@samchon/openapi/lib/structures/IMcpLlmFunction.d.mts

OpenAPI definitions and converters for 'typia' and 'nestia'.

import { ILlmSchema } from "./ILlmSchema";
import { IValidation } from "./IValidation";
/**
 * MCP LLM function.
 *
 * `IMcpLlmFunction` is an interface representing a function metadata
 * of MCP (Model Context Protocol), which has been designed for the LLM
 * (Language Large Model) function calling, instead of providing
 * [`mcp_servers`](https://openai.github.io/openai-agents-python/mcp/#using-mcp-servers)
 * property to the LLM vendor API.
 *
 * It contains the target function's metadata {@link name}, {@link parameters}
 * and {@link description}. You can put these properties to the LLM vendor API
 * to let the LLM know the function's purpose and how to call it.
 *
 * Also, it contains {@link validate} function, which is used to validate the
 * function arguments generated by AI agent. It is very important to the
 * LLM function calling, because LLM like OpenAI takes a lot of mistakes when
 * composing arguments in function calling. Even though `number` like
 * simple type is defined in the {@link parameters} schema, LLM often
 * fills it just by a `string` typed value.
 *
 * > This is the reason why `@samchon/openapi` recommends not to use the
 * > [`mcp_servers`](https://openai.github.io/openai-agents-python/mcp/#using-mcp-servers)
 * > property of LLM API directly, but to use the function calling feature
 * > instead. You have to make the AI agent to correct the type level mistakes
 * > by using the `validate` function.
 *
 * @author Jeongho Nam - https://github.com/samchon
 * @author Byeongjin Oh - https://github.com/sunrabbit123
 */
export interface IMcpLlmFunction<Model extends ILlmSchema.Model> {
    /**
     * Representative name of the function.
     *
     * @maxLength 64
     */
    name: string;
    /**
     * Description of the function.
     *
     * For reference, the `description` is very important property to teach
     * the purpose of the function to the LLM (Language Large Model), and
     * LLM actually determines which function to call by the description.
     *
     * Also, when the LLM conversates with the user, the `description` is
     * used to explain the function to the user. Therefore, the `description`
     * property has the highest priority, and you have to consider it.
     */
    description?: string | undefined;
    /**
     * Parameters of the function.
     */
    parameters: ILlmSchema.IParameters<Model>;
    /**
     * Validate function of the arguments.
     *
     * You know what? LLM (Large Language Model) like OpenAI takes a lot of
     * mistakes when composing arguments in function calling. Even though
     * `number` like simple type is defined in the {@link parameters} schema,
     * LLM often fills it just by a `string` typed value.
     *
     * In that case, you have to give a validation feedback to the LLM by
     * using this `validate` function. The `validate` function will return
     * detailed information about every type errors about the arguments.
     *
     * And in my experience, OpenAI's `gpt-4o-mini` model tends to construct
     * an invalid function calling arguments at the first trial about 50% of
     * the time. However, if correct it through this `validate` function,
     * the success rate soars to 99% at the second trial, and I've never
     * failed at the third trial.
     *
     * > If you've {@link separated} parameters, use the
     * > {@link IMcpLlmFunction.ISeparated.validate} function instead when
     * > validating the LLM composed arguments.
     * >
     * > In that case, This `validate` function would be meaningful only
     * > when you've merged the LLM and human composed arguments by
     * > {@link McpLlm.mergeParameters} function.
     *
     * @param args Arguments to validate
     * @returns Validation result
     */
    validate: (args: unknown) => IValidation<unknown>;
}