🤖 feat: OpenAI Assistants v2 (initial support) (#2781)

* 🤖 Assistants V2 Support: Part 1

- Separated Azure Assistants to its own endpoint
- File Search / Vector Store integration is incomplete, but can toggle and use storage from playground
- Code Interpreter resource files can be added but not deleted
- GPT-4o is supported
- Many improvements to the Assistants Endpoint overall

data-provider v2 changes

copy existing route as v1

chore: rename new endpoint to reduce comparison operations and add new azure filesource

api: add azureAssistants part 1

force use of version for assistants/assistantsAzure

chore: switch name back to azureAssistants

refactor type version: string | number

Ensure assistants endpoints have version set

fix: isArchived type issue in ConversationListParams

refactor: update assistants mutations/queries with endpoint/version definitions, update Assistants Map structure

chore:  FilePreview component ExtendedFile type assertion

feat: isAssistantsEndpoint helper

chore: remove unused useGenerations

chore(buildTree): type issue

chore(Advanced): type issue (unused component, maybe in future)

first pass for multi-assistant endpoint rewrite

fix(listAssistants): pass params correctly

feat: list separate assistants by endpoint

fix(useTextarea): access assistantMap correctly

fix: assistant endpoint switching, resetting ID

fix: broken during rewrite, selecting assistant mention

fix: set/invalidate assistants endpoint query data correctly

feat: Fix issue with assistant ID not being reset correctly

getOpenAIClient helper function

feat: add toast for assistant deletion

fix: assistants delete right after create issue for azure

fix: assistant patching

refactor: actions to use getOpenAIClient

refactor: consolidate logic into helpers file

fix: issue where conversation data was not initially available

v1 chat support

refactor(spendTokens): only early return if completionTokens isNaN

fix(OpenAIClient): ensure spendTokens has all necessary params

refactor: route/controller logic

fix(assistants/initializeClient): use defaultHeaders field

fix: sanitize default operation id

chore: bump openai package

first pass v2 action service

feat: retroactive domain parsing for actions added via v1

feat: delete db records of actions/assistants on openai assistant deletion

chore: remove vision tools from v2 assistants

feat: v2 upload and delete assistant vision images

WIP first pass, thread attachments

fix: show assistant vision files (save local/firebase copy)

v2 image continue

fix: annotations

fix: refine annotations

show analyze as error if is no longer submitting before progress reaches 1 and show file_search as retrieval tool

fix: abort run, undefined endpoint issue

refactor: consolidate capabilities logic and anticipate versioning

frontend version 2 changes

fix: query selection and filter

add endpoint to unknown filepath

add file ids to resource, deleting in progress

enable/disable file search

remove version log

* 🤖 Assistants V2 Support: Part 2

🎹 fix: Autocompletion Chrome Bug on Action API Key Input

chore: remove `useOriginNavigate`

chore: set correct OpenAI Storage Source

fix: azure file deletions, instantiate clients by source for deletion

update code interpret files info

feat: deleteResourceFileId

chore: increase poll interval as azure easily rate limits

fix: openai file deletions, TODO: evaluate rejected deletion settled promises to determine which to delete from db records

file source icons

update table file filters

chore: file search info and versioning

fix: retrieval update with necessary tool_resources if specified

fix(useMentions): add optional chaining in case listMap value is undefined

fix: force assistant avatar roundedness

fix: azure assistants, check correct flag

chore: bump data-provider

* fix: merge conflict

* ci: fix backend tests due to new updates

* chore: update .env.example

* meilisearch improvements

* localization updates

* chore: update comparisons

* feat: add additional metadata: endpoint, author ID

* chore: azureAssistants ENDPOINTS exclusion warning
This commit is contained in:
Danny Avila 2024-05-19 12:56:55 -04:00 committed by GitHub
parent af8bcb08d6
commit 1a452121fa
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
158 changed files with 4184 additions and 1204 deletions

