moderation_pipeline.cod

module moderation_pipeline

def classify_content
  intent "classify a message as safe, unsafe, or uncertain based on keyword signals"
  sig    (message: String) -> Label
  effects {}
  cand
    intent "unsafe — contains harmful keyword"
    when   or(
             contains(lower(message), "spam"),
             contains(lower(message), "hate"),
             contains(lower(message), "violence")
           )
    belief("unsafe", 0.90)
  cand
    intent "safe — contains approval keyword"
    when   or(
             contains(lower(message), "approved"),
             contains(lower(message), "verified")
           )
    belief("safe", 0.90)
  cand
    intent "uncertain — no keyword match"
    belief("uncertain", 0.75)

def moderate
  intent "gate classification on confidence — refuse when confidence is too low"
  sig    (message: String) -> Label
  effects {}
  cand
    result <- classify_content(message)
    believe result
      ge(conf(result), 0.70) => result
      else                   => bottom

def route_message
  intent "route a message to blocked, approved, escalate-to-human, or refuse"
  sig    (message: String) -> String
  effects {}
  cand
    label <- moderate(message)
    believe label
      eq(label, "unsafe")    => "blocked"
      eq(label, "safe")      => "approved"
      eq(label, "uncertain") => "escalate-to-human"
      else                   => bottom

def run_pipeline
  intent "run the full moderation pipeline and print the routing decision"
  sig    (message: String) -> String
  effects {io.stdout}
  cand
    decision <- route_message(message)
    io.say(decision)

def main
  intent "entry point — run the pipeline with a test message"
  sig    () -> String
  effects {io.stdout}
  cand
    run_pipeline("this content is verified and approved")