LibreChat/api/server/services/Endpoints/schemas.js
Danny Avila 583e978a82
feat(Google): Support all Text/Chat Models, Response streaming, PaLM -> Google 🤖 (#1316)
* feat: update PaLM icons

* feat: add additional google models

* POC: formatting inputs for Vertex AI streaming

* refactor: move endpoints services outside of /routes dir to /services/Endpoints

* refactor: shorten schemas import

* refactor: rename PALM to GOOGLE

* feat: make Google editable endpoint

* feat: reusable Ask and Edit controllers based off Anthropic

* chore: organize imports/logic

* fix(parseConvo): include examples in googleSchema

* fix: google only allows odd number of messages to be sent

* fix: pass proxy to AnthropicClient

* refactor: change `google` altName to `Google`

* refactor: update getModelMaxTokens and related functions to handle maxTokensMap with nested endpoint model key/values

* refactor: google Icon and response sender changes (Codey and Google logo instead of PaLM in all cases)

* feat: google support for maxTokensMap

* feat: google updated endpoints with Ask/Edit controllers, buildOptions, and initializeClient

* feat(GoogleClient): now builds prompt for text models and supports real streaming from Vertex AI through langchain

* chore(GoogleClient): remove comments, left before for reference in git history

* docs: update google instructions (WIP)

* docs(apis_and_tokens.md): add images to google instructions

* docs: remove typo apis_and_tokens.md

* Update apis_and_tokens.md

* feat(Google): use default settings map, fully support context for both text and chat models, fully support examples for chat models

* chore: update more PaLM references to Google

* chore: move playwright out of workflows to avoid failing tests
2023-12-10 14:54:13 -05:00

