cactus-react-native 0.2.11 → 1.0.2

package/ios/cactus.xcframework/ios-arm64/cactus.framework/Headers/sampling.h

@@ -1,107 +0,0 @@
-#pragma once
-
-#include "llama.h"
-
-#include "common.h"
-
-#include <string>
-#include <vector>
-
-// common_sampler extends llama_sampler with additional functionality:
-//
-//  - grammar support
-//  - custom sampler logic based on the parameters
-//  - history of the last accepted tokens
-//  - performance metrics
-//
-// This goal is to have a common implementation of the sampling logic shared across the examples.
-// For example, depending on the temperature, the sampling chain can be very simple (greedy) or more
-// complex (top-k, top-p, etc).
-//
-// Another example is related to the grammar. In general, the grammar constraints applied on the full
-// vocabulary can be very taxing. To improve performance, the grammar can be applied only to the sampled
-// token in order to verify if it fits the grammar. And only if the token doesn't fit the grammar, the
-// grammar constraints are applied to the full vocabulary and the token is resampled.
-//
-// The common_sampler also maintains a container with the last accepted tokens. In the future, this can
-// be moved into the core llama library.
-//
-// For convenience, the common_sampler also maintains a container with the current candidate tokens.
-// This can be used to access the probabilities of the rest of the non-sampled tokens.
-//
-// TODO: measure grammar performance
-//
-
-struct common_sampler;
-
-// llama_sampler API overloads
-
-struct common_sampler * common_sampler_init(const struct llama_model * model, const struct common_params_sampling & params);
-
-void common_sampler_free(struct common_sampler * gsmpl);
-
-// if accept_grammar is true, the token is accepted both by the sampling chain and the grammar
-void                    common_sampler_accept(struct common_sampler * gsmpl, llama_token token, bool accept_grammar);
-void                    common_sampler_reset (struct common_sampler * gsmpl);
-struct common_sampler * common_sampler_clone (struct common_sampler * gsmpl);
-
-// arguments can be nullptr to skip printing
-void common_perf_print(const struct llama_context * ctx, const struct common_sampler * gsmpl);
-
-// extended sampling implementation:
-//
-// - set logits
-// - apply the configured sampler chain
-// - check if the token fits the grammar (if any)
-// - if not: resample by first applying the grammar constraints and then sampling again (slower path)
-//
-// if grammar_first is true, the grammar is applied before the samplers (slower)
-// useful in cases where all the resulting candidates (not just the sampled one) must fit the grammar
-//
-llama_token common_sampler_sample(struct common_sampler * gsmpl, struct llama_context * ctx, int idx, bool grammar_first = false);
-
-// generalized version of common_sampler_sample
-//
-// will cross-reference the sampled tokens with a batch of draft tokens and accept those that match
-// if the sampler disagrees at some point, we stop and return the accepted tokens up to now
-//
-//      common_sampler_sample_n(gsmpl, ctx, { idx }, {});
-//
-// is equivalent to
-//
-//      common_sampler_sample(gsmpl, ctx, idx);
-//      common_sampler_accept(gsmpl, token, true);
-//
-// requires: idxs.size() == draft.size() + 1
-//
-// returns at least 1 token, up to idxs.size()
-//
-std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const std::vector<int> & idxs, const llama_tokens & draft, bool grammar_first = false);
-
-// assume idxs == [ 0, 1, 2, ..., draft.size() ]
-std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const llama_tokens & draft, bool grammar_first = false);
-
-uint32_t common_sampler_get_seed(const struct common_sampler * gsmpl);
-
-// helpers
-
-// access the internal list of current candidate tokens
-llama_token_data_array * common_sampler_get_candidates(struct common_sampler * gsmpl);
-
-// get the last accepted token
-llama_token common_sampler_last(const struct common_sampler * gsmpl);
