fix(pydantic-ai): Add error handling to agent execution

[ericapisani]

Fix exception handling in pydantic-ai agent template

The pydantic-ai template's agent execution was not handling exceptions
that could occur during agent.run() calls. This caused test execution
to crash instead of gracefully handling errors.

Wrapping the agent execution in a try-except block allows the test to
continue running and properly capture error information when exceptions
occur, making the tests more robust and providing better visibility
into failures.

This fix enables proper validation of error handling behavior in the
pydantic-ai integration tests.

The tests failing within this suite are due to missing fixes in PRs getsentry/sentry-python#5596 and #89

[github-actions]

🟡 AI SDK Integration Test Results

Status: 168 tests failing (no regressions)

Summary

Metric	main	PR	Change
Total Tests	384	384	—
Passed	192	192	—
Failed	168	168	—

Test Matrix

Agent Tests

SDK	Basic Agent Test	Conversation ID Agent Test	Long Input Agent Test	Tool Call Agent Test	Tool Error Agent Test	Vision Agent Test
nextjs/mastra	❌	❌	—	❌	❌	❌
nextjs/vercel	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str
node/manual	✅	❌	✅	✅	✅	✅
node/mastra	❌	❌	❌	❌	❌	❌
node/vercel	❌	❌	❌	❌	❌	❌
php/laravel	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str
python/langgraph	❌a ❌s	❌a ❌s	❌a ❌s	❌a ❌s	❌a ❌s	❌a ❌s
python/manual	✅a ✅s	✅a ✅s	✅a ✅s	✅a ✅s	✅a ✅s	✅a ✅s
python/openai-agents	✅	✅	❌	❌	❌	❌
python/pydantic-ai	✅	✅	❌	❌	❌	❌

Embedding Tests

SDK	Basic Embeddings Test
browser/google-genai	❌
browser/langchain	❌
browser/openai	✅
nextjs/google-genai	❌
nextjs/langchain	❌
nextjs/openai	✅
nextjs/vercel	❌
node/google-genai	❌
node/langchain	❌
node/openai	✅
node/vercel	❌
php/laravel	❌
python/google-genai	❌a, blk ❌s, blk
python/langchain	❌a, blk ❌s, blk
python/litellm	❌a, blk ✅s, blk
python/manual	✅a, blk ✅s, blk
python/openai	✅a, blk ✅s, blk

LLM Tests

SDK	Basic Error LLM Test	Basic LLM Test	Conversation ID LLM Test	Long Input LLM Test	Multi-Turn LLM Test	Vision LLM Test
browser/anthropic	✅blk ❌str	✅blk ✅str	❌blk ❌str	✅blk ✅str	✅blk ✅str	✅blk ✅str
browser/google-genai	✅blk ✅str	✅blk ✅str	❌blk ❌str	❌blk ❌str	❌blk ❌str	✅blk ✅str
browser/langchain	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str
browser/openai	✅blk ✅str	✅blk ❌str	❌blk ❌str	✅blk ❌str	✅blk ❌str	✅blk ❌str
nextjs/anthropic	✅blk ✅str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str
nextjs/google-genai	✅blk ✅str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str
nextjs/langchain	✅blk ✅str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str	❌blk ❌str
nextjs/openai	✅blk ✅str	✅blk ❌str	✅blk ❌str	✅blk ❌str	✅blk ❌str	✅blk ❌str
node/anthropic	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str
node/google-genai	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str
node/langchain	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str
node/manual	—	✅	✅	✅	✅	✅
node/openai	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str	✅blk ✅str
python/anthropic	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str
python/google-genai	✅a, blk ✅a, str ✅s, blk ✅s, str	✅a, blk ✅a, str ✅s, blk ✅s, str	✅a, blk ✅a, str ✅s, blk ✅s, str	✅a, blk ✅a, str ✅s, blk ✅s, str	❌a, blk ❌a, str ❌s, blk ❌s, str	❌a, blk ❌a, str ❌s, blk ❌s, str
python/langchain	❌a, blk ❌a, str ❌s, blk ❌s, str	✅a, blk ✅a, str ✅s, blk ✅s, str	✅a, blk ✅a, str ✅s, blk ✅s, str	✅a, blk ✅a, str ✅s, blk ✅s, str	✅a, blk ✅a, str ✅s, blk ✅s, str	✅a, blk ✅a, str ✅s, blk ✅s, str
python/litellm	❌a, blk ❌a, str ❌s, blk ❌s, str	❌a, blk ❌a, str ✅s, blk ✅s, str	❌a, blk ❌a, str ✅s, blk ✅s, str	❌a, blk ❌a, str ✅s, blk ✅s, str	❌a, blk ❌a, str ✅s, blk ✅s, str	❌a, blk ❌a, str ✅s, blk ✅s, str
python/manual	—	✅a, blk ✅s, blk	✅a, blk ✅s, blk	✅a, blk ✅s, blk	✅a, blk ✅s, blk	✅a, blk ✅s, blk
python/openai	❌a, blk ❌a, str ❌s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	✅a, blk ❌a, str ✅s, blk ❌s, str	❌a, blk ❌a, str ❌s, blk ❌s, str