445 lines
11 KiB
JavaScript

const { z } = require('zod');
const EModelEndpoint = {
azureOpenAI: 'azureOpenAI',
openAI: 'openAI',
bingAI: 'bingAI',
chatGPTBrowser: 'chatGPTBrowser',
google: 'google',
gptPlugins: 'gptPlugins',
anthropic: 'anthropic',
assistant: 'assistant',
};
const alternateName = {
[EModelEndpoint.openAI]: 'OpenAI',
[EModelEndpoint.assistant]: 'Assistants',
[EModelEndpoint.azureOpenAI]: 'Azure OpenAI',
[EModelEndpoint.bingAI]: 'Bing',
[EModelEndpoint.chatGPTBrowser]: 'ChatGPT',
[EModelEndpoint.gptPlugins]: 'Plugins',
[EModelEndpoint.google]: 'Google',
[EModelEndpoint.anthropic]: 'Anthropic',
};
const endpointSettings = {
[EModelEndpoint.google]: {
model: {
default: 'chat-bison',
},
maxOutputTokens: {
min: 1,
max: 2048,
step: 1,
default: 1024,
},
temperature: {
min: 0,
max: 1,
step: 0.01,
default: 0.2,
},
topP: {
min: 0,
max: 1,
step: 0.01,
default: 0.8,
},
topK: {
min: 1,
max: 40,
step: 0.01,
default: 40,
},
},
};
const google = endpointSettings[EModelEndpoint.google];
const supportsFiles = {
[EModelEndpoint.openAI]: true,
[EModelEndpoint.assistant]: true,
};
const openAIModels = [
'gpt-3.5-turbo-16k-0613',
'gpt-3.5-turbo-16k',
'gpt-4-1106-preview',
'gpt-3.5-turbo',
'gpt-3.5-turbo-1106',
'gpt-4-vision-preview',
'gpt-4',
'gpt-3.5-turbo-instruct-0914',
'gpt-3.5-turbo-0613',
'gpt-3.5-turbo-0301',
'gpt-3.5-turbo-instruct',
'gpt-4-0613',
'text-davinci-003',
'gpt-4-0314',
];
const visionModels = ['gpt-4-vision', 'llava-13b'];
const eModelEndpointSchema = z.nativeEnum(EModelEndpoint);
const tPluginAuthConfigSchema = z.object({
authField: z.string(),
label: z.string(),
description: z.string(),
});
const tPluginSchema = z.object({
name: z.string(),
pluginKey: z.string(),
description: z.string(),
icon: z.string(),
authConfig: z.array(tPluginAuthConfigSchema),
authenticated: z.boolean().optional(),
isButton: z.boolean().optional(),
});
const tExampleSchema = z.object({
input: z.object({
content: z.string(),
}),
output: z.object({
content: z.string(),
}),
});
const tAgentOptionsSchema = z.object({
agent: z.string(),
skipCompletion: z.boolean(),
model: z.string(),
temperature: z.number(),
});
const tConversationSchema = z.object({
conversationId: z.string().nullable(),
title: z.string().nullable().or(z.literal('New Chat')).default('New Chat'),
user: z.string().optional(),
endpoint: eModelEndpointSchema.nullable(),
suggestions: z.array(z.string()).optional(),
messages: z.array(z.string()).optional(),
tools: z.array(tPluginSchema).optional(),
createdAt: z.string(),
updatedAt: z.string(),
systemMessage: z.string().nullable().optional(),
modelLabel: z.string().nullable().optional(),
examples: z.array(tExampleSchema).optional(),
chatGptLabel: z.string().nullable().optional(),
userLabel: z.string().optional(),
model: z.string().nullable().optional(),
promptPrefix: z.string().nullable().optional(),
temperature: z.number().optional(),
topP: z.number().optional(),
topK: z.number().optional(),
context: z.string().nullable().optional(),
top_p: z.number().optional(),
frequency_penalty: z.number().optional(),
presence_penalty: z.number().optional(),
jailbreak: z.boolean().optional(),
jailbreakConversationId: z.string().nullable().optional(),
conversationSignature: z.string().nullable().optional(),
parentMessageId: z.string().optional(),
clientId: z.string().nullable().optional(),
invocationId: z.number().nullable().optional(),
toneStyle: z.string().nullable().optional(),
maxOutputTokens: z.number().optional(),
agentOptions: tAgentOptionsSchema.nullable().optional(),
});
const openAISchema = tConversationSchema
.pick({
model: true,
chatGptLabel: true,
promptPrefix: true,
temperature: true,
top_p: true,
presence_penalty: true,
frequency_penalty: true,
})
.transform((obj) => ({
...obj,
model: obj.model ?? 'gpt-3.5-turbo',
chatGptLabel: obj.chatGptLabel ?? null,
promptPrefix: obj.promptPrefix ?? null,
temperature: obj.temperature ?? 1,
top_p: obj.top_p ?? 1,
presence_penalty: obj.presence_penalty ?? 0,
frequency_penalty: obj.frequency_penalty ?? 0,
}))
.catch(() => ({
model: 'gpt-3.5-turbo',
chatGptLabel: null,
promptPrefix: null,
temperature: 1,
top_p: 1,
presence_penalty: 0,
frequency_penalty: 0,
}));
const googleSchema = tConversationSchema
.pick({
model: true,
modelLabel: true,
promptPrefix: true,
examples: true,
temperature: true,
maxOutputTokens: true,
topP: true,
topK: true,
})
.transform((obj) => ({
...obj,
model: obj.model ?? google.model.default,
modelLabel: obj.modelLabel ?? null,
promptPrefix: obj.promptPrefix ?? null,
examples: obj.examples ?? [{ input: { content: '' }, output: { content: '' } }],
temperature: obj.temperature ?? google.temperature.default,
maxOutputTokens: obj.maxOutputTokens ?? google.maxOutputTokens.default,
topP: obj.topP ?? google.topP.default,
topK: obj.topK ?? google.topK.default,
}))
.catch(() => ({
model: google.model.default,
modelLabel: null,
promptPrefix: null,