-
-// print the sampler chain into a string
-std::string common_sampler_print(const struct common_sampler * gsmpl);
-
-// get a string representation of the last accepted tokens
-std::string common_sampler_prev_str(common_sampler * gsmpl, llama_context * ctx, int n);
-
-char        common_sampler_type_to_chr(enum common_sampler_type cnstr);
-std::string common_sampler_type_to_str(enum common_sampler_type cnstr);
-
-std::vector<enum common_sampler_type> common_sampler_types_from_names(const std::vector<std::string> & names, bool allow_alt_names);
-std::vector<enum common_sampler_type> common_sampler_types_from_chars(const std::string & chars);
-
-llama_sampler * llama_sampler_init_llg(const llama_vocab * vocab,
-                const char * grammar_kind, const char * grammar_data);

package/ios/cactus.xcframework/ios-arm64/cactus.framework/Headers/unicode-data.h

@@ -1,20 +0,0 @@
-#pragma once
-
-#include <cstdint>
-#include <vector>
-#include <unordered_map>
-#include <unordered_set>
-
-struct range_nfd {
-    uint32_t first;
-    uint32_t last;
-    uint32_t nfd;
-};
-
-static const uint32_t MAX_CODEPOINTS = 0x110000;
-
-extern const std::initializer_list<std::pair<uint32_t, uint16_t>> unicode_ranges_flags;
-extern const std::unordered_set<uint32_t> unicode_set_whitespace;
-extern const std::initializer_list<std::pair<uint32_t, uint32_t>> unicode_map_lowercase;
-extern const std::initializer_list<std::pair<uint32_t, uint32_t>> unicode_map_uppercase;
-extern const std::initializer_list<range_nfd> unicode_ranges_nfd;

package/ios/cactus.xcframework/ios-arm64/cactus.framework/Headers/unicode.h

@@ -1,66 +0,0 @@
-#pragma once
-
-#include <cstdint>
-#include <string>
-#include <vector>
-
-struct unicode_cpt_flags {
-    enum {
-        UNDEFINED       = 0x0001,
-        NUMBER          = 0x0002,  // regex: \p{N}
-        LETTER          = 0x0004,  // regex: \p{L}
-        SEPARATOR       = 0x0008,  // regex: \p{Z}
-        ACCENT_MARK     = 0x0010,  // regex: \p{M}
-        PUNCTUATION     = 0x0020,  // regex: \p{P}
-        SYMBOL          = 0x0040,  // regex: \p{S}
-        CONTROL         = 0x0080,  // regex: \p{C}
-        MASK_CATEGORIES = 0x00FF,
-    };
-
-    // codepoint type
-    uint16_t is_undefined   : 1;
-    uint16_t is_number      : 1;  // regex: \p{N}
-    uint16_t is_letter      : 1;  // regex: \p{L}
-    uint16_t is_separator   : 1;  // regex: \p{Z}
-    uint16_t is_accent_mark : 1;  // regex: \p{M}
-    uint16_t is_punctuation : 1;  // regex: \p{P}
-    uint16_t is_symbol      : 1;  // regex: \p{S}
-    uint16_t is_control     : 1;  // regex: \p{C}
-    // helper flags
-    uint16_t is_whitespace  : 1;  // regex: \s
-    uint16_t is_lowercase   : 1;
-    uint16_t is_uppercase   : 1;
-    uint16_t is_nfd         : 1;
-
-    // decode from uint16
-    inline unicode_cpt_flags(const uint16_t flags = 0) {
-        *reinterpret_cast<uint16_t*>(this) = flags;
-    }
-
-    inline uint16_t as_uint() const {
-        return *reinterpret_cast<const uint16_t*>(this);
-    }
-
-    inline uint16_t category_flag() const {
-        return this->as_uint() & MASK_CATEGORIES;
-    }
-};
-
-size_t unicode_len_utf8(char src);
-
-std::string unicode_cpt_to_utf8  (uint32_t cpt);
-uint32_t    unicode_cpt_from_utf8(const std::string & utf8, size_t & offset);
-
-std::vector<uint32_t> unicode_cpts_from_utf8(const std::string & utf8);
-
-std::vector<uint32_t> unicode_cpts_normalize_nfd(const std::vector<uint32_t> & cpts);
-
-unicode_cpt_flags unicode_cpt_flags_from_cpt (uint32_t cpt);
-unicode_cpt_flags unicode_cpt_flags_from_utf8(const std::string & utf8);
-
-std::string unicode_byte_to_utf8(uint8_t byte);
-uint8_t     unicode_utf8_to_byte(const std::string & utf8);
-
-uint32_t unicode_tolower(uint32_t cpt);
-
-std::vector<std::string> unicode_regex_split(const std::string & text, const std::vector<std::string> & regex_exprs);

package/ios/cactus.xcframework/ios-arm64_x86_64-simulator/cactus.framework/Headers/cactus.h

@@ -1,232 +0,0 @@
-#ifndef CACTUS_H
