1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
use proc_macro::{TokenStream, TokenTree};
use quote::{format_ident, quote, ToTokens};
use syn::{parse_macro_input, Attribute, Data, DeriveInput, Expr, Ident, Lit, Meta};
/// The `arg_description` attribute is a procedural macro used to provide additional description for an enum.
///
/// This attribute does not modify the code it annotates but instead attaches metadata in the form of a description.
/// This can be helpful for better code readability and understanding the purpose of different enums.
///
/// # Usage
///
/// ```rust
/// #[arg_description(description = "This is a sample enum.", tokens = 5)]
/// #[derive(EnumDescriptor)]
/// pub enum SampleEnum {
/// Variant1,
/// Variant2,
/// }
/// ```
///
/// Note: The actual usage of the description and tokens provided through this attribute happens
/// in the `EnumDescriptor` derive macro and is retrieved in the `enum_descriptor_derive` function.
///
/// The `arg_description` attribute takes one argument, `description`, which is a string literal.
#[proc_macro_attribute]
pub fn arg_description(_args: TokenStream, input: TokenStream) -> TokenStream {
input
}
/// A derive procedural macro for the `EnumDescriptor` trait.
///
/// The `EnumDescriptor` trait should have a function `name_with_token_count`
/// that returns a tuple with the name of the enum type as a string and the
/// token count for the name as an `usize`.
///
/// This procedural macro generates an implementation of `EnumDescriptor` for
/// the type on which it's applied. The `name_with_token_count` function, in the
/// generated implementation, returns the name of the type and its token count.
///
/// # Usage
///
/// Use the `#[derive(EnumDescriptor)]` attribute on an enum to derive the
/// `EnumDescriptor` trait for it.
///
/// ```
/// #[derive(EnumDescriptor)]
/// enum MyEnum {
/// Variant1,
/// Variant2,
/// }
/// ```
///
/// This will generate:
///
/// ```
/// impl EnumDescriptor for MyEnum {
/// fn name_with_token_count() -> (String, usize) {
/// (String::from("MyEnum"), /* token count of "MyEnum" */)
/// }
/// }
/// ```
///
/// The actual token count is computed during compile time using the
/// `calculate_token_count` function.
#[proc_macro_derive(EnumDescriptor, attributes(arg_description))]
pub fn enum_descriptor_derive(input: TokenStream) -> TokenStream {
let DeriveInput { ident, attrs, .. } = parse_macro_input!(input as DeriveInput);
let name_str = format!("{}", ident);
let name_token_count = calculate_token_count(&name_str);
let mut description = String::new();
let mut desc_tokens = 0_usize;
for attr in &attrs {
if attr.path().is_ident("arg_description") {
let _result = attr.parse_nested_meta(|meta| {
let content = meta.input;
if !content.is_empty() {
if meta.path.is_ident("description") {
let value = meta.value()?;
if let Ok(Lit::Str(value)) = value.parse() {
description = value.value();
}
} else if meta.path.is_ident("tokens") {
let value = meta.value()?;
if let Ok(Lit::Int(value)) = value.parse() {
desc_tokens = value.base10_parse::<usize>()?;
return Ok(());
}
}
return Ok(());
}
Err(meta.error("unrecognized my_attribute"))
});
if _result.is_err() {
println!("Error parsing attribute: {:#?}", _result);
}
}
}
let expanded = quote! {
impl EnumDescriptor for #ident {
fn name_with_token_count() -> (String, usize) {
(String::from(#name_str), #name_token_count)
}
fn arg_description_with_token_count() -> (String, usize) {
(String::from(#description), #desc_tokens)
}
}
};
TokenStream::from(expanded)
}
/// A derive procedural macro for the `VariantDescriptors` trait.
///
/// This macro generates an implementation of the `VariantDescriptors` trait for
/// an enum. The trait provides two methods:
///
/// 1. `variant_names_with_token_counts`: Returns a `Vec` containing tuples,
/// each with a string representation of a variant's name and its token count.
///
/// 2. `variant_name_with_token_count`: Takes an enum variant as input and
/// returns a tuple with the variant's name as a string and its token count.
///
/// Note: This macro will panic if it is used on anything other than an enum.