View file

@ -0,0 +1,650 @@
const { v4 } = require('uuid');
const {
Constants,
RunStatus,
CacheKeys,
ContentTypes,
EModelEndpoint,
ViolationTypes,
ImageVisionTool,
checkOpenAIStorage,
AssistantStreamEvents,
} = require('librechat-data-provider');
const {
initThread,
recordUsage,
saveUserMessage,
checkMessageGaps,
addThreadMetadata,
saveAssistantMessage,
} = require('~/server/services/Threads');
const { sendResponse, sendMessage, sleep, isEnabled, countTokens } = require('~/server/utils');
const { runAssistant, createOnTextProgress } = require('~/server/services/AssistantService');
const { formatMessage, createVisionPrompt } = require('~/app/clients/prompts');
const { createRun, StreamRunManager } = require('~/server/services/Runs');
const { addTitle } = require('~/server/services/Endpoints/assistants');
const { getTransactions } = require('~/models/Transaction');
const checkBalance = require('~/models/checkBalance');
const { getConvo } = require('~/models/Conversation');
const getLogStores = require('~/cache/getLogStores');
const { getModelMaxTokens } = require('~/utils');
const { getOpenAIClient } = require('./helpers');
const { logger } = require('~/config');
const { handleAbortError } = require('~/server/middleware');
const ten_minutes = 1000 * 60 * 10;
/**
* @route POST /
* @desc Chat with an assistant
* @access Public
* @param {Express.Request} req - The request object, containing the request data.
* @param {Express.Response} res - The response object, used to send back a response.
* @returns {void}
*/
const chatV1 = async (req, res) => {
logger.debug('[/assistants/chat/] req.body', req.body);
const {
text,
model,
endpoint,
files = [],
promptPrefix,
assistant_id,
instructions,
thread_id: _thread_id,
messageId: _messageId,
conversationId: convoId,
parentMessageId: _parentId = Constants.NO_PARENT,
} = req.body;
/** @type {Partial<TAssistantEndpoint>} */
const assistantsConfig = req.app.locals?.[endpoint];
if (assistantsConfig) {
const { supportedIds, excludedIds } = assistantsConfig;
const error = { message: 'Assistant not supported' };
if (supportedIds?.length && !supportedIds.includes(assistant_id)) {
return await handleAbortError(res, req, error, {
sender: 'System',
conversationId: convoId,
messageId: v4(),
parentMessageId: _messageId,
error,
});
} else if (excludedIds?.length && excludedIds.includes(assistant_id)) {
return await handleAbortError(res, req, error, {
sender: 'System',
conversationId: convoId,
messageId: v4(),
parentMessageId: _messageId,
});
}
}
/** @type {OpenAIClient} */
let openai;
/** @type {string|undefined} - the current thread id */
let thread_id = _thread_id;
/** @type {string|undefined} - the current run id */
let run_id;
/** @type {string|undefined} - the parent messageId */
let parentMessageId = _parentId;
/** @type {TMessage[]} */
let previousMessages = [];
/** @type {import('librechat-data-provider').TConversation | null} */
let conversation = null;
/** @type {string[]} */
let file_ids = [];
/** @type {Set<string>} */
let attachedFileIds = new Set();
/** @type {TMessage | null} */
let requestMessage = null;
/** @type {undefined | Promise<ChatCompletion>} */
let visionPromise;
const userMessageId = v4();
const responseMessageId = v4();
/** @type {string} - The conversation UUID - created if undefined */
const conversationId = convoId ?? v4();
const cache = getLogStores(CacheKeys.ABORT_KEYS);
const cacheKey = `${req.user.id}:${conversationId}`;
/** @type {Run | undefined} - The completed run, undefined if incomplete */
let completedRun;
const handleError = async (error) => {
const defaultErrorMessage =
'The Assistant run failed to initialize. Try sending a message in a new conversation.';
const messageData = {
thread_id,
assistant_id,
conversationId,
parentMessageId,
sender: 'System',
user: req.user.id,
shouldSaveMessage: false,
messageId: responseMessageId,
endpoint,
};
if (error.message === 'Run cancelled') {
return res.end();
} else if (error.message === 'Request closed' && completedRun) {
return;
} else if (error.message === 'Request closed') {
logger.debug('[/assistants/chat/] Request aborted on close');
} else if (/Files.*are invalid/.test(error.message)) {
const errorMessage = `Files are invalid, or may not have uploaded yet.${
endpoint === EModelEndpoint.azureAssistants
? ' If using Azure OpenAI, files are only available in the region of the assistant\'s model at the time of upload.'
: ''
}`;
return sendResponse(res, messageData, errorMessage);
} else if (error?.message?.includes('string too long')) {
return sendResponse(
res,
messageData,
'Message too long. The Assistants API has a limit of 32,768 characters per message. Please shorten it and try again.',
);
} else if (error?.message?.includes(ViolationTypes.TOKEN_BALANCE)) {
return sendResponse(res, messageData, error.message);
} else {
logger.error('[/assistants/chat/]', error);
}
if (!openai || !thread_id || !run_id) {
return sendResponse(res, messageData, defaultErrorMessage);
}
await sleep(2000);
try {
const status = await cache.get(cacheKey);
if (status === 'cancelled') {
logger.debug('[/assistants/chat/] Run already cancelled');
return res.end();
}
await cache.delete(cacheKey);
const cancelledRun = await openai.beta.threads.runs.cancel(thread_id, run_id);
logger.debug('[/assistants/chat/] Cancelled run:', cancelledRun);
} catch (error) {
logger.error('[/assistants/chat/] Error cancelling run', error);
}
await sleep(2000);
let run;
try {
run = await openai.beta.threads.runs.retrieve(thread_id, run_id);
await recordUsage({
...run.usage,
model: run.model,
user: req.user.id,
conversationId,
});
} catch (error) {
logger.error('[/assistants/chat/] Error fetching or processing run', error);
}
let finalEvent;
try {
const runMessages = await checkMessageGaps({
openai,
run_id,
endpoint,
thread_id,
conversationId,
latestMessageId: responseMessageId,
});
const errorContentPart = {
text: {
value:
error?.message ?? 'There was an error processing your request. Please try again later.',
},
type: ContentTypes.ERROR,
};
if (!Array.isArray(runMessages[runMessages.length - 1]?.content)) {
runMessages[runMessages.length - 1].content = [errorContentPart];
} else {
const contentParts = runMessages[runMessages.length - 1].content;
for (let i = 0; i < contentParts.length; i++) {
const currentPart = contentParts[i];
/** @type {CodeToolCall | RetrievalToolCall | FunctionToolCall | undefined} */
const toolCall = currentPart?.[ContentTypes.TOOL_CALL];
if (
toolCall &&
toolCall?.function &&
!(toolCall?.function?.output || toolCall?.function?.output?.length)
) {
contentParts[i] = {
...currentPart,
[ContentTypes.TOOL_CALL]: {
...toolCall,
function: {
...toolCall.function,
output: 'error processing tool',
},
},
};
}
}
runMessages[runMessages.length - 1].content.push(errorContentPart);
}
finalEvent = {
final: true,
conversation: await getConvo(req.user.id, conversationId),
runMessages,
};
} catch (error) {
logger.error('[/assistants/chat/] Error finalizing error process', error);
return sendResponse(res, messageData, 'The Assistant run failed');
}
return sendResponse(res, finalEvent);
};
try {
res.on('close', async () => {
if (!completedRun) {
await handleError(new Error('Request closed'));
}
});
if (convoId && !_thread_id) {
completedRun = true;
throw new Error('Missing thread_id for existing conversation');
}
if (!assistant_id) {
completedRun = true;
throw new Error('Missing assistant_id');
}
const checkBalanceBeforeRun = async () => {
if (!isEnabled(process.env.CHECK_BALANCE)) {
return;
}
const transactions =
(await getTransactions({
user: req.user.id,
context: 'message',
conversationId,
})) ?? [];
const totalPreviousTokens = Math.abs(
transactions.reduce((acc, curr) => acc + curr.rawAmount, 0),
);
