adds variable filter sampling rate

.env.example

@@ -23,3 +23,4 @@ OPENAI_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
 # LOG_PROMPTS=false
 # REJECT_DISALLOWED=false
 # REJECT_MESSAGE=This content violates /aicg/'s acceptable use policy.
+# REJECT_SAMPLE_RATE=0.2

src/config.ts

@@ -14,6 +14,8 @@ type Config = {
   maxOutputTokens: number;
   /** Whether requests containing disallowed characters should be rejected. */
   rejectDisallowed?: boolean;
+  /** Rejection sample rate (0 - 1). Higher values are more strict but increase server load. */
+  rejectSampleRate?: number;
   /** Message to return when rejecting requests. */
   rejectMessage?: string;
   /** Logging threshold. */
@@ -29,6 +31,7 @@ export const config: Config = {
   modelRateLimit: getEnvWithDefault("MODEL_RATE_LIMIT", 2),
   maxOutputTokens: getEnvWithDefault("MAX_OUTPUT_TOKENS", 256),
   rejectDisallowed: getEnvWithDefault("REJECT_DISALLOWED", false),
+  rejectSampleRate: getEnvWithDefault("REJECT_SAMPLE_RATE", 0.2),
   rejectMessage: getEnvWithDefault(
     "REJECT_MESSAGE",
     "This content violates /aicg/'s acceptable use policy."

src/proxy/rewriters/language-filter.ts

@@ -9,7 +9,7 @@ const DISALLOWED_REGEX =
 // 15k character request ten times a second. So we'll just sample 20% of the
 // characters and hope that's enough.
 const containsDisallowedCharacters = (text: string) => {
-  const sampleSize = Math.floor(text.length * 0.2);
+  const sampleSize = Math.ceil(text.length * (config.rejectSampleRate || 0.2));
   const sample = text
     .split("")
     .sort(() => 0.5 - Math.random())