///
/// # Usage
///
/// ```
/// #[derive(VariantDescriptors)]
/// enum MyEnum {
/// Variant1,
/// Variant2,
/// }
/// ```
///
/// This will generate the following:
///
/// ```
/// impl VariantDescriptors for MyEnum {
/// fn variant_names_with_token_counts() -> Vec<(String, usize)> {
/// vec![
/// (String::from("Variant1"), /* token count of "Variant1" */),
/// (String::from("Variant2"), /* token count of "Variant2" */),
/// ]
/// }
///
/// fn variant_name_with_token_count(&self) -> (String, usize) {
/// match self {
/// Self::Variant1 => (String::from("Variant1"), /* token count of "Variant1" */),
/// Self::Variant2 => (String::from("Variant2"), /* token count of "Variant2" */),
/// }
/// }
/// }
/// ```
///
/// The actual token count is computed during compile time using the
/// `calculate_token_count` function.
#[proc_macro_derive(VariantDescriptors)]
pub fn variant_descriptors_derive(input: TokenStream) -> TokenStream {
let ast = parse_macro_input!(input as DeriveInput);
let enum_name = &ast.ident;
let variants = if let syn::Data::Enum(ref e) = ast.data {
e.variants
.iter()
.map(|v| {
let variant_name = &v.ident;
let token_count = calculate_token_count(&variant_name.to_string());
(variant_name, token_count)
})
.collect::<Vec<_>>()
} else {
panic!("VariantDescriptors can only be used with enums");
};
let variant_names_with_token_counts: Vec<_> = variants
.iter()
.map(|(variant_name, token_count)| {
quote! { (stringify!(#variant_name).to_string(), #token_count) }
})
.collect();
let variant_name_with_token_count: Vec<_> = variants
.iter()
.map(|(variant_name, token_count)| {
quote! { Self::#variant_name => (stringify!(#variant_name).to_string(), #token_count) }
})
.collect();
let expanded = quote! {
impl VariantDescriptors for #enum_name {
fn variant_names_with_token_counts() -> Vec<(String, usize)> {
vec![
#(#variant_names_with_token_counts),*
]
}
fn variant_name_with_token_count(&self) -> (String, usize) {
match self {
#(#variant_name_with_token_count,)*
}
}
}
};
TokenStream::from(expanded)
}
/// A procedural macro to generate information about an enum.
///
/// This macro generates code that uses the `EnumDescriptor` and `VariantDescriptors`
/// traits to extract information about an enum, including its name, variant names,
/// and their corresponding token counts. Additionally, it uses the `FunctionArgument` trait
/// to fetch the argument description. All this information is serialized into JSON.
///
/// The macro returns a tuple containing the JSON and the total token count.
///
/// # Usage
///
/// The generated code will look like this:
///
/// ```rust
/// {
/// use serde_json::Value;
/// let mut total_tokens = 0;
///
/// let (arg_desc, arg_count) = <MyEnum as ::openai_func_enums::FunctionArgument>::argument_description_with_token_count();
/// total_tokens += arg_count;
///
/// // When this is consumed by the function that creates the overall function,
/// // we are going to be requiring all the arguments, which means we will repeat
/// // their names in the "required" part of openai's function schema. So we will
/// // count the tokens associated with this enum name twice here.
/// let enum_name = <MyEnum as EnumDescriptor>::name_with_token_count();
/// total_tokens += enum_name.1;
/// total_tokens += enum_name.1;
///
/// let enum_variants = <MyEnum as VariantDescriptors>::variant_names_with_token_counts();
/// total_tokens += enum_variants.iter().map(|(_, token_count)| *token_count).sum::<usize>();
///
/// let json_enum = serde_json::json!({
/// enum_name.0: {
/// "type": "string",
/// "enum": enum_variants.iter().map(|(name, _)| name.clone()).collect::<Vec<_>>(),
/// "description": arg_desc,
/// }
/// });
///
/// total_tokens += 11;
///
/// (json_enum, total_tokens)
/// }
/// ```
///
/// Note: It is assumed that the enum implements the `EnumDescriptor` and `VariantDescriptors` traits.
/// The actual token count is computed during compile time using these traits' methods.