// TODO: make promptBuffer a config option; buffer for titles, needs buffer for system instructions
const promptBuffer = parentMessageId === Constants.NO_PARENT && !_thread_id ? 200 : 0;
// 5 is added for labels
let promptTokens = (await countTokens(text + (promptPrefix ?? ''))) + 5;
promptTokens += totalPreviousTokens + promptBuffer;
// Count tokens up to the current context window
promptTokens = Math.min(promptTokens, getModelMaxTokens(model));
await checkBalance({
req,
res,
txData: {
model,
user: req.user.id,
tokenType: 'prompt',
amount: promptTokens,
},
});
};
const { openai: _openai, client } = await getOpenAIClient({
req,
res,
endpointOption: req.body.endpointOption,
initAppClient: true,
});
openai = _openai;
if (previousMessages.length) {
parentMessageId = previousMessages[previousMessages.length - 1].messageId;
}
let userMessage = {
role: 'user',
content: text,
metadata: {
messageId: userMessageId,
},
};
/** @type {CreateRunBody | undefined} */
const body = {
assistant_id,
model,
};
if (promptPrefix) {
body.additional_instructions = promptPrefix;
}
if (instructions) {
body.instructions = instructions;
}
const getRequestFileIds = async () => {
let thread_file_ids = [];
if (convoId) {
const convo = await getConvo(req.user.id, convoId);
if (convo && convo.file_ids) {
thread_file_ids = convo.file_ids;
}
}
file_ids = files.map(({ file_id }) => file_id);
if (file_ids.length || thread_file_ids.length) {
userMessage.file_ids = file_ids;
attachedFileIds = new Set([...file_ids, ...thread_file_ids]);
}
};
const addVisionPrompt = async () => {
if (!req.body.endpointOption.attachments) {
return;
}
/** @type {MongoFile[]} */
const attachments = await req.body.endpointOption.attachments;
if (attachments && attachments.every((attachment) => checkOpenAIStorage(attachment.source))) {
return;
}
const assistant = await openai.beta.assistants.retrieve(assistant_id);
const visionToolIndex = assistant.tools.findIndex(
(tool) => tool?.function && tool?.function?.name === ImageVisionTool.function.name,
);
if (visionToolIndex === -1) {
return;
}
let visionMessage = {
role: 'user',
content: '',
};
const files = await client.addImageURLs(visionMessage, attachments);
if (!visionMessage.image_urls?.length) {
return;
}
const imageCount = visionMessage.image_urls.length;
const plural = imageCount > 1;
visionMessage.content = createVisionPrompt(plural);
visionMessage = formatMessage({ message: visionMessage, endpoint: EModelEndpoint.openAI });
visionPromise = openai.chat.completions.create({
model: 'gpt-4-vision-preview',
messages: [visionMessage],
max_tokens: 4000,
});
const pluralized = plural ? 's' : '';
body.additional_instructions = `${
body.additional_instructions ? `${body.additional_instructions}\n` : ''
}The user has uploaded ${imageCount} image${pluralized}.
Use the \`${ImageVisionTool.function.name}\` tool to retrieve ${
plural ? '' : 'a '
}detailed text description${pluralized} for ${plural ? 'each' : 'the'} image${pluralized}.`;
return files;
};
const initializeThread = async () => {
/** @type {[ undefined | MongoFile[]]}*/
const [processedFiles] = await Promise.all([addVisionPrompt(), getRequestFileIds()]);
// TODO: may allow multiple messages to be created beforehand in a future update
const initThreadBody = {
messages: [userMessage],
metadata: {
user: req.user.id,
conversationId,
},
};
if (processedFiles) {
for (const file of processedFiles) {
if (!checkOpenAIStorage(file.source)) {
attachedFileIds.delete(file.file_id);
const index = file_ids.indexOf(file.file_id);
if (index > -1) {
file_ids.splice(index, 1);
}
}
}
userMessage.file_ids = file_ids;
}
const result = await initThread({ openai, body: initThreadBody, thread_id });
thread_id = result.thread_id;
createOnTextProgress({
openai,
conversationId,
userMessageId,
messageId: responseMessageId,
thread_id,
});
requestMessage = {
user: req.user.id,
text,
messageId: userMessageId,
parentMessageId,
// TODO: make sure client sends correct format for `files`, use zod
files,
file_ids,
conversationId,
isCreatedByUser: true,
assistant_id,
thread_id,
model: assistant_id,
endpoint,
};
previousMessages.push(requestMessage);
/* asynchronous */
saveUserMessage({ ...requestMessage, model });
conversation = {
conversationId,
endpoint,
promptPrefix: promptPrefix,
instructions: instructions,
assistant_id,
// model,
};
if (file_ids.length) {
conversation.file_ids = file_ids;
}
};
const promises = [initializeThread(), checkBalanceBeforeRun()];
await Promise.all(promises);
const sendInitialResponse = () => {
sendMessage(res, {
sync: true,
conversationId,
// messages: previousMessages,
requestMessage,
responseMessage: {
user: req.user.id,
messageId: openai.responseMessage.messageId,
parentMessageId: userMessageId,
conversationId,
assistant_id,
thread_id,
model: assistant_id,
},
});
};
/** @type {RunResponse | typeof StreamRunManager | undefined} */
let response;
const processRun = async (retry = false) => {
if (endpoint === EModelEndpoint.azureAssistants) {
body.model = openai._options.model;
openai.attachedFileIds = attachedFileIds;
openai.visionPromise = visionPromise;
if (retry) {
response = await runAssistant({
openai,
thread_id,
run_id,
in_progress: openai.in_progress,
});
return;
}
/* NOTE:
* By default, a Run will use the model and tools configuration specified in Assistant object,
* but you can override most of these when creating the Run for added flexibility:
*/
const run = await createRun({
openai,
thread_id,
body,
});
run_id = run.id;
await cache.set(cacheKey, `${thread_id}:${run_id}`, ten_minutes);
sendInitialResponse();
// todo: retry logic
response = await runAssistant({ openai, thread_id, run_id });
return;
}
/** @type {{[AssistantStreamEvents.ThreadRunCreated]: (event: ThreadRunCreated) => Promise<void>}} */
const handlers = {
[AssistantStreamEvents.ThreadRunCreated]: async (event) => {
await cache.set(cacheKey, `${thread_id}:${event.data.id}`, ten_minutes);
run_id = event.data.id;
sendInitialResponse();
},
};
const streamRunManager = new StreamRunManager({
req,
res,
openai,
handlers,
thread_id,
visionPromise,
attachedFileIds,
responseMessage: openai.responseMessage,
// streamOptions: {
// },
});
await streamRunManager.runAssistant({
thread_id,
body,
});
response = streamRunManager;
};
await processRun();
logger.debug('[/assistants/chat/] response', {
run: response.run,
steps: response.steps,
});
if (response.run.status === RunStatus.CANCELLED) {
logger.debug('[/assistants/chat/] Run cancelled, handled by `abortRun`');
return res.end();
}
if (response.run.status === RunStatus.IN_PROGRESS) {
processRun(true);
}
completedRun = response.run;
/** @type {ResponseMessage} */
const responseMessage = {
...(response.responseMessage ?? response.finalMessage),
parentMessageId: userMessageId,
conversationId,
user: req.user.id,
assistant_id,
thread_id,
model: assistant_id,
endpoint,
};
sendMessage(res, {
final: true,
conversation,
requestMessage: {
parentMessageId,
thread_id,
},
});
res.end();
await saveAssistantMessage({ ...responseMessage, model });
if (parentMessageId === Constants.NO_PARENT && !_thread_id) {
addTitle(req, {
text,
responseText: response.text,
conversationId,
client,
});
}
await addThreadMetadata({
openai,
thread_id,
messageId: responseMessage.messageId,
messages: response.messages,
});
if (!response.run.usage) {
await sleep(3000);
completedRun = await openai.beta.threads.runs.retrieve(thread_id, response.run.id);
if (completedRun.usage) {
await recordUsage({
...completedRun.usage,
user: req.user.id,
model: completedRun.model ?? model,
conversationId,
});
}
} else {
await recordUsage({
...response.run.usage,
user: req.user.id,
model: response.run.model ?? model,
conversationId,
});
}
} catch (error) {
await handleError(error);
}
};
module.exports = chatV1;

