Log and forward requests

Set up test fixtures

Set up test fixtures

Set up MLX test fixtures

Set up test fixtures with logging capture

Reset server_config before each test.

Set up test environment.

Set up test environment.

Set up test environment.

Set up test client

Check server availability

Set up test fixtures

Set up test environment

Set up test fixtures.

Set up integration test environment

Set up test fixtures

Set up test fixtures

Set up OptILLM server for testing

Replace standard attention with memory-efficient version

Stage 2: Use the discovered reasoning structure to solve a specific problem.

Start optillm server for testing Returns the process handle

Stop health checking

Remove <think></think> tags from text (for debugging/logging) Args: text: Text potentially containing thinking tags Returns:

Calculate the success rate of this strategy.

Clean up test fixtures

Clean up after tests

Clean up test fixtures

Restore original server_config after each test.

Restore original server_config.

Restore original server_config.

Restore original server_config.

Clean up test environment

Clean up test fixtures

Clean up server

Test answer extraction specifically with controlled responses

Test API still returns proper errors without auth

Test approach prefix in model name

Test that AzureOpenAI client receives the configured httpx client.

Test basic chat completion

Test basic none approach with conversation logging

Test that multiple requests form a batch

Test error conditions in batch mode

Test that batch statistics are collected correctly

Test that partial batches process after timeout

Test that batching provides performance improvement

Test that Cerebras client receives the configured httpx client.

Test that ChatCompletion.model_dump includes reasoning_tokens in usage

Test ChatCompletion model_dump includes reasoning_tokens

Test that chat completion responses have proper structure

Test ChatCompletionUsage supports reasoning_tokens

Test that CLI arguments are properly parsed

Test --no-ssl-verify CLI flag disables SSL verification.

Test that CLI arguments override environment variables.

Test --ssl-cert-path CLI argument.

Test compact plugin module

Test RStar and Z3 Solver class-based approaches log API calls correctly

Test that concurrent requests are logged properly

Test ConfidenceCalculator functionality.

Test configuration validation.

Test unsupported transport type

Test successful SSE connection

Test SSE connection validation

Test stdio connection validation

Test WebSocket connection validation

Test complete conversation lifecycle: start, log calls, errors, finalize

Test CoT Reflection approach with conversation logging

Test basic functionality of count_reasoning_tokens

Test edge cases with truncated responses

Test edge cases with empty or invalid input

Test with empty or None input

Test with malformed think tags

Test with both complete and truncated think blocks

Test with multiline think blocks

Test with multiple think tag blocks

Test with multiple think tag blocks

Test that we don't count think-like content that isn't actually truncated

Test with text that has no think tags

Test with special characters in think blocks

Test fallback when tokenizer fails

Test counting tokens when response is truncated (no closing </think> tag)

Test with a simple mock tokenizer

Test counting tokens in think tags

Test with a mock tokenizer for precise counting

Test with text that has no think tags

Test creating default configuration file

Test that SSL verification is enabled by default.

Test default configuration for stdio transport

Test OPTILLM_SSL_CERT_PATH environment variable.

Test OPTILLM_SSL_VERIFY=false environment variable.

Test OPTILLM_SSL_VERIFY=true environment variable.