-#define CACTUS_H
-
-#include <sstream>
-#include <iostream>
-#include <chrono>
-#include "chat.h"
-#include "common.h"
-#include "ggml.h"
-#include "gguf.h"
-#include "llama.h"
-#include "llama-impl.h"
-#include "sampling.h"
-#if defined(__ANDROID__)
-#include <android/log.h>
-#endif
-
-struct mtmd_context;
-
-namespace cactus {
-
-std::string tokens_to_output_formatted_string(const llama_context *ctx, const llama_token token);
-
-std::string tokens_to_str(llama_context *ctx, const std::vector<llama_token>::const_iterator begin, const std::vector<llama_token>::const_iterator end);
-
-lm_ggml_type kv_cache_type_from_str(const std::string & s);
-
-enum stop_type
-{
-    STOP_FULL,
-    STOP_PARTIAL,
-};
-
-enum tts_type {
-    TTS_UNKNOWN = -1,
-    TTS_OUTETTS_V0_2 = 1,
-    TTS_OUTETTS_V0_3 = 2,
-};
-
-struct completion_token_output
-{
-    struct token_prob
-    {
-        llama_token tok;
-        float prob;
-    };
-
-    std::vector<token_prob> probs;
-    llama_token tok;
-};
-
-struct conversation_result {
-    std::string text;
-    std::chrono::milliseconds time_to_first_token;
-    std::chrono::milliseconds total_time;
-    int tokens_generated;
-};
-
-struct cactus_tokenize_result {
-    std::vector<llama_token> tokens;
-    bool has_media = false;
-    std::vector<std::string> bitmap_hashes;
-    std::vector<size_t> chunk_pos;
-    std::vector<size_t> chunk_pos_media;
-};
-
-struct cactus_context {
-    bool is_predicting = false;
-    bool is_interrupted = false;
-    bool has_next_token = false;
-    std::string generated_text;
-    std::vector<completion_token_output> generated_token_probs;
-
-    size_t num_prompt_tokens = 0;
-    size_t num_tokens_predicted = 0;
-    size_t n_past = 0;
-    size_t n_remain = 0;
-
-    std::vector<llama_token> embd;
-    common_params params;
-    common_init_result llama_init;
-
-    llama_model *model = nullptr;
-    float loading_progress = 0;
-    bool is_load_interrupted = false;
-
-    llama_context *ctx = nullptr;
-    common_sampler *ctx_sampling = nullptr;
-    common_chat_templates_ptr templates;
-
-    int n_ctx;
-
-    bool truncated = false;
-    bool stopped_eos = false;
-    bool stopped_word = false;
-    bool stopped_limit = false;
-    std::string stopping_word;
-    bool incomplete = false;
-
-    std::vector<common_adapter_lora_info> lora;
-
-    bool context_full = false;
-    std::vector<llama_token> guide_tokens;
-    bool next_token_uses_guide_token = true;
-
-    struct cactus_context_mtmd {
-        mtmd_context* mtmd_ctx = nullptr;
-    };
-    cactus_context_mtmd *mtmd_wrapper = nullptr;
-    bool has_multimodal = false;
-    std::vector<std::string> mtmd_bitmap_past_hashes;
-
-    struct cactus_context_vocoder {
-        common_init_result init_result;
-        llama_model *model = nullptr;
-        llama_context *ctx = nullptr;
-        tts_type type = TTS_UNKNOWN;
-    };
-    cactus_context_vocoder *vocoder_wrapper = nullptr;
-    bool has_vocoder = false;
-    std::vector<llama_token> audio_tokens;
-
-    // Conversation management state
-    bool conversation_active = false;
-    std::string last_chat_template = "";
-
-    ~cactus_context();
-
-    void rewind();
-
-    bool initSampling();
-
-    bool loadModel(common_params &params_);
-
-    bool validateModelChatTemplate(bool use_jinja, const char *name) const;
-
-    common_chat_params getFormattedChatWithJinja(
-      const std::string &messages,
-      const std::string &chat_template,
-      const std::string &json_schema,
-      const std::string &tools,
-      const bool &parallel_tool_calls,
-      const std::string &tool_choice
-    ) const;
-
-    std::string getFormattedChat(
-      const std::string &messages,
-      const std::string &chat_template
-    ) const;
-    
-    void truncatePrompt(std::vector<llama_token> &prompt_tokens);
-
-    void loadPrompt();
-
-    void loadPrompt(const std::vector<std::string> &media_paths);
-
-    void setGuideTokens(const std::vector<llama_token> &tokens);
-   