View file

@ -0,0 +1,618 @@
const { v4 } = require('uuid');
const {
Constants,
RunStatus,
CacheKeys,
ContentTypes,
ToolCallTypes,
EModelEndpoint,
ViolationTypes,
retrievalMimeTypes,
AssistantStreamEvents,
} = require('librechat-data-provider');
const {
initThread,
recordUsage,
saveUserMessage,
checkMessageGaps,
addThreadMetadata,
saveAssistantMessage,
} = require('~/server/services/Threads');
const { sendResponse, sendMessage, sleep, isEnabled, countTokens } = require('~/server/utils');
const { runAssistant, createOnTextProgress } = require('~/server/services/AssistantService');
const { createRun, StreamRunManager } = require('~/server/services/Runs');
const { addTitle } = require('~/server/services/Endpoints/assistants');
const { getTransactions } = require('~/models/Transaction');
const checkBalance = require('~/models/checkBalance');
const { getConvo } = require('~/models/Conversation');
const getLogStores = require('~/cache/getLogStores');
const { getModelMaxTokens } = require('~/utils');
const { getOpenAIClient } = require('./helpers');
const { logger } = require('~/config');
const { handleAbortError } = require('~/server/middleware');
const ten_minutes = 1000 * 60 * 10;
/**
* @route POST /
* @desc Chat with an assistant
* @access Public
* @param {Express.Request} req - The request object, containing the request data.
* @param {Express.Response} res - The response object, used to send back a response.
* @returns {void}
*/
const chatV2 = async (req, res) => {
logger.debug('[/assistants/chat/] req.body', req.body);
/** @type {{ files: MongoFile[]}} */
const {
text,
model,
endpoint,
files = [],
promptPrefix,
assistant_id,
instructions,
thread_id: _thread_id,
messageId: _messageId,
conversationId: convoId,
parentMessageId: _parentId = Constants.NO_PARENT,
} = req.body;
/** @type {Partial<TAssistantEndpoint>} */
const assistantsConfig = req.app.locals?.[endpoint];
if (assistantsConfig) {
const { supportedIds, excludedIds } = assistantsConfig;
const error = { message: 'Assistant not supported' };
if (supportedIds?.length && !supportedIds.includes(assistant_id)) {
return await handleAbortError(res, req, error, {
sender: 'System',
conversationId: convoId,
messageId: v4(),
parentMessageId: _messageId,
error,
});
} else if (excludedIds?.length && excludedIds.includes(assistant_id)) {
return await handleAbortError(res, req, error, {
sender: 'System',
conversationId: convoId,
messageId: v4(),
parentMessageId: _messageId,
});
}
}
/** @type {OpenAIClient} */
let openai;
/** @type {string|undefined} - the current thread id */
let thread_id = _thread_id;
/** @type {string|undefined} - the current run id */
let run_id;
/** @type {string|undefined} - the parent messageId */
let parentMessageId = _parentId;
/** @type {TMessage[]} */
let previousMessages = [];
/** @type {import('librechat-data-provider').TConversation | null} */
let conversation = null;
/** @type {string[]} */
let file_ids = [];
/** @type {Set<string>} */
let attachedFileIds = new Set();
/** @type {TMessage | null} */
let requestMessage = null;
const userMessageId = v4();
const responseMessageId = v4();
/** @type {string} - The conversation UUID - created if undefined */
const conversationId = convoId ?? v4();
const cache = getLogStores(CacheKeys.ABORT_KEYS);
const cacheKey = `${req.user.id}:${conversationId}`;
/** @type {Run | undefined} - The completed run, undefined if incomplete */
let completedRun;
const handleError = async (error) => {
const defaultErrorMessage =
'The Assistant run failed to initialize. Try sending a message in a new conversation.';
const messageData = {
thread_id,
assistant_id,
conversationId,
parentMessageId,
sender: 'System',
user: req.user.id,
shouldSaveMessage: false,
messageId: responseMessageId,
endpoint,
};
if (error.message === 'Run cancelled') {
return res.end();
} else if (error.message === 'Request closed' && completedRun) {
return;
} else if (error.message === 'Request closed') {
logger.debug('[/assistants/chat/] Request aborted on close');
} else if (/Files.*are invalid/.test(error.message)) {
const errorMessage = `Files are invalid, or may not have uploaded yet.${
endpoint === EModelEndpoint.azureAssistants
? ' If using Azure OpenAI, files are only available in the region of the assistant\'s model at the time of upload.'
: ''
}`;
return sendResponse(res, messageData, errorMessage);
} else if (error?.message?.includes('string too long')) {
return sendResponse(
res,
messageData,
'Message too long. The Assistants API has a limit of 32,768 characters per message. Please shorten it and try again.',
);
} else if (error?.message?.includes(ViolationTypes.TOKEN_BALANCE)) {
return sendResponse(res, messageData, error.message);
} else {
logger.error('[/assistants/chat/]', error);
}
if (!openai || !thread_id || !run_id) {
return sendResponse(res, messageData, defaultErrorMessage);
}
await sleep(2000);
try {
const status = await cache.get(cacheKey);
if (status === 'cancelled') {
logger.debug('[/assistants/chat/] Run already cancelled');
return res.end();
}
await cache.delete(cacheKey);
const cancelledRun = await openai.beta.threads.runs.cancel(thread_id, run_id);
logger.debug('[/assistants/chat/] Cancelled run:', cancelledRun);
} catch (error) {
logger.error('[/assistants/chat/] Error cancelling run', error);
}
await sleep(2000);
let run;
try {
run = await openai.beta.threads.runs.retrieve(thread_id, run_id);
await recordUsage({
...run.usage,
model: run.model,
user: req.user.id,
conversationId,
});
} catch (error) {
logger.error('[/assistants/chat/] Error fetching or processing run', error);
}
let finalEvent;
try {
const runMessages = await checkMessageGaps({
openai,
run_id,
endpoint,
thread_id,
conversationId,
latestMessageId: responseMessageId,
});
const errorContentPart = {
text: {
value:
error?.message ?? 'There was an error processing your request. Please try again later.',
},
type: ContentTypes.ERROR,
};
if (!Array.isArray(runMessages[runMessages.length - 1]?.content)) {
runMessages[runMessages.length - 1].content = [errorContentPart];
} else {
const contentParts = runMessages[runMessages.length - 1].content;
for (let i = 0; i < contentParts.length; i++) {
const currentPart = contentParts[i];
/** @type {CodeToolCall | RetrievalToolCall | FunctionToolCall | undefined} */
const toolCall = currentPart?.[ContentTypes.TOOL_CALL];
if (
toolCall &&
toolCall?.function &&
!(toolCall?.function?.output || toolCall?.function?.output?.length)
) {
contentParts[i] = {
...currentPart,
[ContentTypes.TOOL_CALL]: {
...toolCall,
function: {
...toolCall.function,
output: 'error processing tool',
},
},
};
}
}
runMessages[runMessages.length - 1].content.push(errorContentPart);
}
finalEvent = {
final: true,
conversation: await getConvo(req.user.id, conversationId),
runMessages,
};
} catch (error) {
logger.error('[/assistants/chat/] Error finalizing error process', error);
return sendResponse(res, messageData, 'The Assistant run failed');
}
return sendResponse(res, finalEvent);
};
try {
res.on('close', async () => {
if (!completedRun) {
await handleError(new Error('Request closed'));
