@brief brief print request @param request the request

@brief cancel request @param request_id the request ID @return true if request was found and cancelled

\brief Check hash of model files \param model_tag the model tag \return true if all files are present and compatible, false otherwise

\brief Cleanup CURL library (call this once at program shutdown)

\brief Clear the context \param input_list, std::vector<std::string>

\brief Load a model \param input_list, std::vector<std::string>

\brief Save the history \param input_list, std::vector<std::string>

\brief Set the model parameters \param input_list, std::vector<std::string>

\brief Show the model information \param input_list, std::vector<std::string>

\brief Show the status \param input_list, std::vector<std::string>

AVX512 vectorized complex magnitude: sqrt(re^2 + im^2) complex_interleaved: input [n_bins * 2] as (re0, im0, re1, im1, ...) magnitude: outpu

@brief Convert OpenAI-style assistant tool_calls + following tool messages into the Gemma4 chat-template format, which expects a single assistant mess

\brief Convert the codepoint to utf8 \param input the input \return the utf8 string

@brief returns a const reverse iterator to the last element @sa https://json.nlohmann.me/api/basic_json/crbegin/

@brief create lm server @param models the model list @param downloader the downloader @param default_tag the default tag @param port the port @return

@brief returns a const reverse iterator to one before the first @sa https://json.nlohmann.me/api/basic_json/crend/

\brief cut the tag, some program adds a prefix to the tag, like "Ollama/llama3.2-1B", we need to cut the prefix \param tag the tag of the model \retur

\brief Download a file from URL to a local file \param url the URL to download from \param local_path the local path to save the file \param progress_

\brief Download multiple files with progress tracking \param downloads the downloads \param progress_cb the progress callback \return true if the file

! @brief dump a floating-point number Dump a given floating-point number to output stream @a o. Works internally with @a number_buffer.

\brief duration in milliseconds \param start the start time \param stop the stop time \return the duration in milliseconds

\brief duration in microseconds \param start the start time \param stop the stop time \return the duration in microseconds

@brief checks whether the container is empty. @sa https://json.nlohmann.me/api/basic_json/empty/

Fast AVX512 log approximation: natural log via log2 polynomial Accuracy: ~1e-5 relative error over normal float range

\brief Legacy filter method for backward compatibility

! For n != 0, returns k, such that pow10 := 10^(k-1) <= n < 10^k. For n == 0, returns 1 and sets pow10 := 1. */

\brief get all the models \return all the models in json

\brief get all the models \return all the models in json

! For a normalized diyfp w = f * 2^e, this function returns a (normalized) cached power-of-ten c = f_c * 2^e_c, such that the exponent of the product

@brief get current time string, format: hh:mm:ss mm:dd:yyyy @return the current time string

\brief get the error metrics \param y the output value, 256-bits, 8xbf16 \param y_ref the reference value, 256-bits, 8xbf16 \return the error metrics

\brief Get interactive input with arrow key support

\brief Get missing files \param model_tag the model tag \return the missing files

Model-specific factory function for Qwen family and DeepSeek_r1_0528_8b

! Generates V = buffer * 10^decimal_exponent, such that M- <= V <= M+. M- and M+ must be normalized and share the same exponent -60 <= e <= -32. */

@brief Handle the chat request @param request the request @param send_response the send response @param send_streaming_response the send streaming res

@brief Handle the generate request @param request the request @param send_response the send response @param send_streaming_response the send streaming

@brief Handle the models request @param request the request @param send_response the send response @param send_streaming_response the send streaming r

@brief Handle the models request (open ai) @param request the request @param send_response the send response @param send_streaming_response the send s

@brief Handle the openai audio transcriptions request @param request the request @param send_response the send response @param send_streaming_response

@brief Handle the openai chat completion request @param request the request @param send_response the send response @param send_streaming_response the

@brief Handle the openai completion request @param request the request @param send_response the send response @param send_streaming_response the send

@brief Handle the ps request @param request the request @param send_response the send response @param send_streaming_response the send streaming respo

@brief Handle the pull request @param request the request @param send_response the send response @param send_streaming_response the send streaming res

@brief Handle the show request @param request the request @param send_response the send response @param send_streaming_response the send streaming res

Convert frequency from hertz to fractional octave numbers. Adapted from *librosa*. Args: freq (`float` or `np.ndarray`):

\brief Hide the cursor

\brief Initialize CURL library (call this once at program startup) \return true if the CURL library is initialized, false otherwise