import {z} from 'zod'

export type DatasetReference = {name: string, version?: (string | null) | undefined};

export type TrainingJobCreate = {auto_data_sizing?: (boolean | null) | undefined, base_model: string, batch_size?: number | undefined, datasets: Array<DatasetReference>, early_stopping_min_delta?: number | undefined, early_stopping_patience?: number | undefined, encoder_learning_rate?: (number | null) | undefined, eval_steps?: (number | null) | undefined, gradient_accumulation_steps?: (number | null) | undefined, learning_rate?: (number | null) | undefined, lora_alpha?: (number | null) | undefined, lora_dropout?: (number | null) | undefined, lora_r?: (number | null) | undefined, lr_scheduler_type?: string | undefined, max_eval_samples?: (number | null) | undefined, max_gold_per_query?: (number | null) | undefined, max_length?: (number | null) | undefined, max_samples_per_dataset?: (number | null) | undefined, min_samples_per_dataset?: (number | null) | undefined, min_training_steps?: (number | null) | undefined, model_name: string, nr_epochs?: number | undefined, on_capacity_exceeded?: ('raise' | 'truncate_with_warning' | 'skip_sample' | null) | undefined, profile_training?: boolean | undefined, project_id?: (string | null) | undefined, provider_name?: (string | null) | undefined, samples_per_label?: (number | null) | undefined, save_steps?: number | undefined, seed?: (number | null) | undefined, task_learning_rate?: (number | null) | undefined, training_algorithm?: 'sft' | undefined, training_type?: ('full' | 'lora') | undefined, validation_data_percentage?: number | undefined, wandb_api_key?: (string | null) | undefined, warmup_ratio?: (number | null) | undefined, warmup_steps?: (number | null) | undefined, weight_decay?: number | undefined};

export type CreateApiV1TrainingJobsArgs = {'Idempotency-Key'?: (string | null) | undefined, body: TrainingJobCreate};

export type CreateApiV1TrainingJobsOptions = {baseUrl?: string | undefined, ApiKeyAuth?: string | undefined, fetch?: typeof fetch | undefined};

export const datasetReference = z.object({name: z.string(), version: z.string().nullable().optional()});

export const trainingNotice = z.object({code: z.string(), message: z.string()});

export const trainingJobResponse = z.object({artifact_ready: z.boolean().nullable().optional(), base_model: z.string(), batch_size: z.number().int().nullable(), completed_at: z.string().nullable().optional(), created_at: z.string(), current_epoch: z.number().int().nullable().optional(), datasets: z.array(datasetReference), deployability_reason: z.string().nullable().optional(), deployment_status: z.string().nullable(), error_message: z.string().nullable().optional(), example: z.string().nullable().optional(), experiment_id: z.string().nullable().optional(), hub_model_id: z.string().nullable().optional(), hub_model_private: z.boolean().nullable().optional(), id: z.string(), instance_type: z.string().nullable().optional(), is_deployable: z.boolean().nullable().optional(), is_terminal_status: z.boolean().nullable().optional(), job_reference: z.string().nullable().optional(), labels: z.array(z.string()).nullable().optional(), learning_rate: z.number().nullable(), metrics: z.record(z.string(), z.unknown()).nullable().optional(), model_auto_selected: z.boolean().nullable().optional(), model_kind: z.enum(["lora", "full"]).nullable().optional(), model_name: z.string().nullable().optional(), model_selection_reason: z.string().nullable().optional(), normalized_status: z.string().nullable().optional(), notices: z.array(trainingNotice).optional(), nr_epochs: z.number().int().nullable(), progress_percent: z.number().int().nullable().optional(), project_id: z.string().nullable().optional(), provider_deployments: z.record(z.string(), z.unknown()).nullable().optional(), provider_name: z.string().nullable().optional(), provider_ready: z.boolean().nullable().optional(), resolved_recipe: z.record(z.string(), z.unknown()).nullable().optional(), root_job_id: z.string().nullable().optional(), seed: z.number().int().nullable().optional(), source: z.enum(["config_training", "interactive_checkpoint"]).optional(), started_at: z.string().nullable().optional(), status: z.string(), task_type: z.string().nullable().optional(), trained_model_path: z.string().nullable().optional(), training_type: z.string().nullable().optional(), type: z.literal("config").optional(), updated_at: z.string(), user_id: z.string(), validation_data_percentage: z.number().nullable(), version_number: z.string().nullable().optional()});

export const validateResponse = async <Output>(
  schema: {
    readonly '~standard': {
      readonly validate: (value: unknown) => ({ readonly value: Output; readonly issues?: undefined } | { readonly issues: ReadonlyArray<{ readonly message: string; readonly path?: ReadonlyArray<PropertyKey | { readonly key: PropertyKey }> | undefined }> }) | Promise<{ readonly value: Output; readonly issues?: undefined } | { readonly issues: ReadonlyArray<{ readonly message: string; readonly path?: ReadonlyArray<PropertyKey | { readonly key: PropertyKey }> | undefined }> }>
    }
  },
  value: unknown
): Promise<Output> => {
  const result = await schema['~standard'].validate(value)

  if (result.issues !== undefined) {
    const detail = result.issues
      .map(({ message, path }) =>
        path?.length
          ? `${path.map((segment) => String(typeof segment === 'object' ? segment.key : segment)).join('.')}: ${message}`
          : message
      )
      .join('; ')

    throw new Error(`Response does not match the schema: ${detail}`, { cause: { issues: result.issues } })
  }

  return result.value
};

/**
 * Create Training Job
 * 
 * POST /v1/training/jobs
 * 
 * Also accepts auth: BearerAuth
 */
export const createApiV1TrainingJobs = async (args: CreateApiV1TrainingJobsArgs, options?: CreateApiV1TrainingJobsOptions) => {
const {baseUrl = "https://api.fastino.ai", fetch: fetchFn = fetch, ApiKeyAuth} = options ?? {}

const query = new URLSearchParams()
const headers = new Headers()
headers.set('content-type', "application/json")
headers.set('accept', "application/json")

if (args['Idempotency-Key'] != null) headers.set("Idempotency-Key", String(args['Idempotency-Key']))
if (ApiKeyAuth !== undefined) headers.set("X-API-Key", ApiKeyAuth)

const url = `${baseUrl.replace(/\/+$/, '')}/v1/training/jobs${query.size > 0 ? `?${query}` : ''}`

const res = await fetchFn(url, {method: "POST", headers, body: JSON.stringify(args.body)})

if (!res.ok) throw new Error(`Request failed with status ${res.status}`, {cause: {status: res.status, body: await res.text()}})

return validateResponse(trainingJobResponse, await res.json())
};