}
});
if (convoId && !_thread_id) {
completedRun = true;
throw new Error('Missing thread_id for existing conversation');
}
if (!assistant_id) {
completedRun = true;
throw new Error('Missing assistant_id');
}
const checkBalanceBeforeRun = async () => {
if (!isEnabled(process.env.CHECK_BALANCE)) {
return;
}
const transactions =
(await getTransactions({
user: req.user.id,
context: 'message',
conversationId,
})) ?? [];
const totalPreviousTokens = Math.abs(
transactions.reduce((acc, curr) => acc + curr.rawAmount, 0),
);
// TODO: make promptBuffer a config option; buffer for titles, needs buffer for system instructions
const promptBuffer = parentMessageId === Constants.NO_PARENT && !_thread_id ? 200 : 0;
// 5 is added for labels
let promptTokens = (await countTokens(text + (promptPrefix ?? ''))) + 5;
promptTokens += totalPreviousTokens + promptBuffer;
// Count tokens up to the current context window
promptTokens = Math.min(promptTokens, getModelMaxTokens(model));
await checkBalance({
req,
res,
txData: {
model,
user: req.user.id,
tokenType: 'prompt',
amount: promptTokens,
},
});
};
const { openai: _openai, client } = await getOpenAIClient({
req,
res,
endpointOption: req.body.endpointOption,
initAppClient: true,
});
openai = _openai;
if (previousMessages.length) {
parentMessageId = previousMessages[previousMessages.length - 1].messageId;
}
let userMessage = {
role: 'user',
content: [
{
type: ContentTypes.TEXT,
text,
},
],
metadata: {
messageId: userMessageId,
},
};
/** @type {CreateRunBody | undefined} */
const body = {
assistant_id,
model,
};
if (promptPrefix) {
body.additional_instructions = promptPrefix;
}
if (instructions) {
body.instructions = instructions;
}
const getRequestFileIds = async () => {
let thread_file_ids = [];
if (convoId) {
const convo = await getConvo(req.user.id, convoId);
if (convo && convo.file_ids) {
thread_file_ids = convo.file_ids;
}
}
if (files.length || thread_file_ids.length) {
attachedFileIds = new Set([...file_ids, ...thread_file_ids]);
let attachmentIndex = 0;
for (const file of files) {
file_ids.push(file.file_id);
if (file.type.startsWith('image')) {
userMessage.content.push({
type: ContentTypes.IMAGE_FILE,
[ContentTypes.IMAGE_FILE]: { file_id: file.file_id },
});
}
if (!userMessage.attachments) {
userMessage.attachments = [];
}
userMessage.attachments.push({
file_id: file.file_id,
tools: [{ type: ToolCallTypes.CODE_INTERPRETER }],
});
if (file.type.startsWith('image')) {
continue;
}
const mimeType = file.type;
const isSupportedByRetrieval = retrievalMimeTypes.some((regex) => regex.test(mimeType));
if (isSupportedByRetrieval) {
userMessage.attachments[attachmentIndex].tools.push({
type: ToolCallTypes.FILE_SEARCH,
});
}
attachmentIndex++;
}
}
};
const initializeThread = async () => {
await getRequestFileIds();
// TODO: may allow multiple messages to be created beforehand in a future update
const initThreadBody = {
messages: [userMessage],
metadata: {
user: req.user.id,
conversationId,
},
};
const result = await initThread({ openai, body: initThreadBody, thread_id });
thread_id = result.thread_id;
createOnTextProgress({
openai,
conversationId,
userMessageId,
messageId: responseMessageId,
thread_id,
});
requestMessage = {
user: req.user.id,
text,
messageId: userMessageId,
parentMessageId,
// TODO: make sure client sends correct format for `files`, use zod
files,
file_ids,
conversationId,
isCreatedByUser: true,
assistant_id,
thread_id,
model: assistant_id,
endpoint,
};
previousMessages.push(requestMessage);
/* asynchronous */
saveUserMessage({ ...requestMessage, model });
conversation = {
conversationId,
endpoint,
promptPrefix: promptPrefix,
instructions: instructions,
assistant_id,
// model,
};
if (file_ids.length) {
conversation.file_ids = file_ids;
}
};
const promises = [initializeThread(), checkBalanceBeforeRun()];
await Promise.all(promises);
const sendInitialResponse = () => {
sendMessage(res, {
sync: true,
conversationId,
// messages: previousMessages,
requestMessage,
responseMessage: {
user: req.user.id,
messageId: openai.responseMessage.messageId,
parentMessageId: userMessageId,
conversationId,
assistant_id,
thread_id,
model: assistant_id,
},
});
};
/** @type {RunResponse | typeof StreamRunManager | undefined} */
let response;
const processRun = async (retry = false) => {
if (endpoint === EModelEndpoint.azureAssistants) {
body.model = openai._options.model;
openai.attachedFileIds = attachedFileIds;
if (retry) {
response = await runAssistant({
openai,
thread_id,
run_id,
in_progress: openai.in_progress,
});
return;
}
/* NOTE:
* By default, a Run will use the model and tools configuration specified in Assistant object,
* but you can override most of these when creating the Run for added flexibility:
*/
const run = await createRun({
openai,
thread_id,
body,
});
run_id = run.id;
await cache.set(cacheKey, `${thread_id}:${run_id}`, ten_minutes);
sendInitialResponse();
// todo: retry logic
response = await runAssistant({ openai, thread_id, run_id });
return;
}
/** @type {{[AssistantStreamEvents.ThreadRunCreated]: (event: ThreadRunCreated) => Promise<void>}} */
const handlers = {
[AssistantStreamEvents.ThreadRunCreated]: async (event) => {
await cache.set(cacheKey, `${thread_id}:${event.data.id}`, ten_minutes);
run_id = event.data.id;
sendInitialResponse();
},
};
const streamRunManager = new StreamRunManager({
req,
res,
openai,
handlers,
thread_id,
attachedFileIds,
responseMessage: openai.responseMessage,
// streamOptions: {
// },
});
await streamRunManager.runAssistant({
thread_id,
body,
});
response = streamRunManager;
};
await processRun();
logger.debug('[/assistants/chat/] response', {
run: response.run,
steps: response.steps,
});
if (response.run.status === RunStatus.CANCELLED) {
logger.debug('[/assistants/chat/] Run cancelled, handled by `abortRun`');
return res.end();
}
if (response.run.status === RunStatus.IN_PROGRESS) {
processRun(true);
}
completedRun = response.run;
/** @type {ResponseMessage} */
const responseMessage = {
...(response.responseMessage ?? response.finalMessage),
parentMessageId: userMessageId,
conversationId,
user: req.user.id,
assistant_id,
thread_id,
model: assistant_id,
endpoint,
};
sendMessage(res, {
final: true,
conversation,
requestMessage: {
parentMessageId,
thread_id,
},
});
res.end();
await saveAssistantMessage({ ...responseMessage, model });
if (parentMessageId === Constants.NO_PARENT && !_thread_id) {
addTitle(req, {
text,
responseText: response.text,
conversationId,
client,
});
}
await addThreadMetadata({
openai,
thread_id,
messageId: responseMessage.messageId,
messages: response.messages,
});
if (!response.run.usage) {
await sleep(3000);
completedRun = await openai.beta.threads.runs.retrieve(thread_id, response.run.id);
if (completedRun.usage) {
await recordUsage({
...completedRun.usage,
user: req.user.id,
model: completedRun.model ?? model,
conversationId,
});
}
} else {
await recordUsage({
...response.run.usage,
user: req.user.id,
model: response.run.model ?? model,
conversationId,
});
}
} catch (error) {
await handleError(error);
}
};
module.exports = chatV2;

