fix(ollama): bound pending native tool calls

[luvs01]

Motivation

• Prevent a remotely-controlled Ollama-native upstream from exhausting the proxy process by unboundedly retaining provider-controlled tool names and an unlimited number of pending tool calls.
• Ensure retained provider data (tool names and per-call bookkeeping) is accounted to the existing translator budget so aggregate in-memory retention is limited.

Description

• Add hard limits and accounting to the native Ollama parser: introduce NATIVE_TOOL_NAME_MAX_BYTES = 1024, NATIVE_MAX_PENDING_TOOL_CALLS = 128, and a small NATIVE_TOOL_CALL_BOOKKEEPING_BYTES overhead, and enforce them during nativeMessageEvents. (file: src/adapters/ollama-native.ts)
• Charge retained tool-name bytes plus bookkeeping to the translator budget before storing a new pending NativeStreamToolCall, and fail admission when the budget or per-call/pending-count bounds are exceeded. (file: src/adapters/ollama-native.ts)
• Add focused regression tests that exercise oversized tool names, excessive pending-call counts, and aggregate retained-name budget enforcement. (file: tests/ollama-native-parser.test.ts)

Testing

• Ran the focused adapter tests with bun test tests/ollama-native-parser.test.ts and the modified suite passed (29 tests, 0 failures).
• Ran type checking with bun run typecheck and it succeeded.
• Started the full test suite bun run test for broader confidence; the run surfaced unrelated pre-existing failures in Lab automation / API auth and a long-running chat-completions timeout, so only the focused Ollama-native tests and typecheck are relied on for this change.

---

Codex Task

[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: Team

Run ID: a8f20d40-8493-4172-98af-206cd0d54291

---

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 Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-08-31T01:00:57.842256Z	8adc6f4	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

✅ Deterministic PR hygiene checks passed.

[github-actions]

✅ READY

• all PR quality gates passed.

Hygiene

✅ Deterministic PR hygiene checks passed.

[luvs01]

Closing: submitted upstream as lidge-jun#5781, rebased onto current dev and verified — bun test tests/providers/ollama/ollama-native-parser.test.ts 29 pass. Nothing left to merge here.