#[proc_macro]
pub fn generate_enum_info(input: TokenStream) -> TokenStream {
let enum_ident = parse_macro_input!(input as Ident);
let output = quote! {
{
let mut total_tokens = 0;
let (arg_desc, arg_count) = <#enum_ident as EnumDescriptor>::arg_description_with_token_count();
total_tokens += arg_count;
let enum_name = <#enum_ident as EnumDescriptor>::name_with_token_count();
total_tokens += enum_name.1;
total_tokens += enum_name.1;
let enum_variants = <#enum_ident as VariantDescriptors>::variant_names_with_token_counts();
total_tokens += enum_variants.iter().map(|(_, token_count)| *token_count).sum::<usize>();
let json_enum = serde_json::json!({
enum_name.0: {
"type": "string",
"enum": enum_variants.iter().map(|(name, _)| name.clone()).collect::<Vec<_>>(),
"description": arg_desc,
}
});
total_tokens += 11;
(json_enum, total_tokens)
}
};
output.into()
}
#[proc_macro]
pub fn generate_value_arg_info(input: TokenStream) -> TokenStream {
let mut type_and_name_values = Vec::new();
let tokens = input.into_iter().collect::<Vec<TokenTree>>();
for token in tokens {
if let TokenTree::Ident(ident) = &token {
type_and_name_values.push(ident.to_string());
}
}
if type_and_name_values.len() == 2 {
let name = type_and_name_values[1].clone();
let name_tokens = calculate_token_count(name.as_str());
let type_name = type_and_name_values[0].clone();
let type_name_tokens = calculate_token_count(type_name.as_str());
let output = quote! {
{
let mut total_tokens = 0;
total_tokens += #name_tokens;
total_tokens += #type_name_tokens;
if #type_name == "array" {
let json_enum = serde_json::json!({
#name: {
"type": #type_name,
"items": {
"type": "string",
},
}
});
total_tokens += 22;
(json_enum, total_tokens)
} else {
let json_enum = serde_json::json!({
#name: {
"type": #type_name,
}
});
total_tokens += 11;
(json_enum, total_tokens)
}
}
};
return output.into();
}
let gen = quote! {};
gen.into()
}
/// This procedural macro attribute is used to specify a description for an enum variant.
///
/// The `func_description` attribute does not modify the input it is given.
/// It's only used to attach metadata (i.e., a description) to enum variants.
///
/// # Usage
///
/// ```rust
/// enum MyEnum {
/// #[func_description(description="This function does a thing.")]
/// DoAThing,
/// #[func_description(description="This function does another thing.")]
/// DoAnotherThing,
/// }
/// ```
///
/// Note: The actual usage of the description provided through this attribute happens
/// in the `FunctionCallResponse` derive macro and is retrieved in the `impl_function_call_response` function.
#[proc_macro_attribute]
pub fn func_description(_args: TokenStream, input: TokenStream) -> TokenStream {
input
}
/// This procedural macro derives the `FunctionCallResponse` trait for an enum.
///
/// The derive macro expects an enum and it generates a new struct for each variant of the enum.
/// The generated struct is named by appending "Response" to the variant's name. Each struct has the same fields as the variant.
/// Also, a `name`, `to_function_call` and `get_function_json` method is implemented for each struct.
///
/// In the `get_function_json` method, any description provided through the `func_description` attribute is used.
///
/// # Usage
///
/// ```rust
/// #[derive(FunctionCallResponse)]
/// #[func_description]
/// enum MyEnum {
/// Variant1,
/// Variant2,
/// }
/// ```
///
/// Note: This macro can only be applied to enums and it requires the `func_description` attribute to be applied to the enum.
#[proc_macro_derive(FunctionCallResponse, attributes(func_description))]
pub fn derive_function_call_response(input: TokenStream) -> TokenStream {
let ast: DeriveInput = syn::parse(input).unwrap();
let gen = impl_function_call_response(&ast);
gen.into()
}
/// This function generates a `FunctionCallResponse` implementation for each variant of an enum.
///
/// For each enum variant, it creates a new struct with the same fields as the variant and also
/// generates `name`, `to_function_call`, and `get_function_json` methods for the struct.
///
/// In the `get_function_json` method, it utilizes the description provided through the `func_description` attribute.