View file

@ -0,0 +1,158 @@
const { EModelEndpoint, CacheKeys, defaultAssistantsVersion } = require('librechat-data-provider');
const {
initializeClient: initAzureClient,
} = require('~/server/services/Endpoints/azureAssistants');
const { initializeClient } = require('~/server/services/Endpoints/assistants');
const { getLogStores } = require('~/cache');
/**
* @param {Express.Request} req
* @param {string} [endpoint]
* @returns {Promise<string>}
*/
const getCurrentVersion = async (req, endpoint) => {
const index = req.baseUrl.lastIndexOf('/v');
let version = index !== -1 ? req.baseUrl.substring(index + 1, index + 3) : null;
if (!version && req.body.version) {
version = `v${req.body.version}`;
}
if (!version && endpoint) {
const cache = getLogStores(CacheKeys.CONFIG_STORE);
const cachedEndpointsConfig = await cache.get(CacheKeys.ENDPOINT_CONFIG);
version = `v${
cachedEndpointsConfig?.[endpoint]?.version ?? defaultAssistantsVersion[endpoint]
}`;
}
if (!version?.startsWith('v') && version.length !== 2) {
throw new Error(`[${req.baseUrl}] Invalid version: ${version}`);
}
return version;
};
/**
* Asynchronously lists assistants based on provided query parameters.
*
* Initializes the client with the current request and response objects and lists assistants
* according to the query parameters. This function abstracts the logic for non-Azure paths.
*
* @async
* @param {object} params - The parameters object.
* @param {object} params.req - The request object, used for initializing the client.
* @param {object} params.res - The response object, used for initializing the client.
* @param {string} params.version - The API version to use.
* @param {object} params.query - The query parameters to list assistants (e.g., limit, order).
* @returns {Promise<object>} A promise that resolves to the response from the `openai.beta.assistants.list` method call.
*/
const listAssistants = async ({ req, res, version, query }) => {
const { openai } = await getOpenAIClient({ req, res, version });
return openai.beta.assistants.list(query);
};
/**
* Asynchronously lists assistants for Azure configured groups.
*
* Iterates through Azure configured assistant groups, initializes the client with the current request and response objects,
* lists assistants based on the provided query parameters, and merges their data alongside the model information into a single array.
*
* @async
* @param {object} params - The parameters object.
* @param {object} params.req - The request object, used for initializing the client and manipulating the request body.
* @param {object} params.res - The response object, used for initializing the client.
* @param {string} params.version - The API version to use.
* @param {TAzureConfig} params.azureConfig - The Azure configuration object containing assistantGroups and groupMap.
* @param {object} params.query - The query parameters to list assistants (e.g., limit, order).
* @returns {Promise<AssistantListResponse>} A promise that resolves to an array of assistant data merged with their respective model information.
*/
const listAssistantsForAzure = async ({ req, res, version, azureConfig = {}, query }) => {
/** @type {Array<[string, TAzureModelConfig]>} */
const groupModelTuples = [];
const promises = [];
/** @type {Array<TAzureGroup>} */
const groups = [];
const { groupMap, assistantGroups } = azureConfig;
for (const groupName of assistantGroups) {
const group = groupMap[groupName];
groups.push(group);
const currentModelTuples = Object.entries(group?.models);
groupModelTuples.push(currentModelTuples);
/* The specified model is only necessary to
fetch assistants for the shared instance */
req.body.model = currentModelTuples[0][0];
promises.push(listAssistants({ req, res, version, query }));
}
const resolvedQueries = await Promise.all(promises);
const data = resolvedQueries.flatMap((res, i) =>
res.data.map((assistant) => {
const deploymentName = assistant.model;
const currentGroup = groups[i];
const currentModelTuples = groupModelTuples[i];
const firstModel = currentModelTuples[0][0];
if (currentGroup.deploymentName === deploymentName) {
return { ...assistant, model: firstModel };
}
for (const [model, modelConfig] of currentModelTuples) {
if (modelConfig.deploymentName === deploymentName) {
return { ...assistant, model };
}
}
return { ...assistant, model: firstModel };
}),
);
return {
first_id: data[0]?.id,
last_id: data[data.length - 1]?.id,
object: 'list',
has_more: false,
data,
};
};
async function getOpenAIClient({ req, res, endpointOption, initAppClient, overrideEndpoint }) {
let endpoint = overrideEndpoint ?? req.body.endpoint ?? req.query.endpoint;
const version = await getCurrentVersion(req, endpoint);
if (!endpoint) {
throw new Error(`[${req.baseUrl}] Endpoint is required`);
}
let result;
if (endpoint === EModelEndpoint.assistants) {
result = await initializeClient({ req, res, version, endpointOption, initAppClient });
} else if (endpoint === EModelEndpoint.azureAssistants) {
result = await initAzureClient({ req, res, version, endpointOption, initAppClient });
}
return result;
}
const fetchAssistants = async (req, res) => {
const { limit = 100, order = 'desc', after, before, endpoint } = req.query;
const version = await getCurrentVersion(req, endpoint);
const query = { limit, order, after, before };
/** @type {AssistantListResponse} */
let body;
if (endpoint === EModelEndpoint.assistants) {
({ body } = await listAssistants({ req, res, version, query }));
} else if (endpoint === EModelEndpoint.azureAssistants) {
const azureConfig = req.app.locals[EModelEndpoint.azureOpenAI];
body = await listAssistantsForAzure({ req, res, version, azureConfig, query });
}
return body;
};
module.exports = {
getOpenAIClient,
fetchAssistants,
getCurrentVersion,
};

