[WRONG BRANCH] fix(openai-chat): bound Moonshot schema inlining bytes

[luvs01]

Motivation

• Prevent a request-amplification DoS where inlining a large $defs properties map into many sibling-bearing $ref nodes could multiply serialized output and exhaust process memory.
• Preserve the chosen Moonshot $ref-with-siblings semantics (conjunction for required/properties, node-narrowing otherwise) while making inlining resource-bounded and safe.

Description

• Add a serialized-byte budget MOONSHOT_MAX_INLINED_SCHEMA_BYTES (1 MiB) and an iterative estimator serializedJsonBytesUpTo to bound how many bytes of a referenced schema may be inlined. (src/adapters/openai-chat.ts)
• Track per-normalization state with remainingInlineBytes and inlineSizeCache so large referenced targets are measured once and charged before being copied, and reuse cached sizes for repeated references. (src/adapters/openai-chat.ts)
• When the inlining budget would be exceeded, preserve a Moonshot-compatible bare $ref instead of duplicating the large map, and keep the existing depth/node/expansion guards. (src/adapters/openai-chat.ts)
• Add a focused regression test that constructs a 10k-entry boolean properties definition referenced many times and asserts that emitted output stays bounded and that some references remain as bare $ref. (tests/moonshot-tool-schema.test.ts)
• Update the adapter registry decision note to document the new serialized-byte budget and its rationale. (structure/10_adapter-registry.md)

Testing

• Ran the focused Moonshot suite with bun test tests/moonshot-tool-schema.test.ts, which passed (17 tests, 0 failures).
• Ran typecheck with bun run typecheck, which completed without type errors.
• Ran the privacy check with bun run privacy:scan, which passed.
• Attempting the full repository bun run test was not used for acceptance here because unrelated, pre-existing management/auth tests timed out/failed in this environment; the change was validated by the focused tests above.

---

Codex Task

[github-actions]

✅ Deterministic PR hygiene checks passed.

[github-actions]

⏳ DRAFT

• wrong target branch (main); retarget to dev.

What to do

• Retarget this PR to dev — all contributions go to dev.

Its title has been prefixed with [WRONG BRANCH].
This pull request was already a draft. Its draft status will be preserved after every issue above is resolved.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 3f6a8b4d-611c-4d18-86cf-a8d9eea9f795

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ab7f684bfc

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]

Share the byte allowance across the whole tool catalog

The allowance is initialized inside normalizeMoonshotToolParameters, so toolsToChatFormat grants a fresh 1 MiB amplification budget to every tool. Because the inbound tool arrays have no aggregate count or output-size limit, a catalog containing many individually small schemas can still exhaust memory while constructing and serializing the final request; for example, I reproduced a 128-tool, 3.36 MB catalog expanding to about 82 MB. Allocate the budget once per request/catalog and pass it through each normalization, or enforce an aggregate serialized-output cap.

Useful? React with 👍 / 👎.