///
/// This function is used by the `FunctionCallResponse` derive macro.
fn impl_function_call_response(ast: &DeriveInput) -> proc_macro2::TokenStream {
match &ast.data {
Data::Enum(enum_data) => {
let mut generated_structs = Vec::new();
let mut json_generator_functions = Vec::new();
for variant in &enum_data.variants {
let variant_name = &variant.ident;
let struct_name = format_ident!("{}Response", variant_name);
let mut description = String::new();
let mut desc_tokens = 0_usize;
for attr in &variant.attrs {
if attr.path().is_ident("func_description") {
let attribute_parsed = attr.parse_nested_meta(|meta| {
let content = meta.input;
if !content.is_empty() {
if meta.path.is_ident("description") {
let value = meta.value()?;
if let Ok(Lit::Str(value)) = value.parse() {
description = value.value();
desc_tokens = calculate_token_count(description.as_str());
}
}
return Ok(());
}
Err(meta.error("unrecognized attribute"))
});
match attribute_parsed {
Ok(_attribute_parsed) => {}
Err(e) => {
println!("Error parsing attribute: {:#?}", e);
}
}
}
}
let fields: Vec<_> = variant
.fields
.iter()
.map(|f| {
let field_name =
format_ident!("{}", to_snake_case(&f.ty.to_token_stream().to_string()));
let field_type = &f.ty;
quote! {
pub #field_name: #field_type,
}
})
.collect();
let field_info: Vec<_> = variant
.fields
.iter()
.map(|f| {
let field_type = &f.ty;
quote! {
generate_enum_info!(#field_type)
}
})
.collect();
json_generator_functions.push(quote! {
impl #struct_name {
pub fn name() -> String {
stringify!(#struct_name).to_string()
}
pub fn to_function_call() -> ChatCompletionFunctionCall {
ChatCompletionFunctionCall::Function {
name: stringify!(#struct_name).to_string(),
}
}
pub fn to_tool_choice() -> ChatCompletionToolChoiceOption {
ChatCompletionToolChoiceOption::Named(ChatCompletionNamedToolChoice {
r#type: ChatCompletionToolType::Function,
function: FunctionName { name: stringify!(#struct_name).to_string() }
})
}
pub fn get_function_json() -> (serde_json::Value, usize) {
let mut parameters = serde_json::Map::new();
let mut total_tokens = 0;
for (arg_json, arg_tokens) in vec![#(#field_info),*] {
total_tokens += arg_tokens;
parameters.insert(
arg_json.as_object().unwrap().keys().next().unwrap().clone(),
arg_json
.as_object()
.unwrap()
.values()
.next()
.unwrap()
.clone(),
);
}
let function_json = json!({
"name": stringify!(#struct_name),
"description": #description,
"parameters": {
"type": "object",
"properties": parameters,
"required": parameters.keys().collect::<Vec<_>>()
}
});
total_tokens += 12;
total_tokens += #desc_tokens;
(function_json, total_tokens)
}
}
});
generated_structs.push(quote! {
#[derive(serde::Deserialize, Debug)]
#[serde(rename_all = "PascalCase")]
pub struct #struct_name {
#(#fields)*
}
});
}
let gen = quote! {
use async_openai::types::{ChatCompletionFunctionCall, ChatCompletionNamedToolChoice, ChatCompletionToolChoiceOption, FunctionName,};
use serde_json::{json, Value};
use openai_func_enums::generate_enum_info;
#(#generated_structs)*
#(#json_generator_functions)*
};
gen
}
_ => panic!("FunctionCallResponse can only be derived for enums"),
}
}
/// The `SubcommandGPT` procedural macro is used to derive a structure
/// which encapsulates various chat completion commands.
///
/// This macro should be applied to an enum. It generates various supporting
/// structures and methods, including structures representing the command arguments,
/// methods for converting between the argument structures and the original enum,
/// JSON conversion methods, and an implementation of the original enum that provides
/// methods for executing the commands and dealing with the responses.
///
/// Each variant of the original enum will be converted into a corresponding structure,
/// and each field in the variant will become a field in the generated structure.
/// The generated structures will derive `serde::Deserialize` and `Debug` automatically.
///
/// This macro also generates methods for calculating the token count of a string and
/// for executing commands based on function calls received from the chat API.
///
/// The types of fields in the enum variants determine how the corresponding fields in the
/// generated structures are treated. For example, fields of type `String` or `&str` are
/// converted to JSON value arguments with type `"string"`, while fields of type `u8`, `u16`,
/// `u32`, `u64`, `usize`, `i8`, `i16`, `i32`, `i64`, `isize`, `f32` or `f64` are converted
/// to JSON value arguments with type `"integer"` or `"number"` respectively.
/// For fields with a tuple type, currently this macro simply prints that the field is of a tuple type.
/// For fields with an array type, they are converted to JSON value arguments with type `"array"`.