View file

@ -0,0 +1,262 @@
const { FileContext } = require('librechat-data-provider');
const { getStrategyFunctions } = require('~/server/services/Files/strategies');
const { deleteAssistantActions } = require('~/server/services/ActionService');
const { uploadImageBuffer } = require('~/server/services/Files/process');
const { updateAssistant, getAssistants } = require('~/models/Assistant');
const { getOpenAIClient, fetchAssistants } = require('./helpers');
const { deleteFileByFilter } = require('~/models/File');
const { logger } = require('~/config');
/**
* Create an assistant.
* @route POST /assistants
* @param {AssistantCreateParams} req.body - The assistant creation parameters.
* @returns {Assistant} 201 - success response - application/json
*/
const createAssistant = async (req, res) => {
try {
const { openai } = await getOpenAIClient({ req, res });
const { tools = [], endpoint, ...assistantData } = req.body;
assistantData.tools = tools
.map((tool) => {
if (typeof tool !== 'string') {
return tool;
}
return req.app.locals.availableTools[tool];
})
.filter((tool) => tool);
let azureModelIdentifier = null;
if (openai.locals?.azureOptions) {
azureModelIdentifier = assistantData.model;
assistantData.model = openai.locals.azureOptions.azureOpenAIApiDeploymentName;
}
assistantData.metadata = {
author: req.user.id,
endpoint,
};
const assistant = await openai.beta.assistants.create(assistantData);
if (azureModelIdentifier) {
assistant.model = azureModelIdentifier;
}
logger.debug('/assistants/', assistant);
res.status(201).json(assistant);
} catch (error) {
logger.error('[/assistants] Error creating assistant', error);
res.status(500).json({ error: error.message });
}
};
/**
* Retrieves an assistant.
* @route GET /assistants/:id
* @param {string} req.params.id - Assistant identifier.
* @returns {Assistant} 200 - success response - application/json
*/
const retrieveAssistant = async (req, res) => {
try {
/* NOTE: not actually being used right now */
const { openai } = await getOpenAIClient({ req, res });
const assistant_id = req.params.id;
const assistant = await openai.beta.assistants.retrieve(assistant_id);
res.json(assistant);
} catch (error) {
logger.error('[/assistants/:id] Error retrieving assistant', error);
res.status(500).json({ error: error.message });
}
};
/**
* Modifies an assistant.
* @route PATCH /assistants/:id
* @param {object} req - Express Request
* @param {object} req.params - Request params
* @param {string} req.params.id - Assistant identifier.
* @param {AssistantUpdateParams} req.body - The assistant update parameters.
* @returns {Assistant} 200 - success response - application/json
*/
const patchAssistant = async (req, res) => {
try {
const { openai } = await getOpenAIClient({ req, res });
const assistant_id = req.params.id;
const { endpoint: _e, ...updateData } = req.body;
updateData.tools = (updateData.tools ?? [])
.map((tool) => {
if (typeof tool !== 'string') {
return tool;
}
return req.app.locals.availableTools[tool];
})
.filter((tool) => tool);
if (openai.locals?.azureOptions && updateData.model) {
updateData.model = openai.locals.azureOptions.azureOpenAIApiDeploymentName;
}
const updatedAssistant = await openai.beta.assistants.update(assistant_id, updateData);
res.json(updatedAssistant);
} catch (error) {
logger.error('[/assistants/:id] Error updating assistant', error);
res.status(500).json({ error: error.message });
}
};
/**
* Deletes an assistant.
* @route DELETE /assistants/:id
* @param {object} req - Express Request
* @param {object} req.params - Request params
* @param {string} req.params.id - Assistant identifier.
* @returns {Assistant} 200 - success response - application/json
*/
const deleteAssistant = async (req, res) => {
try {
const { openai } = await getOpenAIClient({ req, res });
const assistant_id = req.params.id;
const deletionStatus = await openai.beta.assistants.del(assistant_id);
if (deletionStatus?.deleted) {
await deleteAssistantActions({ req, assistant_id });
}
res.json(deletionStatus);
} catch (error) {
logger.error('[/assistants/:id] Error deleting assistant', error);
res.status(500).json({ error: 'Error deleting assistant' });
}
};
/**
* Returns a list of assistants.
* @route GET /assistants
* @param {object} req - Express Request
* @param {AssistantListParams} req.query - The assistant list parameters for pagination and sorting.
* @returns {AssistantListResponse} 200 - success response - application/json
*/
const listAssistants = async (req, res) => {
try {
const body = await fetchAssistants(req, res);
if (req.app.locals?.[req.query.endpoint]) {
/** @type {Partial<TAssistantEndpoint>} */
const assistantsConfig = req.app.locals[req.query.endpoint];
const { supportedIds, excludedIds } = assistantsConfig;
if (supportedIds?.length) {
body.data = body.data.filter((assistant) => supportedIds.includes(assistant.id));
} else if (excludedIds?.length) {
body.data = body.data.filter((assistant) => !excludedIds.includes(assistant.id));
}
}
res.json(body);
} catch (error) {
logger.error('[/assistants] Error listing assistants', error);
res.status(500).json({ message: 'Error listing assistants' });
}
};
/**
* Returns a list of the user's assistant documents (metadata saved to database).
* @route GET /assistants/documents
* @returns {AssistantDocument[]} 200 - success response - application/json
*/
const getAssistantDocuments = async (req, res) => {
try {
res.json(await getAssistants({ user: req.user.id }));
} catch (error) {
logger.error('[/assistants/documents] Error listing assistant documents', error);
res.status(500).json({ error: error.message });
}
};
/**
* Uploads and updates an avatar for a specific assistant.
* @route POST /avatar/:assistant_id
* @param {object} req - Express Request
* @param {object} req.params - Request params
* @param {string} req.params.assistant_id - The ID of the assistant.
* @param {Express.Multer.File} req.file - The avatar image file.
* @param {object} req.body - Request body
* @param {string} [req.body.metadata] - Optional metadata for the assistant's avatar.
* @returns {Object} 200 - success response - application/json
*/
const uploadAssistantAvatar = async (req, res) => {
try {
const { assistant_id } = req.params;
if (!assistant_id) {
return res.status(400).json({ message: 'Assistant ID is required' });
}
let { metadata: _metadata = '{}' } = req.body;
const { openai } = await getOpenAIClient({ req, res });
const image = await uploadImageBuffer({
req,
context: FileContext.avatar,
metadata: {
buffer: req.file.buffer,
},
});
try {
_metadata = JSON.parse(_metadata);
} catch (error) {
logger.error('[/avatar/:assistant_id] Error parsing metadata', error);
_metadata = {};
}
if (_metadata.avatar && _metadata.avatar_source) {
const { deleteFile } = getStrategyFunctions(_metadata.avatar_source);
try {
await deleteFile(req, { filepath: _metadata.avatar });
await deleteFileByFilter({ filepath: _metadata.avatar });
} catch (error) {
logger.error('[/avatar/:assistant_id] Error deleting old avatar', error);
}
}
const metadata = {
..._metadata,
avatar: image.filepath,
avatar_source: req.app.locals.fileStrategy,
};
const promises = [];
promises.push(
updateAssistant(
{ assistant_id },
{
avatar: {
filepath: image.filepath,
source: req.app.locals.fileStrategy,
},
user: req.user.id,
},
),
);
promises.push(openai.beta.assistants.update(assistant_id, { metadata }));
const resolved = await Promise.all(promises);
res.status(201).json(resolved[1]);
} catch (error) {
const message = 'An error occurred while updating the Assistant Avatar';
logger.error(message, error);
res.status(500).json({ message });
}
};
module.exports = {
createAssistant,
retrieveAssistant,
patchAssistant,
deleteAssistant,
listAssistants,
getAssistantDocuments,
uploadAssistantAvatar,
};