Legend: ✅ Pass | ❌ Fail | ✅🔧 Fixed | ❌📉 Regressed | ✅🆕 New (pass) | ❌🆕 New (fail) | 🗑️ Removed | _str=streaming _blk=blocking _a=async _s=sync

---

Generated by AI SDK Integration Tests

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Swallowed exceptions prevent sentry_sdk.capture_exception from firing
  • Added 'raise' statement after printing exception to allow it to propagate to base template's error handler.

Or push these changes by commenting:

@cursor push 4a222eb538

Preview (4a222eb538)

diff --git a/src/runner/templates/agents/python/pydantic-ai/template.njk b/src/runner/templates/agents/python/pydantic-ai/template.njk
--- a/src/runner/templates/agents/python/pydantic-ai/template.njk
+++ b/src/runner/templates/agents/python/pydantic-ai/template.njk
@@ -73,6 +73,7 @@
         print(f"Turn {{ loop.index }} Response: {result.output}")
     except Exception as e:
         print(f"Turn {{ loop.index }} Error: {type(e).__name__}: {e}")
+        raise
 {% if not loop.last %}
     print()  # Blank line between turns
 {% endif %}

_{This Bugbot Autofix run was free. To enable autofix for future PRs, go to the Cursor dashboard.}

[cursor]

Swallowed exceptions prevent sentry_sdk.capture_exception from firing

High Severity

The new except Exception as e block catches and prints the error but never re-raises it. When the causeAPIError flag is set, base.python.njk wraps the main() call in its own try-except that calls sentry_sdk.capture_exception(e) — but the exception is now silently swallowed inside main(), so it never propagates. This means Sentry never captures the exception, breaking the error-handling test scenarios that depend on it.

 

[ericapisani]

I don't think this is entirely accurate as I verified during testing that an error span is being returned from the Python SDK (and not being swallowed as Cursor is saying here) as a result of the generic exception thrown, and is being asserted correctly within the testing framework (example below from a test run):

{
    trace_id: '25a8c19c306a42a2b1bd22b34daa3e84',
    span_id: '85d009fd8743fdb6',
    parent_span_id: '905cfffa695c1f38',
    same_process_as_parent: true,
    op: 'gen_ai.execute_tool',
    description: 'execute_tool read_file',
    start_timestamp: '2026-03-06T16:13:37.239475Z',
    timestamp: '2026-03-06T16:13:37.240453Z',
    origin: 'auto.ai.pydantic_ai',
    status: 'internal_error',
    tags: { status: 'internal_error' },
    data: {
      'thread.id': '8372809984',
      'thread.name': 'MainThread',
      'gen_ai.operation.name': 'execute_tool',
      'gen_ai.tool.type': 'function',
      'gen_ai.tool.name': 'read_file',
      'gen_ai.agent.name': 'file_assistant',
      'gen_ai.tool.input': '{"path": "/nonexistent/file.txt"}'
    }
  }

[constantinius]

Looks good, thank you!