Refactor code interpreter tool

[fcogidi]

Summary

Refactors the E2B CodeInterpreter tool: clearer constructor parameters (sandbox vs code vs HTTP timeouts), stronger typing and validation, structured JSON errors for timeouts/stream failures when return_errors_as_json=True, and a fresh sandbox per run_code with cleanup in finally. Bumps aieng-agents to 0.3.0 and updates the monorepo dependency and README example accordingly.

Clickup Ticket(s): N/A

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)
• 📝 Documentation update
• 🔧 Refactoring (no functional changes)
• ⚡ Performance improvement
• 🧪 Test improvements
• 🔒 Security fix

Changes Made

• Replace timeout_seconds with explicit sandbox_timeout_seconds, code_execution_timeout_seconds, and request_timeout_seconds; validate ranges and document behavior in the class docstring.
• Wire E2B AsyncSandbox.create / run_code with template, envs, metadata, allow_internet_access, and httpx-style request timeouts.
• Add _failure_json / return_errors_as_json path so TimeoutException, httpx.TimeoutException, and httpx.RemoteProtocolError can return CodeInterpreterOutput JSON instead of raising.
• Add _validate_str_dict for envs / metadata; tighten CodeInterpreterOutput typing (**kwargs: Any); use contextlib.suppress when killing the sandbox in finally.
• Update integration tests to use sandbox_timeout_seconds=15.
• Bump aieng-agents to 0.3.0 in aieng-agents/pyproject.toml, root pyproject.toml, and uv.lock; simplify the README CodeInterpreter snippet.

Testing

• Tests pass locally (uv run pytest tests/)
• Type checking passes (uv run mypy <src_dir>)
• Linting passes (uv run ruff check src_dir/)
• Manual testing performed (describe below)

Related Issues

Relates to consumers that instantiated CodeInterpreter(timeout_seconds=...) — they must migrate to sandbox_timeout_seconds (and optionally the new timeout knobs).

Deployment Notes

• Semver: package version 0.3.0 reflects the breaking CodeInterpreter API.
• Downstream apps should pin or bump aieng-agents[all]>=0.3.0 and update constructor kwargs.

Checklist

• Code follows the project's style guidelines
• Self-review of code completed
• Documentation updated (if applicable)
• No sensitive information (API keys, credentials) exposed

[Copilot]

Pull request overview

Refactors the aieng-agents E2B CodeInterpreter tool API to separate sandbox lifetime, code execution timeout, and HTTP request timeout, while bumping the package to 0.3.0 and updating consumer dependency pins and examples.

Changes:

• Refactor CodeInterpreter constructor + execution flow (fresh sandbox per run_code, validation, structured JSON error outputs when enabled).
• Update integration tests to use sandbox_timeout_seconds.
• Bump aieng-agents to 0.3.0 and update monorepo dependency pins + README snippet.

Reviewed changes

Copilot reviewed 5 out of 6 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
aieng-agents/aieng/agents/tools/code_interpreter.py	Refactors API + timeout handling, adds validation and JSON error-return path, ensures sandbox cleanup.
aieng-agents/tests/tools/test_code_interpreter.py	Updates constructor usage for the breaking timeout parameter rename.
aieng-agents/README.md	Updates CodeInterpreter example (currently mismatched with new API/return type).
aieng-agents/pyproject.toml	Bumps package version to 0.3.0.
pyproject.toml	Updates dependency constraint to aieng-agents[all]>=0.3.0.
uv.lock	Locks aieng-agents to 0.3.0.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The new return_errors_as_json behavior (converting TimeoutException, httpx.TimeoutException, and httpx.RemoteProtocolError into structured CodeInterpreterOutput JSON) is not exercised by the existing tests in tests/tools/test_code_interpreter.py. Please add tests that force each failure mode (or mock AsyncSandbox.run_code) and assert the returned JSON has error.name set and remains parseable by CodeInterpreterOutput.