View file

@ -0,0 +1,208 @@
const { ToolCallTypes } = require('librechat-data-provider');
const { validateAndUpdateTool } = require('~/server/services/ActionService');
const { getOpenAIClient } = require('./helpers');
const { logger } = require('~/config');
/**
* Create an assistant.
* @route POST /assistants
* @param {AssistantCreateParams} req.body - The assistant creation parameters.
* @returns {Assistant} 201 - success response - application/json
*/
const createAssistant = async (req, res) => {
try {
/** @type {{ openai: OpenAIClient }} */
const { openai } = await getOpenAIClient({ req, res });
const { tools = [], endpoint, ...assistantData } = req.body;
assistantData.tools = tools
.map((tool) => {
if (typeof tool !== 'string') {
return tool;
}
return req.app.locals.availableTools[tool];
})
.filter((tool) => tool);
let azureModelIdentifier = null;
if (openai.locals?.azureOptions) {
azureModelIdentifier = assistantData.model;
assistantData.model = openai.locals.azureOptions.azureOpenAIApiDeploymentName;
}
assistantData.metadata = {
author: req.user.id,
endpoint,
};
const assistant = await openai.beta.assistants.create(assistantData);
if (azureModelIdentifier) {
assistant.model = azureModelIdentifier;
}
logger.debug('/assistants/', assistant);
res.status(201).json(assistant);
} catch (error) {
logger.error('[/assistants] Error creating assistant', error);
res.status(500).json({ error: error.message });
}
};
/**
* Modifies an assistant.
* @param {object} params
* @param {Express.Request} params.req
* @param {OpenAIClient} params.openai
* @param {string} params.assistant_id
* @param {AssistantUpdateParams} params.updateData
* @returns {Promise<Assistant>} The updated assistant.
*/
const updateAssistant = async ({ req, openai, assistant_id, updateData }) => {
const tools = [];
let hasFileSearch = false;
for (const tool of updateData.tools ?? []) {
let actualTool = typeof tool === 'string' ? req.app.locals.availableTools[tool] : tool;
if (!actualTool) {
continue;
}
if (actualTool.type === ToolCallTypes.FILE_SEARCH) {
hasFileSearch = true;
}
if (!actualTool.function) {
tools.push(actualTool);
continue;
}
const updatedTool = await validateAndUpdateTool({ req, tool: actualTool, assistant_id });
if (updatedTool) {
tools.push(updatedTool);
}
}
if (hasFileSearch && !updateData.tool_resources) {
const assistant = await openai.beta.assistants.retrieve(assistant_id);
updateData.tool_resources = assistant.tool_resources ?? null;
}
if (hasFileSearch && !updateData.tool_resources?.file_search) {
updateData.tool_resources = {
...(updateData.tool_resources ?? {}),
file_search: {
vector_store_ids: [],
},
};
}
updateData.tools = tools;
if (openai.locals?.azureOptions && updateData.model) {
updateData.model = openai.locals.azureOptions.azureOpenAIApiDeploymentName;
}
return await openai.beta.assistants.update(assistant_id, updateData);
};
/**
* Modifies an assistant with the resource file id.
* @param {object} params
* @param {Express.Request} params.req
* @param {OpenAIClient} params.openai
* @param {string} params.assistant_id
* @param {string} params.tool_resource
* @param {string} params.file_id
* @param {AssistantUpdateParams} params.updateData
* @returns {Promise<Assistant>} The updated assistant.
*/
const addResourceFileId = async ({ req, openai, assistant_id, tool_resource, file_id }) => {
const assistant = await openai.beta.assistants.retrieve(assistant_id);
const { tool_resources = {} } = assistant;
if (tool_resources[tool_resource]) {
tool_resources[tool_resource].file_ids.push(file_id);
} else {
tool_resources[tool_resource] = { file_ids: [file_id] };
}
delete assistant.id;
return await updateAssistant({
req,
openai,
assistant_id,
updateData: { tools: assistant.tools, tool_resources },
});
};
/**
* Deletes a file ID from an assistant's resource.
* @param {object} params
* @param {Express.Request} params.req
* @param {OpenAIClient} params.openai
* @param {string} params.assistant_id
* @param {string} [params.tool_resource]
* @param {string} params.file_id
* @param {AssistantUpdateParams} params.updateData
* @returns {Promise<Assistant>} The updated assistant.
*/
const deleteResourceFileId = async ({ req, openai, assistant_id, tool_resource, file_id }) => {
const assistant = await openai.beta.assistants.retrieve(assistant_id);
const { tool_resources = {} } = assistant;
if (tool_resource && tool_resources[tool_resource]) {
const resource = tool_resources[tool_resource];
const index = resource.file_ids.indexOf(file_id);
if (index !== -1) {
resource.file_ids.splice(index, 1);
}
} else {
for (const resourceKey in tool_resources) {
const resource = tool_resources[resourceKey];
const index = resource.file_ids.indexOf(file_id);
if (index !== -1) {
resource.file_ids.splice(index, 1);
break;
}
}
}
delete assistant.id;
return await updateAssistant({
req,
openai,
assistant_id,
updateData: { tools: assistant.tools, tool_resources },
});
};
/**
* Modifies an assistant.
* @route PATCH /assistants/:id
* @param {object} req - Express Request
* @param {object} req.params - Request params
* @param {string} req.params.id - Assistant identifier.
* @param {AssistantUpdateParams} req.body - The assistant update parameters.
* @returns {Assistant} 200 - success response - application/json
*/
const patchAssistant = async (req, res) => {
try {
const { openai } = await getOpenAIClient({ req, res });
const assistant_id = req.params.id;
const { endpoint: _e, ...updateData } = req.body;
updateData.tools = updateData.tools ?? [];
const updatedAssistant = await updateAssistant({ req, openai, assistant_id, updateData });
res.json(updatedAssistant);
} catch (error) {
logger.error('[/assistants/:id] Error updating assistant', error);
res.status(500).json({ error: error.message });
}
};
module.exports = {
patchAssistant,
createAssistant,
updateAssistant,
addResourceFileId,
deleteResourceFileId,
};