2
0
Files
bot/packages/schemas/features/blocks/integrations/openai.ts

151 lines
3.6 KiB
TypeScript

import { z } from 'zod'
import { variableStringSchema } from '../../utils'
import { blockBaseSchema, credentialsBaseSchema } from '../baseSchemas'
import { IntegrationBlockType } from './enums'
export const openAITasks = ['Create chat completion', 'Create image'] as const
export const chatCompletionModels = [
'gpt-4',
'gpt-4-0314',
'gpt-4-32k',
'gpt-4-32k-0314',
'gpt-3.5-turbo',
'gpt-3.5-turbo-0301',
] as const
export const modelLimit = {
'gpt-3.5-turbo': 4096,
'gpt-3.5-turbo-0301': 4096,
'gpt-4': 8192,
'gpt-4-0314': 8192,
'gpt-4-32k': 32768,
'gpt-4-32k-0314': 32768,
} as const
export const chatCompletionMessageRoles = [
'system',
'user',
'assistant',
] as const
export const chatCompletionMessageCustomRoles = [
'Messages sequence ✨',
] as const
export const chatCompletionResponseValues = [
'Message content',
'Total tokens',
] as const
const openAIBaseOptionsSchema = z.object({
credentialsId: z.string().optional(),
})
const initialOptionsSchema = z
.object({
task: z.undefined(),
})
.merge(openAIBaseOptionsSchema)
const chatCompletionMessageSchema = z.object({
id: z.string(),
role: z.enum(chatCompletionMessageRoles).optional(),
content: z.string().optional(),
})
const chatCompletionCustomMessageSchema = z.object({
id: z.string(),
role: z.enum(chatCompletionMessageCustomRoles),
content: z
.object({
assistantMessagesVariableId: z.string().optional(),
userMessagesVariableId: z.string().optional(),
})
.optional(),
})
const chatCompletionOptionsSchema = z
.object({
task: z.literal(openAITasks[0]),
model: z.enum(chatCompletionModels),
messages: z.array(
z.union([chatCompletionMessageSchema, chatCompletionCustomMessageSchema])
),
advancedSettings: z
.object({
temperature: z.number().or(variableStringSchema).optional(),
})
.optional(),
responseMapping: z.array(
z.object({
id: z.string(),
valueToExtract: z.enum(chatCompletionResponseValues),
variableId: z.string().optional(),
})
),
})
.merge(openAIBaseOptionsSchema)
const createImageOptionsSchema = z
.object({
task: z.literal(openAITasks[1]),
prompt: z.string().optional(),
advancedOptions: z.object({
size: z.enum(['256x256', '512x512', '1024x1024']).optional(),
}),
responseMapping: z.array(
z.object({
id: z.string(),
valueToExtract: z.enum(['Image URL']),
variableId: z.string().optional(),
})
),
})
.merge(openAIBaseOptionsSchema)
export const openAIBlockSchema = blockBaseSchema.merge(
z.object({
type: z.enum([IntegrationBlockType.OPEN_AI]),
options: z.discriminatedUnion('task', [
initialOptionsSchema,
chatCompletionOptionsSchema,
createImageOptionsSchema,
]),
})
)
export const openAICredentialsSchema = z
.object({
type: z.literal('openai'),
data: z.object({
apiKey: z.string(),
}),
})
.merge(credentialsBaseSchema)
export const defaultChatCompletionOptions = (
createId: () => string
): ChatCompletionOpenAIOptions => ({
task: 'Create chat completion',
messages: [
{
id: createId(),
},
],
responseMapping: [
{
id: createId(),
valueToExtract: 'Message content',
},
],
model: 'gpt-3.5-turbo',
})
export type OpenAICredentials = z.infer<typeof openAICredentialsSchema>
export type OpenAIBlock = z.infer<typeof openAIBlockSchema>
export type ChatCompletionOpenAIOptions = z.infer<
typeof chatCompletionOptionsSchema
>
export type CreateImageOpenAIOptions = z.infer<typeof createImageOptionsSchema>