Test environment variable expansion in SSE headers. Note: The current implementation only expands values that are entirely environmen

Test error handling and fallback.

Test that errors in approaches are properly logged

Test tool execution with config load failure

Test tool execution with non-existent server

Test SSE tool execution without URL

Test stdio tool execution without command

Test tool execution with unsupported transport

Test WebSocket tool execution without URL

Test that existing approaches work without reasoning token changes

Test approach specification via extra_body

Test approach specification via extra_body parameter

Test schema extraction from OpenAI response format.

Test complete integration from approach execution to file logging

Test JSON generation with new outlines API.

Test getting capabilities description with no servers

Test getting tools when no tools are available

Test httpx.Client created with custom certificate path.

Test httpx.Client created with verify=False when SSL disabled.

Test httpx.Client created with verify=True by default.

Test MARS on IMO25 Problem 3 - Functional Equation (Expected: c = 4)

Test MARS on IMO25 Problem 4 - Number Theory (Expected: 6J·12^K formula)

Test that all DeepConf components can be imported.

Test that incompatible requests are properly handled

Test that inference pipeline behavior is unchanged

Test that ChatCompletionUsage defaults reasoning_tokens to 0

Test that ChatCompletionUsage includes reasoning_tokens

Test the info function.

Test that an info message is logged when using custom certificate.

Test MCPServer initialization

Test MCPServerManager initialization

Test custom configuration path

Test default configuration path

Test handling of invalid request IDs and logger statistics

Test JSONGenerator initialization with new API.

Test that JSON plugin falls back gracefully when schema is invalid

Test JSON plugin with actual local inference

Test loading non-existent configuration

Test loading valid configuration

Test that log files have correct JSONL structure and required fields

Test logger initialization and disabled logger behavior

Test approaches work with logging disabled, no request_id, and API errors

Test that logging doesn't significantly impact response time

Test basic MARS functionality with mock client

Test MARS consensus and verification mechanism

Test error handling in parallel execution

Test MARS on challenging problems that require deep reasoning

Test that MARS can be imported correctly

Test that MARS logging provides useful monitoring information

Test that parallel execution shows improvement over theoretical sequential

Test that MARS actually uses ThreadPoolExecutor for parallel execution

Test that worker pool size is calculated correctly

Test that requests with different models are properly separated

Test that MLX batch processing can be created

Test MLX batch processing with actual generation

Test MLX batch processing parameter validation

Test BON, MCTS, and RTO approaches log API calls correctly

Test multiple different approaches get logged correctly

Test handling multiple concurrent conversations and log file naming

Test reasoning tokens with multiple responses (n > 1)

Test n parameter for multiple completions

Test that optillm works exactly the same without --batch-mode

Test that OpenAI client receives the configured httpx client.

Test JSON schema to Pydantic model conversion.

Test that processor errors are handled correctly

Test that PyTorch InferencePipeline has process_batch method

Test PyTorch batch processing with small model

Test RE2 approach with conversation logging

Test readurls plugin uses custom certificate path.

Test readurls plugin respects SSL verification disabled.

Test readurls plugin respects SSL verification enabled.

Test reasoning token calculation with mock content

Test that responses without thinking still work normally

Test fallback estimation when tokenizer fails

Test that reasoning tokens are included in API responses

Test that reasoning tokens are properly logged

Test reasoning tokens with n > 1

Test reasoning tokens with a prompt designed to trigger thinking

Test that required modules can be imported

Test that API responses include completion_tokens_details

Test API response when there are no reasoning tokens

Test that plugin has run function defined

Test the main run function with a valid schema.