tool output — none of which are observations. (2026-06-23 audit, improvements #74: the prior signature let 38/117 obvious-noise rows through — spec fixtures like ‘kind: “decision”`, CHANGELOG table rows, benchmark tree output, the distiller’s own source comments — and they were being injected into SessionStart.)

Regexp.union(
  /\bdef\s|\bclass\s|\bmodule\s/,                      # Ruby definitions
  /=>|::|","|":\s*"|[{}]|\$\(|&&|\|\|/,                # code / JSON / shell punctuation
  /\w+:\s*["\[{\d]/,                                   # code/JSON key: "value" / key: 1 / key: [
  /\w\(/,                                              # method/function call: expect(, insert_observation(
  /\s\|\s/,                                            # spaced table pipe (doc / CHANGELOG rows)
  /[\u{2500}-\u{257f}]/,                               # box-drawing glyphs (tree / benchmark output)
  /\(vector\)|\(text\)/,                               # benchmark mode labels
  /parentUuid|isSidechain|toolUseID|hookName|"type":/  # raw JSONL transcript fields
)

Instance Method Summary

• #distill(text, content_item_id: nil) ⇒ Object

Instance Method Details

#distill(text, content_item_id: nil) ⇒ Object

# File 'lib/claude_memory/distill/null_distiller.rb', line 71

def distill(text, content_item_id: nil)
  entities = extract_entities(text)
  facts = extract_facts(text, entities)
  decisions = extract_decisions(text)
  signals = extract_signals(text)
  observations = extract_observations(text)

  Extraction.new(
    entities: entities,
    facts: facts,
    decisions: decisions,
    signals: signals,
    observations: observations
  )
end