feat(agent-core): feed AskUserQuestion answers back as question text and option labels by RealKai42 · Pull Request #1414 · MoonshotAI/kimi-code

RealKai42 · 2026-07-06T06:48:13Z

Problem

The flattened answers record fed back to the model was keyed by synthesized ids:

{"answers":{"q_0":"opt_0_1"}}

To understand what the user picked, the model had to map positional ids back to the original tool-call arguments across messages — unreadable in transcripts, and an observed badcase where the model misread the user's choice and went down the wrong path.

Change

After this PR the tool result is self-explanatory (all client paths now produce the same shape):

{"answers":{"Which database?":"Postgres"}}

toAgentCoreResponse translates ids back to text: it now receives the original broker request and maps q_<i> → question text, opt_<q>_<o> → option label (comma-joined for multi-select; Other stays free text). Unknown ids are kept verbatim for diagnosability; a missing request falls back to the old raw-id behavior.
Wire protocol unchanged: clients still answer with structured option ids over REST/WS; the resolve route reads the pending request (before settling drops it) and passes it to the translator. packages/protocol is untouched.
Uniqueness constraints (the precondition for text keys/values): question texts must be unique per call, option labels unique within each question. Enforced in the tool execution path — runtime arg validation is AJV against the JSON Schema, where a zod refine is unrepresentable — and mirrored on the exported zod schemas. Violations return an instructive error so the model can rephrase and retry.
Web transcript card resolves the new label form (whole-string match first, so labels containing commas resolve for single-select) and keeps the legacy opt_<q>_<o> decode so old session transcripts still render.
TUI and ACP paths already produced the text form; they are now the uniform shape rather than one of three inconsistent formats.

Tests

adapter: single/multi/other/multi_with_other/skipped, unknown-id passthrough, no-request fallback
tool: uniqueness rejection on both foreground and background paths before any UI interaction; cross-question duplicate labels stay allowed
pinned contract: AJV accepts duplicates (enforcement is in the tool), JSON Schema generation unaffected by the refine
server e2e: wire ids in → text record out
web: dual-form (label + legacy id) resolution

Full suite: 9253 passed; the 6 failures in test/cli/update/preflight.test.ts (rollout gating) reproduce on a clean checkout and are unrelated (time-dependent baseline).

…and option labels The flattened answers record the model receives was keyed by synthesized ids (q_0 / opt_0_1), forcing a cross-message positional lookup against the original tool call to understand what the user picked — both unreadable in transcripts and a real model-misreads-the-choice badcase. - toAgentCoreResponse now takes the original broker request and translates wire ids back to question text (keys) and option labels (values); unknown ids are kept verbatim, missing request falls back to raw ids - wire protocol unchanged: clients still answer with option ids; the resolve route reads the pending request before settling it - question texts must be unique per call and option labels unique per question, enforced in the tool execution path (AJV cannot express the zod refine) and mirrored on the exported schemas - web transcript card resolves both the new label form and legacy id transcripts; TUI and ACP paths already produced the text form

changeset-bot · 2026-07-06T06:48:23Z

🦋 Changeset detected

Latest commit: 10fa5be

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@moonshot-ai/kimi-code	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

pkg-pr-new · 2026-07-06T06:50:28Z

pnpm dlx https://pkg.pr.new/@moonshot-ai/kimi-code@10fa5be

npx https://pkg.pr.new/@moonshot-ai/kimi-code@10fa5be

commit: 10fa5be

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 12df36f9b1

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-06T06:53:02Z

        break;
      case 'multi':
-        flattened[qid] = ans.option_ids.join(',');
+        flattened[key] = ans.option_ids.map(optionText).join(',');


Reject comma-bearing labels for multi-select answers

When an AskUserQuestion is multi_select and any selected option label contains a comma, this comma-join makes the model-facing answer ambiguous; e.g. selecting Fast, risky and Safe emits Fast, risky,Safe, and the web transcript resolver splits on commas so it displays Fast, risky as Other text rather than the selected option. The new uniqueness check still allows such labels, so either reject/escape comma-bearing labels for multi-select and multi_with_other, or use an unambiguous value shape.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-07-06T06:53:02Z

+
  const flattened: InProcessQuestionAnswers = {};
  for (const [qid, ans] of Object.entries(resp.answers)) {
+    const key = itemsById.get(qid)?.question ?? qid;


Enforce unique question text at the broker boundary

When requestQuestion reaches the broker from anything other than AskUserQuestionTool, the QuestionRequest/IQuestionService boundary still accepts duplicate question strings, and both qids then map to the same flattened key here. Resolving q_0 and q_1 for two same-text questions silently overwrites one answer, so the uniqueness contract needs to be enforced in the broker/adapter path as well or this translation needs a disambiguating fallback for duplicates.

Useful? React with 👍 / 👎.

…rden question schema - Join multi-select labels with ', ' in the server translator, matching what the TUI reverse-RPC path already emits, so the model sees one format regardless of which client answered - Trim segments in the web transcript resolver before label matching: TUI-answered multi-select transcripts (', '-joined) previously lost their highlight to a spurious leading-space Other row - Move the question-text/legacy-q_<i> answer lookup out of the SFC into askUserToolParse as answerFor(), per that module's testability intent - Require non-empty question text and option labels (.min(1)) so empty strings are rejected by AJV at the tool boundary instead of failing deeper in the protocol layer

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: afdf84fc7b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-06T07:53:28Z

+    itemsById.set(item.id, item);
+    for (const opt of item.options) labelsByOptionId.set(opt.id, opt.label);
+  }
+  const optionText = (id: string): string => labelsByOptionId.get(id) ?? id;


Reject cross-question option ids

When a client submits a stale or malformed response that pairs one question id with an option id from another question (for example q_1 with opt_0_0), the REST schema accepts the shape and this global lookup translates opt_0_0 to question 0's label before storing it under question 1's text. That gives the model a label that was never offered for that question instead of preserving the raw id for diagnosability, so the lookup should be scoped to the answered qid or reject cross-question ids.

Useful? React with 👍 / 👎.

The translator's option-id lookup was a single flat map across all questions, so a stale or malformed response pairing one question with another question's option id (q_1 + opt_0_0) was silently translated into a label that was never offered for that question. Scope the lookup to the answered question's own options; cross-question and unknown ids now both pass through verbatim, staying diagnosable.

chatgpt-codex-connector Bot reviewed Jul 6, 2026

View reviewed changes

RealKai42 merged commit c12f309 into main Jul 6, 2026
10 checks passed

RealKai42 deleted the kaiyi/douala branch July 6, 2026 08:37

This was referenced Jul 6, 2026

ci: release packages #1378

Merged

ci: release packages #1443

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat(agent-core): feed AskUserQuestion answers back as question text and option labels - #1414

feat(agent-core): feed AskUserQuestion answers back as question text and option labels#1414
RealKai42 merged 3 commits into
mainfrom
kaiyi/douala

RealKai42 commented Jul 6, 2026

Uh oh!

changeset-bot Bot commented Jul 6, 2026 •

edited

Loading

Uh oh!

pkg-pr-new Bot commented Jul 6, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

RealKai42 commented Jul 6, 2026

Problem

Change

Tests

Uh oh!

changeset-bot Bot commented Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦋 Changeset detected

Uh oh!

pkg-pr-new Bot commented Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

changeset-bot Bot commented Jul 6, 2026 •

edited

Loading

pkg-pr-new Bot commented Jul 6, 2026 •

edited

Loading