-    void beginCompletion();
-
-    void endCompletion();
-    
-    completion_token_output nextToken();
-   
-    size_t findStoppingStrings(const std::string &text, const size_t last_token_size, const stop_type type);
-   
-    completion_token_output doCompletion();
-   
-    std::vector<float> getEmbedding(common_params &embd_params);
-    
-    std::string bench(int pp, int tg, int pl, int nr);
-   
-    int applyLoraAdapters(std::vector<common_adapter_lora_info> lora);
-   
-    void removeLoraAdapters();
-    
-    std::vector<common_adapter_lora_info> getLoadedLoraAdapters();
-
-    cactus_tokenize_result tokenize(const std::string &text, const std::vector<std::string> &media_paths);
-
-    bool initMultimodal(const std::string &mmproj_path, bool use_gpu);
-    bool isMultimodalEnabled() const;
-    bool isMultimodalSupportVision() const;
-    bool isMultimodalSupportAudio() const;
-    void releaseMultimodal();
-    void processMedia(const std::string &prompt, const std::vector<std::string> &media_paths);
-
-    bool initVocoder(const std::string &vocoder_model_path);
-    bool isVocoderEnabled() const;
-    tts_type getTTSType() const;
-    std::string getFormattedAudioCompletion(const std::string &speaker_json_str, const std::string &text_to_speak);
-    std::vector<llama_token> getAudioCompletionGuideTokens(const std::string &text_to_speak);
-    std::vector<float> decodeAudioTokens(const std::vector<llama_token> &tokens);
-    void releaseVocoder();
-};
-
-extern bool cactus_verbose;
-
-#if CACTUS_VERBOSE != 1
-#define LOG_VERBOSE(MSG, ...)
-#else
-#define LOG_VERBOSE(MSG, ...)                                       \
-    do                                                              \
-    {                                                               \
-        if (cactus_verbose)                                        \
-        {                                                           \
-            log("VERBOSE", __func__, __LINE__, MSG, ##__VA_ARGS__); \
-        }                                                           \
-    } while (0)
-#endif
-
-#define LOG_ERROR(MSG, ...) log("ERROR", __func__, __LINE__, MSG, ##__VA_ARGS__)
-
-#define LOG_WARNING(MSG, ...) log("WARNING", __func__, __LINE__, MSG, ##__VA_ARGS__)
-
-#define LOG_INFO(MSG, ...) log("INFO", __func__, __LINE__, MSG, ##__VA_ARGS__)
-
-void log(const char *level, const char *function, int line, const char *format, ...);
-
-void llama_batch_clear(llama_batch *batch);
-
-void llama_batch_add(llama_batch *batch, llama_token id, llama_pos pos, const std::vector<llama_seq_id>& seq_ids, bool logits);
-
-size_t common_part(const std::vector<llama_token> &a, const std::vector<llama_token> &b);
-
-bool ends_with(const std::string &str, const std::string &suffix);
-
-size_t find_partial_stop_string(const std::string &stop, const std::string &text);
-
-} // namespace cactus
-
-#endif /* CACTUS_H */

package/ios/cactus.xcframework/ios-arm64_x86_64-simulator/cactus.framework/Headers/cactus_ffi.h

@@ -1,256 +0,0 @@
-#ifndef CACTUS_FFI_H
-#define CACTUS_FFI_H
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <stddef.h>
-
-#if defined _WIN32 || defined __CYGWIN__
-  #ifdef CACTUS_FFI_BUILDING_DLL
-    #ifdef __GNUC__
-      #define CACTUS_FFI_EXPORT __attribute__ ((dllexport))
-    #else
-      #define CACTUS_FFI_EXPORT __declspec(dllexport)
-    #endif
-  #else
-    #ifdef __GNUC__
-      #define CACTUS_FFI_EXPORT __attribute__ ((dllimport))
-    #else
-      #define CACTUS_FFI_EXPORT __declspec(dllimport)
-    #endif
-  #endif
-  #define CACTUS_FFI_LOCAL
-#else
-  #if __GNUC__ >= 4
-    #define CACTUS_FFI_EXPORT __attribute__ ((visibility ("default")))
-    #define CACTUS_FFI_LOCAL  __attribute__ ((visibility ("hidden")))
-  #else
-    #define CACTUS_FFI_EXPORT
-    #define CACTUS_FFI_LOCAL
-  #endif
-#endif
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-typedef struct cactus_context_opaque* cactus_context_handle_t;
-
-
-typedef struct cactus_init_params_c {
-    const char* model_path;
