feat(pattern-disruptor): add prompt disruption module by vi70x4 · Pull Request #61 · animaios/anima

vi70x4 · 2026-06-27T23:16:17Z

Summary

add the offline @proj-airi/pattern-disruptor package with EN/RU word banks, random/contextual/double-pass word injection, and synonym freshness scanning
persist per-card pattern disruptor settings and compose its prompt supplement with existing toolset guidance
add AIRI card editor controls and English i18n entries

Tests

pnpm -F @proj-airi/pattern-disruptor test:run
pnpm -F @proj-airi/pattern-disruptor typecheck
pnpm -F @proj-airi/pattern-disruptor build
pnpm -F @proj-airi/stage-ui exec vitest run src/stores/modules/airi-card.test.ts src/stores/chat.contract.test.ts
pnpm -F @proj-airi/stage-pages typecheck

Known Verification Gap

pnpm -F @proj-airi/stage-ui typecheck currently fails on unrelated test typing errors already present on the origin/main base (for example tts-session.test.ts and character/orchestrator/index.test.ts).

Ready for Gemini review.

Add the offline pattern disruptor package, per-card settings, chat prompt integration, and card editor controls. Tests: pnpm -F @proj-airi/pattern-disruptor test:run; pnpm -F @proj-airi/pattern-disruptor typecheck; pnpm -F @proj-airi/pattern-disruptor build; pnpm -F @proj-airi/stage-ui exec vitest run src/stores/modules/airi-card.test.ts src/stores/chat.contract.test.ts; pnpm -F @proj-airi/stage-pages typecheck. Known verification gap: pnpm -F @proj-airi/stage-ui typecheck fails on unrelated test typing errors already present on the origin/main base.

deepsource-io · 2026-06-27T23:16:51Z

DeepSource Code Review

We reviewed changes in f0ebbad...f32dcb1 on this pull request. Below is the summary for the review, and you can see the individual issues we found as inline review comments.

See full review on DeepSource ↗

Important

Some issues found as part of this review are outside of the diff in this pull request and aren't shown in the inline review comments due to GitHub's API limitations. You can see those issues on the DeepSource dashboard.

PR Report Card

Overall Grade	Security Reliability Complexity Hygiene

Code Review Summary

Analyzer	Updated (UTC)	Details
JavaScript	Jun 27, 2026 11:31p.m.	Review ↗
Shell	Jun 27, 2026 11:31p.m.	Review ↗
C & C++	Jun 27, 2026 11:31p.m.	Review ↗

Important

AI Review is run only on demand for your team. We're only showing results of static analysis review right now. To trigger AI Review, comment @deepsourcebot review on this thread.

deepsource-io · 2026-06-27T23:17:58Z

+  ResolvedPatternDisruptorLanguage,
+} from '../types'
+
+const CYRILLIC_RE = /[\u0400-\u04FF]/


Use the 'u' flag with regular expressions

It is recommended to use the u flag with regular expressions.

deepsource-io · 2026-06-27T23:17:58Z

+  return text.match(WORD_RE)?.map((token) => token.toLowerCase()) ?? []
+}
+
+export function normalizeToken(token: string, language: ResolvedPatternDisruptorLanguage): string {


`normalizeToken` has a cyclomatic complexity of 15 with "medium" risk

A function with high cyclomatic complexity can be hard to understand and
maintain. Cyclomatic complexity is a software metric that measures the number of
independent paths through a function. A higher cyclomatic complexity indicates
that the function has more decision points and is more complex.

deepsource-io · 2026-06-27T23:17:58Z

+  const blacklist = normalizeWordSet(input.settings.randomWords.blacklist)
+  const exactLength = input.settings.randomWords.wordLength
+
+  return input.bank.filter(([word, partOfSpeech]) => {


Function has a cyclomatic complexity of 7 with "medium" risk

A function with high cyclomatic complexity can be hard to understand and
maintain. Cyclomatic complexity is a software metric that measures the number of
independent paths through a function. A higher cyclomatic complexity indicates
that the function has more decision points and is more complex.

deepsource-io · 2026-06-27T23:17:58Z

+    const lower = word.toLowerCase()
+    if (blacklist.has(lower) || input.history.has(lower)) return false
+    if (partsOfSpeech.size > 0 && !partsOfSpeech.has(partOfSpeech)) return false
+    if (exactLength > 0 && word.length !== exactLength) return false


Boolean return can be simplified

The following pattern:

deepsource-io · 2026-06-27T23:17:58Z

+  })
+}
+
+function uniqueWords(words: string[], count: number, history: Set<string>, settings: PatternDisruptorSettings) {


`uniqueWords` has a cyclomatic complexity of 7 with "medium" risk

A function with high cyclomatic complexity can be hard to understand and
maintain. Cyclomatic complexity is a software metric that measures the number of
independent paths through a function. A higher cyclomatic complexity indicates
that the function has more decision points and is more complex.

deepsource-io · 2026-06-27T23:17:58Z

+  return uniqueWords([...associations, ...fallback], input.count, input.history, input.settings)
+}
+
+export function generateRandomWords(input: GenerateRandomWordsInput): string[] {


`generateRandomWords` has a cyclomatic complexity of 6 with "medium" risk

A function with high cyclomatic complexity can be hard to understand and
maintain. Cyclomatic complexity is a software metric that measures the number of
independent paths through a function. A higher cyclomatic complexity indicates
that the function has more decision points and is more complex.

deepsource-io · 2026-06-27T23:17:58Z

+  count: number
+}
+
+function collectTokenStats(messages: string[], language: ResolvedPatternDisruptorLanguage): TokenStats[] {


`collectTokenStats` has a cyclomatic complexity of 6 with "medium" risk

A function with high cyclomatic complexity can be hard to understand and
maintain. Cyclomatic complexity is a software metric that measures the number of
independent paths through a function. A higher cyclomatic complexity indicates
that the function has more decision points and is more complex.

deepsource-io · 2026-06-27T23:17:58Z

 import { useI18n } from 'vue-i18n'

 import CardCreationTabArtistry from './tabs/CardCreationTabArtistry.vue'
+import CardCreationTabPatternDisruptor from './tabs/CardCreationTabPatternDisruptor.vue'


No default export found in imported module "./tabs/CardCreationTabPatternDisruptor.vue"

It is recommended to use default imports only if there is a default export in the module being imported. If there are no default exports in the source module, it is better to use named imports, and it helps the build tools while optimizing the code.

gemini-code-assist

Code Review

This pull request introduces the @proj-airi/pattern-disruptor package, an offline prompt disruption engine designed to vary AI character responses through random word injection and synonym overuse tracking, alongside its integration into the settings UI and chat orchestrator. The review feedback highlights several critical areas for improvement: the template renderer should allow optional whitespace inside double braces; the word length filters in the random words generator and synonym scanner should be adjusted to support 3-letter Russian words like 'сад'; basic Russian inflection handling is needed to prevent dictionary matching failures; and the pattern disruptor state should be reset when switching sessions or characters to avoid word history leakage.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

gemini-code-assist · 2026-06-27T23:18:07Z

+function renderTemplate(template: string, values: Record<string, string | number>): string {
+  return template.replace(/\{\{(\w+)\}\}/g, (_, key: string) => String(values[key] ?? ''))
+}


The template renderer uses a strict regular expression /\{\{(\w+)\}\}/g which does not allow any whitespace inside the double braces. If a user customizes their prompt and writes {{ words }} instead of {{words}}, the replacement will silently fail. Adding optional whitespace handling makes the template rendering much more robust.

Suggested change

function renderTemplate(template: string, values: Record<string, string | number>): string {

return template.replace(/\{\{(\w+)\}\}/g, (_, key: string) => String(values[key] ?? ''))

}

function renderTemplate(template: string, values: Record<string, string | number>): string {

return template.replace(/\{\{\s*(\w+)\s*\}\}/g, (_, key: string) => String(values[key] ?? ''))

}

gemini-code-assist · 2026-06-27T23:18:07Z

+}): string[] {
+  const keywords = tokenizeWords(input.userMessage)
+    .map((token) => normalizeToken(token, input.language))
+    .filter((token) => token.length >= 4 && !isStopword(token, input.language) && Boolean(input.synonyms[token]))


The filter enforces token.length >= 4, which prevents valid 3-letter words like сад (garden) from being matched as contextual anchors. Since сад is a key in ru/synonyms.json, it will be completely ignored in contextual mode. Since we already check Boolean(input.synonyms[token]), the length check is redundant and can be safely removed.

Suggested change

.filter((token) => token.length >= 4 && !isStopword(token, input.language) && Boolean(input.synonyms[token]))

.filter((token) => !isStopword(token, input.language) && Boolean(input.synonyms[token]))

gemini-code-assist · 2026-06-27T23:18:08Z

+  for (const message of messages) {
+    for (const token of tokenizeWords(message)) {
+      const normalizedWord = normalizeToken(token, language)
+      if (normalizedWord.length < 4 || isStopword(normalizedWord, language)) continue


The scanner enforces normalizedWord.length < 4, which prevents 3-letter words like сад (garden) from being scanned for overuse. Lowering this limit to 3 ensures that all valid words in the Russian vocabulary are correctly tracked.

Suggested change

if (normalizedWord.length < 4 || isStopword(normalizedWord, language)) continue

if (normalizedWord.length < 3 || isStopword(normalizedWord, language)) continue

gemini-code-assist · 2026-06-27T23:18:08Z

+
+export function normalizeToken(token: string, language: ResolvedPatternDisruptorLanguage): string {
+  const lower = token.toLowerCase().replace(/^[^\p{L}]+|[^\p{L}]+$/gu, '')
+  if (language === 'ru') return lower


Since Russian is a highly inflected language, returning the token as-is (if (language === 'ru') return lower) means that inflected forms (e.g., 'саду', 'яркая', 'историю') will fail to match the dictionary keys in synonyms.json (like 'сад', 'яркий', 'история'). This significantly reduces the effectiveness of the contextual mode and synonym scanner for Russian conversations. Consider implementing a basic suffix-stripper or mapping for the small Russian vocabulary to ensure effective matching.

gemini-code-assist · 2026-06-27T23:18:08Z

+      patternDisruptorStore.prepareForUserTurn({
+        settings: cardStore.activeCard?.extensions?.airi?.modules?.patternDisruptor,
+        messageText,
+        sessionMessages,
+      })


The patternDisruptorStore.resetState() function is defined but never called when switching sessions or characters. This can lead to word history leakage, where recently injected words from a previous character or session are carried over and affect the random word selection of a new session. Consider calling patternDisruptorStore.resetState() when the active session or active card changes.

Tests: pnpm -F @proj-airi/pattern-disruptor test:run; pnpm -F @proj-airi/pattern-disruptor typecheck; pnpm -F @proj-airi/pattern-disruptor build; pnpm -F @proj-airi/stage-ui exec vitest run src/stores/chat.contract.test.ts src/stores/modules/airi-card.test.ts; pnpm -F @proj-airi/stage-pages typecheck. Known: pnpm -F @proj-airi/stage-ui typecheck still fails in unrelated existing test files (tts-session, character orchestrator, characters, speech, provider catalog).

deepsource-io · 2026-06-27T23:33:07Z

+  return isDroppableEnglishPlural(withoutContraction) ? withoutContraction.slice(0, -1) : withoutContraction
+}
+
+function getKnownRussianSuffixCandidate(token: string): string | undefined {


Expected to return a value at the end of function 'getKnownRussianSuffixCandidate'

Any code paths that do not have explicit returns will return undefined. It is recommended to replace any implicit dead-ends that return undefined with a return null statement.

deepsource-io · 2026-06-27T23:33:07Z

+  return new Set(Array.from(words ?? []).map((word) => word.toLowerCase()))
+}
+
+function isAllowedWordBankEntry(input: {


`isAllowedWordBankEntry` has a cyclomatic complexity of 6 with "medium" risk

A function with high cyclomatic complexity can be hard to understand and
maintain. Cyclomatic complexity is a software metric that measures the number of
independent paths through a function. A higher cyclomatic complexity indicates
that the function has more decision points and is more complex.

deepsource-io Bot reviewed Jun 27, 2026

View reviewed changes

gemini-code-assist Bot reviewed Jun 27, 2026

View reviewed changes

vi70x4 force-pushed the vi/feat/pattern-disruptor branch from 692d355 to f32dcb1 Compare June 27, 2026 23:31

deepsource-io Bot reviewed Jun 27, 2026

View reviewed changes

vi70x4 merged commit d888c0f into main Jun 28, 2026
6 of 9 checks passed

vi70x4 deleted the vi/feat/pattern-disruptor branch June 28, 2026 00:11

	.filter((token) => token.length >= 4 && !isStopword(token, input.language) && Boolean(input.synonyms[token]))
	.filter((token) => !isStopword(token, input.language) && Boolean(input.synonyms[token]))

	if (normalizedWord.length < 4 \|\| isStopword(normalizedWord, language)) continue
	if (normalizedWord.length < 3 \|\| isStopword(normalizedWord, language)) continue

Uh oh!

Conversation

vi70x4 commented Jun 27, 2026

Summary

Tests

Known Verification Gap

Uh oh!

deepsource-io Bot commented Jun 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

DeepSource Code Review

PR Report Card

Code Review Summary

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

Use the 'u' flag with regular expressions

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

`normalizeToken` has a cyclomatic complexity of 15 with "medium" risk

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

Function has a cyclomatic complexity of 7 with "medium" risk

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

Boolean return can be simplified

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

`uniqueWords` has a cyclomatic complexity of 7 with "medium" risk

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

`generateRandomWords` has a cyclomatic complexity of 6 with "medium" risk

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

`collectTokenStats` has a cyclomatic complexity of 6 with "medium" risk

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

No default export found in imported module "./tabs/CardCreationTabPatternDisruptor.vue"

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Jun 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 27, 2026

Choose a reason for hiding this comment

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

Expected to return a value at the end of function 'getKnownRussianSuffixCandidate'

Uh oh!

deepsource-io Bot Jun 27, 2026

Choose a reason for hiding this comment

`isAllowedWordBankEntry` has a cyclomatic complexity of 6 with "medium" risk

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

deepsource-io Bot commented Jun 27, 2026 •

edited

Loading