Module api_extensions 

Extended /v1/chat/completions handler.

Adds support for tools (function calling), logprobs, n > 1 completions, response format constraints (JSON mode / JSON Schema), stop sequences, and frequency/presence penalties on top of the base server implementation.

Structs

JsonModeEnforcer

Wraps generation to produce valid JSON output.

StopChecker

Detects and truncates text at stop sequences.

Functions

apply_frequency_penalty

Apply frequency and presence penalties in-place to a logit vector.

extended_chat_completions

Handler for POST /v1/chat/completions/extended.

generate_n_completions

Generate n independent completions from the same prompt, seeding each run with base_seed + i for determinism.