[WRONG BRANCH] web-search: bound Anthropic sidecar SSE and error bodies to avoid unbounded buffering

[luvs01]

Motivation

• The Anthropic web-search sidecar previously accumulated upstream-controlled SSE frames, answer text, and citation lists with no caps, and used res.text() for non-2xx bodies, which allows a malicious or compromised upstream to exhaust proxy memory.

Description

• Add explicit memory/quantity limits: ANTHROPIC_MAX_SSE_BYTES, ANTHROPIC_MAX_FRAME_CHARS, ANTHROPIC_MAX_TEXT_CHARS, and ANTHROPIC_MAX_SOURCES and enforce them in parseAnthropicSidecarSSE to bound total streamed bytes, per-frame size, accumulated answer length, and retained sources.
• Cancel the stream reader and surface a safe sidecar error when any cap is exceeded instead of retaining attacker-controlled data in memory.
• Replace unbounded res.text() on non-OK responses with the shared bounded reader readBoundedResponseBody and return a status-only fallback when the body is oversized or unavailable.
• Add focused regression tests to tests/web-search-anthropic.test.ts that exercise oversized SSE responses, oversized frames, overlong answer deltas, capped sources, and bounded non-success response bodies.

Testing

• Ran the updated unit tests with bun test --timeout 20000 tests/web-search-anthropic.test.ts, and all tests in that file passed (15 pass).
• Ran type checking with bun x tsc --noEmit and bun run privacy:scan, both of which completed successfully in the local run.
• Ran broader test invocation (bun run test), which exercised the test suite and showed the change did not regress the Anthropic web-search tests, but one unrelated external-provider discovery test (tests/grok-models-effort-list.test.ts) timed out in this environment due to external network/provider discovery dependencies.

---

Codex Task

[coderabbitai]

Warning

Review limit reached

@luvs01, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 54 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 5fd4db54-fac4-4640-a460-bfb01c2a53c4

📥 Commits

Reviewing files that changed from the base of the PR and between 121f1ad and 0538418.

📒 Files selected for processing (2)

• src/web-search/anthropic-executor.ts
• tests/web-search-anthropic.test.ts

---

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

[github-actions]

✅ Deterministic PR hygiene checks passed.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0538418c21

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]

Enforce the frame limit while buffering

When an upstream sends more than 1 MiB without a frame separator and keeps the stream open, this check never runs because processFrame is called only after \n\n or EOF. The parser therefore retains and repeatedly concatenates/scans the oversized frame until the 4 MiB body limit or sidecar timeout, defeating the intended per-frame ceiling and allowing unnecessary memory/CPU consumption; check the unterminated buffer after every decoded chunk and return the existing frame-limit error immediately.

AGENTS.md reference: src/AGENTS.md:L17-L17

Useful? React with 👍 / 👎.

[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.