Fix test_tool_choice_none_after_tool_results test failure

[fede-kamel]

Problem

The test_tool_choice_none_after_tool_results test added in PR #50 fails with:

AttributeError: module 'oci.generative_ai_inference.models' has no attribute 'Tool'

This was reported by @paxiaatucsdedu in #50 (comment)

Root Cause

The test used non-existent OCI SDK classes and had several issues:

1. Used models.Tool which doesn't exist in the OCI SDK
2. Incorrect test expectations - Expected tool_choice=none after any ToolMessage, but the actual implementation only sets it when:
   • max_sequential_tool_calls limit is exceeded, OR
   • An infinite loop is detected
3. Missing stop parameter in _prepare_request() call
4. Missing kwargs - Didn't pass bound model kwargs containing tools

Solution

Fixed the test to:

1. ✅ Use Python function instead of OCI SDK mock objects (following pattern from other tests)
2. ✅ Create scenario with 3 sequential tool calls to exceed max_sequential_tool_calls=3 limit
3. ✅ Add required stop parameter to _prepare_request() call
4. ✅ Pass bound model kwargs to ensure tools are included in the request

Changes

• libs/oci/tests/unit_tests/chat_models/test_oci_generative_ai.py
  • Replaced models.Tool with Python function get_weather(city: str)
  • Updated test to create 3 ToolMessages to trigger the limit
  • Fixed _prepare_request() call signature
  • Updated docstring to accurately describe the test

Test Results

Before: ❌ FAILED

AttributeError: module 'oci.generative_ai_inference.models' has no attribute 'Tool'

After: ✅ PASSED

tests/unit_tests/chat_models/test_oci_generative_ai.py::test_tool_choice_none_after_tool_results PASSED

All 6 tool-related unit tests now pass:

test_meta_tool_calling PASSED
test_cohere_tool_choice_validation PASSED
test_meta_tool_conversion PASSED
test_ai_message_tool_calls_direct_field PASSED
test_ai_message_tool_calls_additional_kwargs PASSED
test_tool_choice_none_after_tool_results PASSED ✅

Impact

• ✅ Unblocks developers encountering this test failure
• ✅ No changes to production code - test-only fix
• ✅ Backward compatible
• ✅ Ready to merge immediately

---

Related:

• Fixes issue reported in PR Fix infinite tool calling loop with Meta Llama models #50: Fix infinite tool calling loop with Meta Llama models #50 (comment)
• Originally fixed as part of PR Optimize tool call conversions to eliminate redundant API lookups #53, now separated for expedited merge

[fede-kamel]

✅ All Unit Tests Verified

Ran all unit tests for ChatOCIGenAI to verify the fix:

cd ~/Projects/langchain-oracle/libs/oci
.venv/bin/python -m pytest tests/unit_tests/chat_models/test_oci_generative_ai.py -v

Test Results: 12/12 PASSED ✅

tests/unit_tests/chat_models/test_oci_generative_ai.py::test_llm_chat[cohere.command-r-16k] PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_llm_chat[meta.llama-3.3-70b-instruct] PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_meta_tool_calling PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_cohere_tool_choice_validation PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_meta_tool_conversion PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_json_mode_output PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_json_schema_output PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_auth_file_location PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_include_raw_output PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_ai_message_tool_calls_direct_field PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_ai_message_tool_calls_additional_kwargs PASSED
tests/unit_tests/chat_models/test_oci_generative_ai.py::test_tool_choice_none_after_tool_results PASSED ✅

All tool-related tests pass, including the previously failing test_tool_choice_none_after_tool_results!

This fix is ready to merge and will unblock developers encountering the test failure.