API: @loopstack/typesafe-module

Classes

TypeSafeClientService

Creates TypeSafe SDK clients with the API key read from the environment.

Inject it to call systemOne() directly when you want answer types inferred from your own question literals.

import { TypeSafeClientService } from '@loopstack/typesafe-module';

Provided by: TypeSafeModule

export class TypeSafeClientService {
  getClient(config?: TypeSafeModelConfig): TypeSafeClient;
  systemOne<const Q extends Questions>(
    request: SystemOneRequest<Q>,
    options?: { envApiKey?: string; signal?: AbortSignal },
  ): Promise<SystemOneResult<Q>>;
}

TypeSafeModule

NestJS module that provides the TypeSafeSystemOneTool (typesafe_system_one) and the TypeSafeClientService for typed yes/no, choice and score decisions with TypeSafe AI.

Registration:

• TypeSafeModule — bare import is all that is needed; there are no static methods.

Requires: a TypeSafe API key, read from the TYPESAFE_API_KEY env var by default (override the env var name per call via envApiKey). The model defaults to TYPESAFE_DEFAULT_MODEL, then jev-latest.

import { TypeSafeModule } from '@loopstack/typesafe-module';

export class TypeSafeModule {}

TypeSafeSystemOneTool

Tool that answers named yes/no (noul), choice and score questions about a state with TypeSafe AI.

Build questions with the noul, choice and score helpers exported by this package. The model and API key env var come from options.config. Returns TypeSafeSystemOneResult with the answers keyed by question name, and TypeSafeResultMeta with the model and token usage.

import { TypeSafeSystemOneTool } from '@loopstack/typesafe-module';

Provided by: TypeSafeModule

export class TypeSafeSystemOneTool extends BaseTool<
  TypeSafeSystemOneArgs,
  TypeSafeSystemOneConfig,
  TypeSafeSystemOneResult,
  TypeSafeResultMeta
> {
  protected handle(
    args: TypeSafeSystemOneArgs,
    ctx: RunContext,
    options?: ToolCallOptions<TypeSafeSystemOneConfig>,
  ): Promise<ToolEnvelope<TypeSafeSystemOneResult, TypeSafeResultMeta>>;
}

Interfaces

TypeSafeModelConfig

Per-call client settings for TypeSafeClientService.

import { TypeSafeModelConfig } from '@loopstack/typesafe-module';

export interface TypeSafeModelConfig {
  /** Model name (e.g. 'jev-latest'). Falls back to the SDK default: TYPESAFE_DEFAULT_MODEL, then 'jev-latest'. */
  model?: string;
  /** Environment variable name containing the API key. Falls back to TYPESAFE_API_KEY. */
  envApiKey?: string;
}

Type Aliases

TypeSafeAnswer

A TypeSafe answer (noul, choice or score).

import { TypeSafeAnswer } from '@loopstack/typesafe-module';

export type TypeSafeAnswer = z.infer<typeof TypeSafeAnswerSchema>;

TypeSafeQuestion

A TypeSafe question (noul, choice or score).

import { TypeSafeQuestion } from '@loopstack/typesafe-module';

export type TypeSafeQuestion = z.infer<typeof TypeSafeQuestionSchema>;

TypeSafeResultMeta

Metadata returned by typesafe_system_one: provider typesafe, the model that answered, and token usage.

import { TypeSafeResultMeta } from '@loopstack/typesafe-module';

export type TypeSafeResultMeta = {
  provider: 'typesafe';
  model: string;
  usage: {
    inputTokens: number;
    outputTokens: number;
  };
};

TypeSafeSystemOneArgs

Args for TypeSafeSystemOneTool.

import { TypeSafeSystemOneArgs } from '@loopstack/typesafe-module';

export type TypeSafeSystemOneArgs = z.infer<typeof TypeSafeSystemOneArgsSchema>;

TypeSafeSystemOneConfig

Config for TypeSafeSystemOneTool.

import { TypeSafeSystemOneConfig } from '@loopstack/typesafe-module';

export type TypeSafeSystemOneConfig = z.infer<typeof TypeSafeSystemOneConfigSchema>;

TypeSafeSystemOneResult

Result of the typesafe_system_one tool.

import { TypeSafeSystemOneResult } from '@loopstack/typesafe-module';

export type TypeSafeSystemOneResult = z.infer<typeof TypeSafeSystemOneResultSchema>;

Variables

TypeSafeAnswerSchema

Zod schema for any TypeSafe answer, discriminated by type: TypeSafeNoulAnswerSchema, TypeSafeChoiceAnswerSchema or TypeSafeScoreAnswerSchema.

import { TypeSafeAnswerSchema } from '@loopstack/typesafe-module';

TypeSafeChoiceAnswerSchema

Zod schema for a choice answer: the selected label, its confidence and the probability of every label.

import { TypeSafeChoiceAnswerSchema } from '@loopstack/typesafe-module';

TypeSafeChoiceQuestionSchema

Zod schema for a question that selects one of the labels in criteria (choice).

import { TypeSafeChoiceQuestionSchema } from '@loopstack/typesafe-module';

TypeSafeEntrySchema

Text, a JSON object or array, or null — the shape TypeSafe accepts for state, instructions and criteria.

import { TypeSafeEntrySchema } from '@loopstack/typesafe-module';

TypeSafeNoulAnswerSchema

Zod schema for a yes/no answer: the probability of yes.

import { TypeSafeNoulAnswerSchema } from '@loopstack/typesafe-module';

TypeSafeNoulQuestionSchema

Zod schema for a yes/no question (noul).

import { TypeSafeNoulQuestionSchema } from '@loopstack/typesafe-module';

TypeSafeQuestionSchema

Zod schema for any TypeSafe question, discriminated by type: TypeSafeNoulQuestionSchema, TypeSafeChoiceQuestionSchema or TypeSafeScoreQuestionSchema.

import { TypeSafeQuestionSchema } from '@loopstack/typesafe-module';

TypeSafeScoreAnswerSchema

Zod schema for a score answer: the expected score (may fall between levels), its confidence, the rubric and the probability of every level.

import { TypeSafeScoreAnswerSchema } from '@loopstack/typesafe-module';

TypeSafeScoreQuestionSchema

Zod schema for a question that scores against an ordered rubric of at least two levels (score).

import { TypeSafeScoreQuestionSchema } from '@loopstack/typesafe-module';

TypeSafeSystemOneArgsSchema

Zod schema for typesafe_system_one tool args: the state to evaluate and the named questions to answer.

import { TypeSafeSystemOneArgsSchema } from '@loopstack/typesafe-module';

TypeSafeSystemOneConfigSchema

Zod schema for typesafe_system_one tool config (model, envApiKey).

import { TypeSafeSystemOneConfigSchema } from '@loopstack/typesafe-module';

TypeSafeSystemOneConfigSchema: z.ZodObject<
  {
    model: z.ZodOptional<z.ZodString>;
    envApiKey: z.ZodOptional<z.ZodString>;
  },
  z.core.$strip
>;

TypeSafeSystemOneResultSchema

Zod schema for the typesafe_system_one result: the model that answered, the answers keyed by question name, and token usage as reported by the API.

import { TypeSafeSystemOneResultSchema } from '@loopstack/typesafe-module';

Re-exports from @typesafe-ai/sdk

The question builders choice, noul and score; the types Questions, SystemOneRequest and SystemOneResult; and the error classes TypeSafeError, APIError, APIConnectionError, APITimeoutError, APIUserAbortError, AuthenticationError, BadRequestError, InternalServerError, NotFoundError, PermissionDeniedError, RateLimitError and UnprocessableEntityError.