examples: [{ input: { content: '' }, output: { content: '' } }],
temperature: google.temperature.default,
maxOutputTokens: google.maxOutputTokens.default,
topP: google.topP.default,
topK: google.topK.default,
}));
const bingAISchema = tConversationSchema
.pick({
jailbreak: true,
systemMessage: true,
context: true,
toneStyle: true,
jailbreakConversationId: true,
conversationSignature: true,
clientId: true,
invocationId: true,
})
.transform((obj) => ({
...obj,
model: '',
jailbreak: obj.jailbreak ?? false,
systemMessage: obj.systemMessage ?? null,
context: obj.context ?? null,
toneStyle: obj.toneStyle ?? 'creative',
jailbreakConversationId: obj.jailbreakConversationId ?? null,
conversationSignature: obj.conversationSignature ?? null,
clientId: obj.clientId ?? null,
invocationId: obj.invocationId ?? 1,
}))
.catch(() => ({
model: '',
jailbreak: false,
systemMessage: null,
context: null,
toneStyle: 'creative',
jailbreakConversationId: null,
conversationSignature: null,
clientId: null,
invocationId: 1,
}));
const anthropicSchema = tConversationSchema
.pick({
model: true,
modelLabel: true,
promptPrefix: true,
temperature: true,
maxOutputTokens: true,
topP: true,
topK: true,
})
.transform((obj) => ({
...obj,
model: obj.model ?? 'claude-1',
modelLabel: obj.modelLabel ?? null,
promptPrefix: obj.promptPrefix ?? null,
temperature: obj.temperature ?? 1,
maxOutputTokens: obj.maxOutputTokens ?? 4000,
topP: obj.topP ?? 0.7,
topK: obj.topK ?? 5,
}))
.catch(() => ({
model: 'claude-1',
modelLabel: null,
promptPrefix: null,
temperature: 1,
maxOutputTokens: 4000,
topP: 0.7,
topK: 5,
}));
const chatGPTBrowserSchema = tConversationSchema
.pick({
model: true,
})
.transform((obj) => ({
...obj,
model: obj.model ?? 'text-davinci-002-render-sha',
}))
.catch(() => ({
model: 'text-davinci-002-render-sha',
}));
const gptPluginsSchema = tConversationSchema
.pick({
model: true,
chatGptLabel: true,
promptPrefix: true,
temperature: true,
top_p: true,
presence_penalty: true,
frequency_penalty: true,
tools: true,
agentOptions: true,
})
.transform((obj) => ({
...obj,
model: obj.model ?? 'gpt-3.5-turbo',
chatGptLabel: obj.chatGptLabel ?? null,
promptPrefix: obj.promptPrefix ?? null,
temperature: obj.temperature ?? 0.8,
top_p: obj.top_p ?? 1,
presence_penalty: obj.presence_penalty ?? 0,
frequency_penalty: obj.frequency_penalty ?? 0,
tools: obj.tools ?? [],
agentOptions: obj.agentOptions ?? {
agent: 'functions',
skipCompletion: true,
model: 'gpt-3.5-turbo',
temperature: 0,
},
}))
.catch(() => ({
model: 'gpt-3.5-turbo',
chatGptLabel: null,
promptPrefix: null,
temperature: 0.8,
top_p: 1,
presence_penalty: 0,
frequency_penalty: 0,
tools: [],
agentOptions: {
agent: 'functions',
skipCompletion: true,
model: 'gpt-3.5-turbo',
temperature: 0,
},
}));
const assistantSchema = tConversationSchema
.pick({
model: true,
assistant_id: true,
thread_id: true,
})
.transform((obj) => {
const newObj = { ...obj };
Object.keys(newObj).forEach((key) => {
const value = newObj[key];
if (value === undefined || value === null) {
delete newObj[key];
}
});
return newObj;
})
.catch(() => ({}));
const endpointSchemas = {
[EModelEndpoint.openAI]: openAISchema,
[EModelEndpoint.assistant]: assistantSchema,
[EModelEndpoint.azureOpenAI]: openAISchema,
[EModelEndpoint.google]: googleSchema,
[EModelEndpoint.bingAI]: bingAISchema,
[EModelEndpoint.anthropic]: anthropicSchema,
[EModelEndpoint.chatGPTBrowser]: chatGPTBrowserSchema,
[EModelEndpoint.gptPlugins]: gptPluginsSchema,
};
function getFirstDefinedValue(possibleValues) {
let returnValue;
for (const value of possibleValues) {
if (value) {
returnValue = value;
break;
}
}
return returnValue;
}
const parseConvo = (endpoint, conversation, possibleValues) => {
const schema = endpointSchemas[endpoint];
if (!schema) {
throw new Error(`Unknown endpoint: ${endpoint}`);
}
const convo = schema.parse(conversation);
if (possibleValues && convo) {
convo.model = getFirstDefinedValue(possibleValues.model) ?? convo.model;
}
return convo;
};
const getResponseSender = (endpointOption) => {
const { model, endpoint, chatGptLabel, modelLabel, jailbreak } = endpointOption;
if (
[
EModelEndpoint.openAI,
EModelEndpoint.azureOpenAI,
EModelEndpoint.gptPlugins,
EModelEndpoint.chatGPTBrowser,
].includes(endpoint)
) {
if (chatGptLabel) {
return chatGptLabel;
} else if (model && model.includes('gpt-3')) {
return 'GPT-3.5';
} else if (model && model.includes('gpt-4')) {
return 'GPT-4';
}
return alternateName[endpoint] ?? 'ChatGPT';
}
if (endpoint === EModelEndpoint.bingAI) {
return jailbreak ? 'Sydney' : 'BingAI';
}
if (endpoint === EModelEndpoint.anthropic) {
return modelLabel ?? 'Claude';
}
if (endpoint === EModelEndpoint.google) {
if (modelLabel) {
return modelLabel;
} else if (model && model.includes('code')) {
return 'Codey';
}
return 'PaLM2';
}
return '';
};
module.exports = {
parseConvo,
getResponseSender,
EModelEndpoint,
supportsFiles,
openAIModels,
visionModels,
alternateName,
endpointSettings,
};