-    const char* chat_template; 
-
-    int32_t n_ctx;
-    int32_t n_batch;
-    int32_t n_ubatch;
-    int32_t n_gpu_layers;
-    int32_t n_threads;
-    bool use_mmap;
-    bool use_mlock;
-    bool embedding; 
-    int32_t pooling_type; 
-    int32_t embd_normalize;
-    bool flash_attn;
-    const char* cache_type_k; 
-    const char* cache_type_v; 
-    void (*progress_callback)(float progress); 
-
-} cactus_init_params_c_t;
-
-typedef struct cactus_completion_params_c {
-    const char* prompt;
-    int32_t n_predict; 
-    int32_t n_threads; 
-    int32_t seed;
-    double temperature;
-    int32_t top_k;
-    double top_p;
-    double min_p;
-    double typical_p;
-    int32_t penalty_last_n;
-    double penalty_repeat;
-    double penalty_freq;
-    double penalty_present;
-    int32_t mirostat;
-    double mirostat_tau;
-    double mirostat_eta;
-    bool ignore_eos;
-    int32_t n_probs; 
-    const char** stop_sequences; 
-    int stop_sequence_count;
-    const char* grammar; 
-    bool (*token_callback)(const char* token_json);
-
-} cactus_completion_params_c_t;
-
-
-typedef struct cactus_token_array_c {
-    int32_t* tokens;
-    int32_t count;
-} cactus_token_array_c_t;
-
-typedef struct cactus_float_array_c {
-    float* values;
-    int32_t count;
-} cactus_float_array_c_t;
-
-typedef struct cactus_completion_result_c {
-    char* text; 
-    int32_t tokens_predicted;
-    int32_t tokens_evaluated;
-    bool truncated;
-    bool stopped_eos;
-    bool stopped_word;
-    bool stopped_limit;
-    char* stopping_word; 
-} cactus_completion_result_c_t;
-
-typedef struct cactus_tokenize_result_c {
-    cactus_token_array_c_t tokens;
-    bool has_media;
-    char** bitmap_hashes;
-    int bitmap_hash_count;
-    size_t* chunk_positions;
-    int chunk_position_count;
-    size_t* chunk_positions_media;
-    int chunk_position_media_count;
-} cactus_tokenize_result_c_t;
-
-CACTUS_FFI_EXPORT cactus_context_handle_t cactus_init_context_c(const cactus_init_params_c_t* params);
-
-CACTUS_FFI_EXPORT void cactus_free_context_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT int cactus_completion_c(
-    cactus_context_handle_t handle,
-    const cactus_completion_params_c_t* params,
-    cactus_completion_result_c_t* result
-);
-
-// **MULTIMODAL COMPLETION**
-CACTUS_FFI_EXPORT int cactus_multimodal_completion_c(
-    cactus_context_handle_t handle,
-    const cactus_completion_params_c_t* params,
-    const char** media_paths,
-    int media_count,
-    cactus_completion_result_c_t* result
-);
-
-CACTUS_FFI_EXPORT void cactus_stop_completion_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT cactus_token_array_c_t cactus_tokenize_c(cactus_context_handle_t handle, const char* text);
-
-CACTUS_FFI_EXPORT char* cactus_detokenize_c(cactus_context_handle_t handle, const int32_t* tokens, int32_t count);
-
-CACTUS_FFI_EXPORT cactus_float_array_c_t cactus_embedding_c(cactus_context_handle_t handle, const char* text);
-
-CACTUS_FFI_EXPORT void cactus_free_string_c(char* str);
-
-CACTUS_FFI_EXPORT void cactus_free_token_array_c(cactus_token_array_c_t arr);
-
-CACTUS_FFI_EXPORT void cactus_free_float_array_c(cactus_float_array_c_t arr);
-
-CACTUS_FFI_EXPORT void cactus_free_completion_result_members_c(cactus_completion_result_c_t* result);
-
-CACTUS_FFI_EXPORT cactus_tokenize_result_c_t cactus_tokenize_with_media_c(cactus_context_handle_t handle, const char* text, const char** media_paths, int media_count);
-
-CACTUS_FFI_EXPORT void cactus_free_tokenize_result_c(cactus_tokenize_result_c_t* result);
-
-CACTUS_FFI_EXPORT void cactus_set_guide_tokens_c(cactus_context_handle_t handle, const int32_t* tokens, int32_t count);
-
-CACTUS_FFI_EXPORT int cactus_init_multimodal_c(cactus_context_handle_t handle, const char* mmproj_path, bool use_gpu);
-