///
/// When running the chat command, a custom system message can be optionally provided.
/// If provided, this message will be used as the system message in the chat request.
/// If not provided, a default system message will be used.
///
/// If the total token count of the request exceeds a specified limit, an error will be returned.
///
/// The `derive_subcommand_gpt` function consumes a `TokenStream` representing the enum
/// to which the macro is applied and produces a `TokenStream` representing the generated code.
///
/// # Panics
/// This macro will panic (only at compile time) if it is applied to a non-enum item.
#[proc_macro_derive(SubcommandGPT)]
pub fn derive_subcommand_gpt(input: TokenStream) -> TokenStream {
let input = parse_macro_input!(input as DeriveInput);
let name = input.ident;
let data = match input.data {
Data::Enum(data) => data,
_ => panic!("SubcommandGPT can only be implemented for enums"),
};
let mut generated_structs = Vec::new();
let mut json_generator_functions = Vec::new();
let mut generated_clap_gpt_enum = Vec::new();
let mut generated_struct_names = Vec::new();
for variant in data.variants.iter() {
let variant_name = &variant.ident;
let struct_name = format_ident!("{}", variant_name);
generated_struct_names.push(struct_name.clone());
let mut variant_desc = String::new();
let mut variant_desc_tokens = 0_usize;
// let description = get_comment_from_attr(&variant_attrs);
for variant_attrs in &variant.attrs {
let description = get_comment_from_attr(variant_attrs);
if let Some(description) = description {
variant_desc = description;
variant_desc_tokens = calculate_token_count(variant_desc.as_str());
}
}
let fields: Vec<_> = variant
.fields
.iter()
.map(|f| {
// If the field has an identifier (i.e., it is a named field),
// use it. Otherwise, use the type as the name.
let field_name = if let Some(ident) = &f.ident {
format_ident!("{}", ident)
} else {
format_ident!("{}", to_snake_case(&f.ty.to_token_stream().to_string()))
};
let field_type = &f.ty;
quote! {
pub #field_name: #field_type,
}
})
.collect();
let execute_command_parameters: Vec<_> = variant
.fields
.iter()
.map(|field| {
let field_name = &field.ident;
quote! { #field_name: self.#field_name.clone() }
})
.collect();
let number_type = "number";
let number_ident = format_ident!("{}", number_type);
let integer_type = "integer";
let integer_ident = format_ident!("{}", integer_type);
let string_type = "string";
let string_ident = format_ident!("{}", string_type);
let array_type = "array";
let array_ident = format_ident!("{}", array_type);
let field_info: Vec<_> = variant
.fields
.iter()
.map(|f| {
let field_name = if let Some(ident) = &f.ident {
format_ident!("{}", ident)
} else {
format_ident!("{}", to_snake_case(&f.ty.to_token_stream().to_string()))
};
let field_type = &f.ty;
match field_type {
syn::Type::Path(typepath) if typepath.qself.is_none() => {
let type_ident = &typepath.path.segments.last().unwrap().ident;
match type_ident.to_string().as_str() {
"f32" | "f64" => {
return quote! {
generate_value_arg_info!(#number_ident, #field_name)
};
}
"u8" | "u16" | "u32" | "u64" | "u128" | "usize" | "i8" | "i16"
| "i32" | "i64" | "i128" | "isize" => {
return quote! {
generate_value_arg_info!(#integer_ident, #field_name)
};
}
"String" | "&str" => {
return quote! {
generate_value_arg_info!(#string_ident, #field_name)
};
}
"Vec" => {
return quote! {
generate_value_arg_info!(#array_ident, #field_name)
};
}
_ => {
return quote! {
generate_enum_info!(#field_type)
};
}
}
}
syn::Type::Tuple(_) => {
println!("Field {} is of tuple type", field_name);
}
syn::Type::Array(_) => {
println!("Field {} is of array type", field_name);
return quote! {
generate_value_arg_info!(#array_ident, #field_name)
};
}
_ => {
println!("Field {} is of another type.", field_name);
}
}
quote! {}
})
.collect();
json_generator_functions.push(quote! {
impl #struct_name {
pub fn name() -> String {
stringify!(#struct_name).to_string()
}
pub fn to_function_call() -> ChatCompletionFunctionCall {
ChatCompletionFunctionCall::Function {
name: stringify!(#struct_name).to_string(),
}
}
pub fn to_tool_choice() -> ChatCompletionToolChoiceOption {
ChatCompletionToolChoiceOption::Named(ChatCompletionNamedToolChoice {
r#type: ChatCompletionToolType::Function,
function: FunctionName { name: stringify!(#struct_name).to_string() }
})
}
pub fn execute_command(&self) -> #name {
#name::#variant_name {
#(#execute_command_parameters),*
}
}
pub fn get_function_json() -> (serde_json::Value, usize) {
let mut parameters = serde_json::Map::new();
let mut total_tokens = 0;
for (arg_json, arg_tokens) in vec![#(#field_info),*] {
total_tokens += arg_tokens;
parameters.insert(
arg_json.as_object().unwrap().keys().next().unwrap().clone(),
arg_json
.as_object()
.unwrap()
.values()
.next()
.unwrap()
.clone(),
);
}
let function_json = serde_json::json!({
"name": stringify!(#struct_name),
"description": #variant_desc,
"parameters": {
"type": "object",
"properties": parameters,
"required": parameters.keys().collect::<Vec<_>>()
}
});
total_tokens += 12;
total_tokens += #variant_desc_tokens;
(function_json, total_tokens)
}
}
});
generated_structs.push(quote! {
#[derive(Clone, serde::Deserialize, Debug)]
pub struct #struct_name {
#(#fields)*
}
});
}
let all_function_calls = quote! {
pub fn all_function_jsons() -> (serde_json::Value, usize) {
let results = vec![#(#generated_struct_names::get_function_json(),)*];
let combined_json = serde_json::Value::Array(results.iter().map(|(json, _)| json.clone()).collect());
let total_tokens = results.iter().map(|(_, tokens)| tokens).sum();
(combined_json, total_tokens)
}
};
generated_clap_gpt_enum.push(quote! {
#[derive(Subcommand)]
pub enum CommandsGPT {
GPT { a: String },
}
});
let struct_names: Vec<String> = generated_struct_names
.iter()
.map(|name| format!("{}", name))
.collect();
let match_arms: Vec<_> = generated_struct_names
.iter()
.map(|struct_name| {
let response_name = format_ident!("{}", struct_name);
quote! {
Ok(FunctionResponse::#response_name(response)) => {
let result = response.execute_command();
let run_result = result.run(execution_strategy_clone).await;
match run_result {
Ok(run_result) => {
{
let prior_result_clone = prior_result.clone();
let mut prior_result_lock = prior_result_clone.lock().await;
*prior_result_lock = run_result;
}
return Ok(());
}
Err(e) => {
println!("{:#?}", e);
}
}
}
}
})
.collect();
let match_arms_no_return: Vec<_> = generated_struct_names
.iter()
.map(|struct_name| {
let response_name = format_ident!("{}", struct_name);
quote! {
Ok(FunctionResponse::#response_name(response)) => {
let result = response.execute_command();
let run_result = result.run(execution_strategy_clone).await;
match run_result {
Ok(run_result) => {
{
let mut prior_result_lock = prior_result_clone.lock().await;
*prior_result_lock = run_result;
}
}
Err(e) => {
println!("{:#?}", e);
}
}
}
}
})
.collect();
let commands_gpt_impl = quote! {
#[derive(Clone, Debug, Deserialize)]
pub enum FunctionResponse {
#(
#generated_struct_names(#generated_struct_names),
)*
}
impl CommandsGPT {
#all_function_calls
fn to_snake_case(camel_case: &str) -> String {
let mut snake_case = String::new();
for (i, ch) in camel_case.char_indices() {
if i > 0 && ch.is_uppercase() {
snake_case.push('_');
}
snake_case.extend(ch.to_lowercase());
}
snake_case
}
pub fn parse_gpt_function_call(function_call: &FunctionCall) -> Result<FunctionResponse, Box<dyn std::error::Error + Send + Sync + 'static>> {
match function_call.name.as_str() {
#(
#struct_names => {
match serde_json::from_str::<#generated_struct_names>(&function_call.arguments) {
Ok(arguments) => Ok(FunctionResponse::#generated_struct_names(arguments)),
Err(_) => {
let snake_case_args = function_call.arguments
.as_str()
.split(',')
.map(|s| {
let mut parts = s.split(':');
match (parts.next(), parts.next()) {
(Some(key), Some(value)) => {
let key_trimmed = key.trim_matches(|c: char| !c.is_alphanumeric()).trim();
let key_snake_case = Self::to_snake_case(key_trimmed);
format!("\"{}\":{}", key_snake_case, value)
},
_ => s.to_string()
}
})
.collect::<Vec<String>>()
.join(",");
let snake_case_args = format!("{{{}", snake_case_args);
match serde_json::from_str::<#generated_struct_names>(&snake_case_args) {
Ok(arguments) => {
Ok(FunctionResponse::#generated_struct_names(arguments))
}
Err(e) => {
Err(Box::new(CommandError::new("There was an issue deserializing function arguments.")))
}
}
}
}
},
)*
_ => {
println!("{:#?}", function_call);
Err(Box::new(CommandError::new("Unknown function name")))
}
}
}
fn calculate_token_count(text: &str) -> usize {
let bpe = tiktoken_rs::cl100k_base().unwrap();
bpe.encode_ordinary(&text).len()
}
pub async fn run(
prompt: &String,
model_name: &str,
request_token_limit: usize,
max_response_tokens: u16,
custom_system_message: Option<String>,
prior_result: Arc<Mutex<Option<String>>>,
execution_strategy: ToolCallExecutionStrategy,
) -> Result<(), Box<dyn std::error::Error + Send + Sync + 'static>> {
let tool_args = get_tool_chat_completion_args(CommandsGPT::all_function_jsons)?;
let mut system_message_tokens = 7;
let mut system_message = String::from("You are a helpful function calling bot.");
if let Some(custom_system_message) = custom_system_message {
system_message = custom_system_message;
system_message_tokens = Self::calculate_token_count(system_message.as_str());
}
let request_token_total = tool_args.1 + system_message_tokens + Self::calculate_token_count(prompt.as_str());
if request_token_total > request_token_limit {
return Err(Box::new(CommandError::new("Request token count is too high")));
}
let request = CreateChatCompletionRequestArgs::default()
.max_tokens(max_response_tokens)
.model(model_name)
// .temperature(0.0)
.messages([ChatCompletionRequestMessage::System(ChatCompletionRequestSystemMessageArgs::default()
.content(system_message)
.build()?),
ChatCompletionRequestMessage::User(ChatCompletionRequestUserMessageArgs::default()
.content(prompt.to_string())
.build()?)])
.tools(tool_args.0)
// What is going on here? Fails on their end if this is included.
// .tool_choice("auto")
.build()?;
let client = Client::new();
let response_message = client
.chat()
.create(request)
.await?
.choices
.get(0)
.unwrap()
.message
.clone();
if let Some(tool_calls) = response_message.tool_calls {
if tool_calls.len() == 1 {
let execution_strategy_clone = execution_strategy.clone();
match Self::parse_gpt_function_call(&tool_calls.first().unwrap().function) {
#(#match_arms,)*
Err(e) => {
println!("{:#?}", e);
return Err(Box::new(CommandError::new("Error running GPT command")));
}
};
} else {
// TODO: FML. Can't get feature flags to get passed through from consuming
// code. Compilation context issue.
// #[cfg(feature = "verbose")]
// {
// println!("Called {} tools. Execution strategy set to {:#?}.", tool_calls.len(), execution_strategy);
// println!();
//
// }
println!("Called {} tools. Execution strategy set to {:#?}.", tool_calls.len(), execution_strategy);
println!();
match execution_strategy {
ToolCallExecutionStrategy::Async => {
let mut tasks = Vec::new();
for tool_call in tool_calls.iter() {
match tool_call.r#type {
ChatCompletionToolType::Function => {
let function = tool_call.function.clone();
let prior_result_clone = prior_result.clone();
let execution_strategy_clone = execution_strategy.clone();
let task = tokio::spawn( async move {
match Self::parse_gpt_function_call(&function) {
#(#match_arms_no_return,)*
Err(e) => {
println!("{:#?}", e);
}
}
});
tasks.push(task);
},
}
}
for task in tasks {
let _ = task.await;
}
},
ToolCallExecutionStrategy::Synchronous => {
for tool_call in tool_calls.iter() {
match tool_call.r#type {
ChatCompletionToolType::Function => {
let prior_result_clone = prior_result.clone();
let execution_strategy_clone = execution_strategy.clone();
match Self::parse_gpt_function_call(&tool_call.function) {
#(#match_arms_no_return,)*
Err(e) => {
println!("{:#?}", e);
}
}
},
}
}
},
ToolCallExecutionStrategy::Parallel => {
let mut handles = Vec::new();
for tool_call in tool_calls.iter() {
match tool_call.r#type {
ChatCompletionToolType::Function => {
let function = tool_call.function.clone();
let prior_result_clone = prior_result.clone();
// let execution_strategy_clone = execution_strategy.clone();
// TODO: Think through. There's a lot of overhead to
// make os threads this way. For now assume that if
// strategy is set to "Parallel" that we only want to
// put the intially returned tool calls on threads, and
// if they themselves contain something multi-step we
// will run those as if they are io-bound. Potentially
// makes sense to support letting variants get
// decorated with a execution strategy preference like
// "this is io bound" or "this is cpu bound" and then
let execution_strategy_clone = ToolCallExecutionStrategy::Async;
let handle = std::thread::spawn(move || {
let rt = tokio::runtime::Runtime::new().unwrap();
rt.block_on(async {
match Self::parse_gpt_function_call(&function) {
#(#match_arms_no_return,)*
Err(e) => {
println!("{:#?}", e);
}
}
})
});
handles.push(handle);
},
}
}
for handle in handles {
let _ = handle.join();
}
},
}
}
Ok(())
} else {
return Ok(());
}
}
}
};
let gen = quote! {
use serde::Deserialize;
use serde_json::{json, Value};
use openai_func_enums::{
generate_enum_info, generate_value_arg_info, get_tool_chat_completion_args,
};
use async_trait::async_trait;
use async_openai::{
types::{
ChatCompletionFunctionCall, ChatCompletionNamedToolChoice, ChatCompletionRequestMessage,
ChatCompletionRequestSystemMessageArgs, ChatCompletionRequestUserMessageArgs,
ChatCompletionToolChoiceOption, ChatCompletionToolType, CreateChatCompletionRequestArgs,
FunctionCall, FunctionName,
},
Client,
};
#(#generated_structs)*
#(#json_generator_functions)*
#(#generated_clap_gpt_enum)*
#commands_gpt_impl
};
gen.into()
}
fn get_comment_from_attr(attr: &Attribute) -> Option<String> {
if attr.path().is_ident("doc") {
if let Meta::NameValue(meta) = &attr.meta {
if meta.path.is_ident("doc") {
let value = meta.value.clone();
match value {
Expr::Lit(value) => match value.lit {
Lit::Str(value) => {
return Some(value.value());
}
_ => {
return None;
}
},
_ => {
return None;
}
}
}
}
}
None
}
/// Calculate the token count of a given text string using the Byte Pair Encoding (BPE) tokenizer.
///
/// This function utilizes the BPE tokenizer from the `cl100k_base` library. It tokenizes the given text and
/// returns the count of the tokens. This can be used to measure how many tokens a particular text string
/// consumes, which is often relevant in the context of natural language processing tasks.
///
/// # Arguments
///
/// * `text` - A string slice that holds the text to tokenize.
///
/// # Returns
///
/// * `usize` - The count of tokens in the text.
///
/// # Example
///
/// ```
/// let text = "Hello, world!";
/// let token_count = calculate_token_count(text);
/// println!("Token count: {}", token_count);
/// ```
///
/// Note: This function can fail if the `cl100k_base` tokenizer is not properly initialized or the text cannot be tokenized.
fn calculate_token_count(text: &str) -> usize {
let bpe = tiktoken_rs::cl100k_base().unwrap();
bpe.encode_ordinary(text).len()
}
/// Convert a camelCase or PascalCase string into a snake_case string.
///
/// This function iterates over each character in the input string. If the character is an uppercase letter, it adds an
/// underscore before it (except if it's the first character) and then appends the lowercase version of the character
/// to the output string.
///
/// # Arguments
///
/// * `camel_case` - A string slice that holds the camelCase or PascalCase string to convert.
///
/// # Returns
///
/// * `String` - The converted snake_case string.
///
/// # Example
///
/// ```
/// let camel_case = "HelloWorld";
/// let snake_case = to_snake_case(camel_case);
/// assert_eq!(snake_case, "hello_world");
/// ```
fn to_snake_case(camel_case: &str) -> String {
let mut snake_case = String::new();
for (i, ch) in camel_case.char_indices() {
if i > 0 && ch.is_uppercase() {
snake_case.push('_');
}
snake_case.extend(ch.to_lowercase());
}
snake_case
}