llama_cpp_bindings/
lib.rs

1//! Bindings to the llama.cpp library.
2//!
3//! As llama.cpp is a very fast moving target, this crate does not attempt to create a stable API
4//! with all the rust idioms. Instead it provided safe wrappers around nearly direct bindings to
5//! llama.cpp. This makes it easier to keep up with the changes in llama.cpp, but does mean that
6//! the API is not as nice as it could be.
7//!
8//! # Feature Flags
9//!
10//! - `cuda` enables CUDA gpu support.
11
12pub mod batch_add_error;
13pub mod chat_message_parse_outcome;
14pub mod context;
15pub mod error;
16pub mod extract_tool_call_markers_from_haystack;
17pub mod ffi_error_reader;
18pub mod ffi_status_is_ok;
19pub mod ffi_status_to_i32;
20pub mod ggml_time_us;
21pub mod gguf_context;
22pub mod gguf_context_error;
23pub mod gguf_type;
24pub mod ingest_outcome;
25pub mod ingest_prompt_chunk;
26pub mod invalid_numa_strategy;
27pub mod json_schema_to_grammar;
28pub mod llama_backend;
29pub mod llama_backend_device;
30pub mod llama_backend_device_type;
31pub mod llama_backend_numa_strategy;
32pub mod llama_batch;
33pub mod llama_time_us;
34pub mod llama_token_attr;
35pub mod llama_token_attrs;
36pub mod llama_token_attrs_from_int_error;
37pub mod llguidance_sampler;
38#[cfg(feature = "dynamic-backends")]
39pub mod load_backends;
40#[cfg(feature = "dynamic-backends")]
41pub mod load_backends_error;
42#[cfg(feature = "dynamic-backends")]
43pub mod load_backends_from_path;
44pub mod log_options;
45pub mod max_devices;
46pub mod mlock_supported;
47pub mod mmap_supported;
48pub mod model;
49pub mod mtmd;
50pub mod raw_chat_message;
51pub mod resolved_tool_call_markers;
52pub mod sampled_token;
53pub mod sampled_token_classifier;
54pub mod sampled_token_section;
55pub mod sampling;
56pub mod send_logs_to_log;
57pub mod streaming_json_probe;
58pub mod streaming_markers;
59pub mod timing;
60pub mod token;
61pub mod tool_call_format;
62pub mod tool_call_marker_pair;
63pub mod tool_call_template_overrides;
64
65pub use error::{
66    ApplyChatTemplateError, ChatTemplateError, DecodeError, EmbeddingsError, EncodeError,
67    EvalMultimodalChunksError, GrammarError, JsonSchemaToGrammarError, KvCacheSeqAddError,
68    KvCacheSeqDivError, LlamaContextLoadError, LlamaCppError, LlamaLoraAdapterInitError,
69    LlamaLoraAdapterRemoveError, LlamaLoraAdapterSetError, LlamaModelLoadError, LogitsError,
70    MarkerDetectionError, MetaValError, ModelParamsError, NewLlamaChatMessageError,
71    ParseChatMessageError, Result, SampleError, SamplerAcceptError, SamplingError,
72    StringToTokenError, TokenSamplingError, TokenToStringError,
73};
74
75pub use chat_message_parse_outcome::ChatMessageParseOutcome;
76pub use llama_backend_device::{LlamaBackendDevice, list_llama_ggml_backend_devices};
77pub use llama_backend_device_type::LlamaBackendDeviceType;
78pub use llama_cpp_bindings_types::{
79    BracketedJsonShape, KeyValueXmlTagsShape, PairedQuoteShape, ParsedChatMessage, ParsedToolCall,
80    ReasoningMarkers, TokenUsage, TokenUsageError, ToolCallArgsShape, ToolCallArguments,
81    ToolCallMarkers, ToolCallValueQuote, XmlTagsShape,
82};
83pub use raw_chat_message::RawChatMessage;
84pub use sampled_token::SampledToken;
85pub use sampled_token_classifier::SampledTokenClassifier;
86pub use sampled_token_section::SampledTokenSection;
87
88pub use ffi_status_is_ok::status_is_ok;
89pub use ffi_status_to_i32::status_to_i32;
90pub use ggml_time_us::ggml_time_us;
91pub use ingest_prompt_chunk::ingest_prompt_chunk;
92pub use json_schema_to_grammar::json_schema_to_grammar;
93pub use llama_time_us::llama_time_us;
94pub use max_devices::max_devices;
95pub use mlock_supported::mlock_supported;
96pub use mmap_supported::mmap_supported;
97
98pub use log_options::LogOptions;
99pub use send_logs_to_log::send_logs_to_log;
llama_cpp_bindings/lib.rs

llama_cpp_bindings/
lib.rs
