@samchon/openapi
Version:
OpenAPI definitions and converters for 'typia' and 'nestia'.
85 lines (84 loc) • 4.61 kB
text/typescript
import { ILlmSchemaV3_1 } from "./ILlmSchemaV3_1";
/**
* Type schema info of the Llama.
*
* `ILlamaSchema` is a type schema info of the Llama function calling.
*
* `ILlamaSchema` seems fully supporting the JSON schema definition of the OpenAPI v3.1
* specification; {@link OpenApiV3_1.IJsonSchema}. By the way, as the
* {@link OpenApiV3_1.IJsonSchema} has too much ambiguous and duplicated expressions,
* `ILlamaSchema` is designed to be clear and simple for the Llama function calling,
* by utilizng {@link ILlmSchemaV3_1} which has been transformed from the
* {@link OpenApi.IJsonSchema} for the convenience and clarity.
*
* Therefore, `ILlamaSchema` does not follow the entire specification of
* the OpenAPI v3.1. It has own specific restrictions and definitions. Here is the
* list of how `ILlmSchemaV3_1` is different with the OpenAPI v3.1 JSON schema.
*
* - Decompose mixed type: {@link OpenApiV3_1.IJsonSchema.IMixed}
* - Resolve nullable property: {@link OpenApiV3_1.IJsonSchema.__ISignificant.nullable}
* - Tuple type is banned: {@link OpenApiV3_1.IJsonSchema.ITuple.prefixItems}
* - Constant type is banned: {@link OpenApiV3_1.IJsonSchema.IConstant}
* - Merge {@link OpenApiV3_1.IJsonSchema.IAnyOf} to {@link ILlamaSchema.IOneOf}
* - Merge {@link OpenApiV3_1.IJsonSchema.IAllOf} to {@link ILlamaSchema.IObject}
* - Merge {@link OpenApiV3_1.IJsonSchema.IRecursiveReference} to {@link ILlamaSchema.IReference}
* - Do not support {@link OpenApiV3_1.IJsonSchema.ITuple} type
*
* If compare with the {@link OpenApi.IJsonSchema}, the emended JSON schema specification,
*
* - {@link ILlamaSchema.IParameters.$defs} instead of the {@link OpenApi.IJsonSchema.schemas}
* - Do not support {@link OpenApi.IJsonSchema.ITuple} type
* - {@link ILlamaSchema.properties} and {@link ILlamaSchema.required} are always defined
*
* For reference, if you've composed the `ILlamaSchema` type with the
* {@link ILlamaSchema.IConfig.reference} `false` option (default is `false`), only the
* recursive named types would be archived into the {@link ILlamaSchema.IParameters.$defs},
* and the others would be escaped from the {@link ILlamaSchema.IReference} type.
*
* @reference https://www.llama.com/docs/model-cards-and-prompt-formats/llama3_2
* @reference https://github.com/ggerganov/llama.cpp/blob/master/examples/json_schema_to_grammar.py
* @warning Specified not only by the official documentation, but also by my
* experiments. Therefore, its definitions can be inaccurate or be
* changed in the future. If you find any wrong or outdated definitions,
* please let me know by issue.
* @author Jeongho Nam - https://github.com/samchon
*/
export type ILlamaSchema = ILlamaSchema.IConstant | ILlamaSchema.IBoolean | ILlamaSchema.IInteger | ILlamaSchema.INumber | ILlamaSchema.IString | ILlamaSchema.IArray | ILlamaSchema.IObject | ILlamaSchema.IReference | ILlamaSchema.IOneOf | ILlamaSchema.INull | ILlamaSchema.IUnknown;
export declare namespace ILlamaSchema {
export import IParameters = ILlmSchemaV3_1.IParameters;
export import IConstant = ILlmSchemaV3_1.IConstant;
export import IBoolean = ILlmSchemaV3_1.IBoolean;
export import IInteger = ILlmSchemaV3_1.IInteger;
export import INumber = ILlmSchemaV3_1.INumber;
export import IString = ILlmSchemaV3_1.IString;
export import IObject = ILlmSchemaV3_1.IObject;
export import IArray = ILlmSchemaV3_1.IArray;
export import IReference = ILlmSchemaV3_1.IReference;
export import IOneOf = ILlmSchemaV3_1.IOneOf;
export import INull = ILlmSchemaV3_1.INull;
export import IUnknown = ILlmSchemaV3_1.IUnknown;
/**
* Configuration for Llama schema composition.
*/
interface IConfig {
/**
* Whether to allow reference type in everywhere.
*
* If you configure this property to `false`, most of reference types
* represented by {@link ILlamaSchema.IReference} would be escaped to
* a plain type unless recursive type case.
*
* This is because some low sized LLM models does not understand the
* reference type well, and even the large size LLM models sometimes occur
* the hallucination.
*
* However, the reference type makes the schema size smaller, so that reduces
* the LLM token cost. Therefore, if you're using the large size of LLM model,
* and want to reduce the LLM token cost, you can configure this property to
* `true`.
*
* @default false
*/
reference: boolean;
}
}