-CACTUS_FFI_EXPORT bool cactus_is_multimodal_enabled_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT bool cactus_supports_vision_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT bool cactus_supports_audio_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT void cactus_release_multimodal_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT int cactus_init_vocoder_c(cactus_context_handle_t handle, const char* vocoder_model_path);
-
-CACTUS_FFI_EXPORT bool cactus_is_vocoder_enabled_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT int cactus_get_tts_type_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT char* cactus_get_formatted_audio_completion_c(cactus_context_handle_t handle, const char* speaker_json_str, const char* text_to_speak);
-
-CACTUS_FFI_EXPORT cactus_token_array_c_t cactus_get_audio_guide_tokens_c(cactus_context_handle_t handle, const char* text_to_speak);
-
-CACTUS_FFI_EXPORT cactus_float_array_c_t cactus_decode_audio_tokens_c(cactus_context_handle_t handle, const int32_t* tokens, int32_t count);
-
-CACTUS_FFI_EXPORT void cactus_release_vocoder_c(cactus_context_handle_t handle);
-
-
-typedef struct {
-    const char* path;
-    float scale;
-} cactus_lora_adapter_c_t;
-
-typedef struct {
-    cactus_lora_adapter_c_t* adapters;
-    int32_t count;
-} cactus_lora_adapters_c_t;
-
-typedef struct {
-    char* model_name;
-    int64_t model_size;
-    int64_t model_params;
-    double pp_avg;
-    double pp_std;
-    double tg_avg;
-    double tg_std;
-} cactus_bench_result_c_t;
-
-CACTUS_FFI_EXPORT cactus_bench_result_c_t cactus_bench_c(cactus_context_handle_t handle, int pp, int tg, int pl, int nr);
-CACTUS_FFI_EXPORT int cactus_apply_lora_adapters_c(cactus_context_handle_t handle, const cactus_lora_adapters_c_t* adapters);
-CACTUS_FFI_EXPORT void cactus_remove_lora_adapters_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT cactus_lora_adapters_c_t cactus_get_loaded_lora_adapters_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT bool cactus_validate_chat_template_c(cactus_context_handle_t handle, bool use_jinja, const char* name);
-CACTUS_FFI_EXPORT char* cactus_get_formatted_chat_c(cactus_context_handle_t handle, const char* messages, const char* chat_template);
-
-typedef struct {
-    char* prompt;
-    char* json_schema;
-    char* tools;
-    char* tool_choice;
-    bool parallel_tool_calls;
-} cactus_chat_result_c_t;
-
-CACTUS_FFI_EXPORT cactus_chat_result_c_t cactus_get_formatted_chat_with_jinja_c(
-    cactus_context_handle_t handle, 
-    const char* messages,
-    const char* chat_template,
-    const char* json_schema,
-    const char* tools,
-    bool parallel_tool_calls,
-    const char* tool_choice
-);
-
-CACTUS_FFI_EXPORT void cactus_rewind_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT bool cactus_init_sampling_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT void cactus_begin_completion_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT void cactus_end_completion_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT void cactus_load_prompt_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT void cactus_load_prompt_with_media_c(cactus_context_handle_t handle, const char** media_paths, int media_count);
-
-CACTUS_FFI_EXPORT int cactus_do_completion_step_c(cactus_context_handle_t handle, char** token_text);
-CACTUS_FFI_EXPORT size_t cactus_find_stopping_strings_c(cactus_context_handle_t handle, const char* text, size_t last_token_size, int stop_type);
-
-CACTUS_FFI_EXPORT int32_t cactus_get_n_ctx_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT int32_t cactus_get_n_embd_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT char* cactus_get_model_desc_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT int64_t cactus_get_model_size_c(cactus_context_handle_t handle);
-CACTUS_FFI_EXPORT int64_t cactus_get_model_params_c(cactus_context_handle_t handle);
-
-CACTUS_FFI_EXPORT void cactus_free_bench_result_members_c(cactus_bench_result_c_t* result);
-CACTUS_FFI_EXPORT void cactus_free_lora_adapters_c(cactus_lora_adapters_c_t* adapters);
-CACTUS_FFI_EXPORT void cactus_free_chat_result_members_c(cactus_chat_result_c_t* result);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif