fix(reading-chat): include memos for memo-comparison questions

[Rasukarusan]

Motivation

• Some user questions that compare or rank memos (e.g. “一番感想が長い本”) failed because the planner returned includeMemo: false or searchText: "感想", which caused the app to omit full memo text or accidentally filter out records.

Description

• Clarify the planner prompt so it explicitly requests includeMemo=true for questions about memo body, length, character count, or content comparisons. (apps/api/src/application/usecases/reading-chat/ask-reading-chat.ts)
• Add applyQuestionIntent which detects memo-related and comparison-related intent from the question and enforces includeMemo: true, and for comparisons sets searchText to undefined and limit to MAX_LIMIT. (applyQuestionIntent in the same file)
• Wire planQuery to call applyQuestionIntent after normalizing the planner result so the server side compensates for planner variation. (planQuery changes)
• Add a regression test that verifies a “longest impression” style question requests memo bodies and up to 100 candidates. (apps/api/src/application/usecases/reading-chat/ask-reading-chat.spec.ts)

Testing

• Ran pnpm --filter api test -- --runInBand application/usecases/reading-chat/ask-reading-chat.spec.ts, and the spec passed: 5 tests, 0 failures.
• Ran ESLint via pnpm --filter api exec eslint src/application/usecases/reading-chat/ask-reading-chat.ts src/application/usecases/reading-chat/ask-reading-chat.spec.ts, which succeeded.
• Ran type check with pnpm --filter api check-types, which failed due to pre-existing Prisma-generated typing errors in unrelated files (ai-summary.ts and discovery.ts); there are no type errors introduced by the changed files.

---

Codex Task

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
kidoku	Ready	Preview	Aug 11, 2026 3:01am