+ The specific chat model that will be used for this workspace. If + empty, will use the system LLM preference. +
++ The specific chat model that will be used for this workspace. If + empty, will use the system LLM preference. +
++ {" "} +
{workspace?.slug}
@@ -101,13 +102,7 @@ export default function WorkspaceSettings({ workspace }) {Total number of vectors in your vector database.
- {totalVectors !== null ? ( -- {totalVectors} -
- ) : ( -- Referenced {references} times. -
- )} -+ Referenced {references} times. +
+ )} diff --git a/frontend/src/components/WorkspaceChat/ChatContainer/ChatHistory/HistoricalMessage/Actions/index.jsx b/frontend/src/components/WorkspaceChat/ChatContainer/ChatHistory/HistoricalMessage/Actions/index.jsx new file mode 100644 index 000000000..12fa7dc73 --- /dev/null +++ b/frontend/src/components/WorkspaceChat/ChatContainer/ChatHistory/HistoricalMessage/Actions/index.jsx @@ -0,0 +1,43 @@ +import useCopyText from "@/hooks/useCopyText"; +import { Check, ClipboardText } from "@phosphor-icons/react"; +import { memo } from "react"; +import { Tooltip } from "react-tooltip"; + +const Actions = ({ message }) => { + return ( ++
Welcome to your new workspace.
+
To get started either{" "}
);
})}
-
{showing && (
` + - hljs.highlight(lang, str, true).value + + `+" ); } catch (__) {} } return ( - `++++ +${lang || ""}+` + + hljs.highlight(code, { language: lang, ignoreIllegals: true }).value + "` + - HTMLEncode(str) + + `+" ); }, }); -window.copySnippet = function (uuid = "") { - const target = document.getElementById(`code-${uuid}`); - const markdown = - target.parentElement?.parentElement?.querySelector( - "pre:first-of-type" - )?.innerText; - if (!markdown) return false; - - window.navigator.clipboard.writeText(markdown); - target.classList.add("text-green-500"); - const originalText = target.innerHTML; - target.innerText = "Copied!"; - target.setAttribute("disabled", true); - - setTimeout(() => { - target.classList.remove("text-green-500"); - target.innerHTML = originalText; - target.removeAttribute("disabled"); - }, 5000); -}; - export default function renderMarkdown(text = "") { return markdown.render(text); } diff --git a/frontend/src/utils/constants.js b/frontend/src/utils/constants.js index 11b8da976..2fde1ee00 100644 --- a/frontend/src/utils/constants.js +++ b/frontend/src/utils/constants.js @@ -7,3 +7,8 @@ export const COMPLETE_QUESTIONNAIRE = "anythingllm_completed_questionnaire"; export const USER_BACKGROUND_COLOR = "bg-historical-msg-user"; export const AI_BACKGROUND_COLOR = "bg-historical-msg-system"; + +export function fullApiUrl() { + if (API_BASE !== "/api") return API_BASE; + return `${window.location.origin}/api`; +} diff --git a/frontend/yarn.lock b/frontend/yarn.lock index e7b223df9..fa1e71331 100644 --- a/frontend/yarn.lock +++ b/frontend/yarn.lock @@ -365,6 +365,26 @@ resolved "https://registry.yarnpkg.com/@eslint/js/-/js-8.53.0.tgz#bea56f2ed2b5baea164348ff4d5a879f6f81f20d" integrity sha512-Kn7K8dx/5U6+cT1yEhpX1w4PCSg0M+XyRILPgvwcEBjerFWCwQj5sbr3/VmxqV0JGHCBCzyd6LxypEuehypY1w== +"@floating-ui/core@^1.5.3": + version "1.5.3" + resolved "https://registry.yarnpkg.com/@floating-ui/core/-/core-1.5.3.tgz#b6aa0827708d70971c8679a16cf680a515b8a52a" + integrity sha512-O0WKDOo0yhJuugCx6trZQj5jVJ9yR0ystG2JaNAemYUWce+pmM6WUEFIibnWyEJKdrDxhm75NoSRME35FNaM/Q== + dependencies: + "@floating-ui/utils" "^0.2.0" + +"@floating-ui/dom@^1.0.0": + version "1.5.4" + resolved "https://registry.yarnpkg.com/@floating-ui/dom/-/dom-1.5.4.tgz#28df1e1cb373884224a463235c218dcbd81a16bb" + integrity sha512-jByEsHIY+eEdCjnTVu+E3ephzTOzkQ8hgUfGwos+bg7NlH33Zc5uO+QHz1mrQUOgIKKDD1RtS201P9NvAfq3XQ== + dependencies: + "@floating-ui/core" "^1.5.3" + "@floating-ui/utils" "^0.2.0" + +"@floating-ui/utils@^0.2.0": + version "0.2.1" + resolved "https://registry.yarnpkg.com/@floating-ui/utils/-/utils-0.2.1.tgz#16308cea045f0fc777b6ff20a9f25474dd8293d2" + integrity sha512-9TANp6GPoMtYzQdt54kfAyMmz1+osLlXdg2ENroU7zzrtflTLrrC/lgrIfaSe+Wu0b89GKccT7vxXA0MoAIO+Q== + "@humanwhocodes/config-array@^0.11.13": version "0.11.13" resolved "https://registry.yarnpkg.com/@humanwhocodes/config-array/-/config-array-0.11.13.tgz#075dc9684f40a531d9b26b0822153c1e832ee297" @@ -846,6 +866,11 @@ chokidar@^3.5.3: optionalDependencies: fsevents "~2.3.2" +classnames@^2.3.0: + version "2.5.1" + resolved "https://registry.yarnpkg.com/classnames/-/classnames-2.5.1.tgz#ba774c614be0f016da105c858e7159eae8e7687b" + integrity sha512-saHYOzhIQs6wy2sVxTM6bUDsQO4F50V9RQ22qBpEdCW+I+/Wmke2HOl6lS6dTpdxVhb88/I6+Hs+438c3lfUow== + cliui@^8.0.1: version "8.0.1" resolved "https://registry.yarnpkg.com/cliui/-/cliui-8.0.1.tgz#0c04b075db02cbfe60dc8e6cf2f5486b1a3608aa" @@ -1021,6 +1046,11 @@ doctrine@^3.0.0: dependencies: esutils "^2.0.2" +dompurify@^3.0.8: + version "3.0.8" + resolved "https://registry.yarnpkg.com/dompurify/-/dompurify-3.0.8.tgz#e0021ab1b09184bc8af7e35c7dd9063f43a8a437" + integrity sha512-b7uwreMYL2eZhrSCRC4ahLTeZcPZxSmYfmcQGXGkXiZSNW1X85v+SDM5KsWcpivIiUBH47Ji7NtyUdpLeF5JZQ== + electron-to-chromium@^1.4.535: version "1.4.576" resolved "https://registry.yarnpkg.com/electron-to-chromium/-/electron-to-chromium-1.4.576.tgz#0c6940fdc0d60f7e34bd742b29d8fa847c9294d1" @@ -2538,6 +2568,14 @@ react-toastify@^9.1.3: dependencies: clsx "^1.1.1" +react-tooltip@^5.25.2: + version "5.25.2" + resolved "https://registry.yarnpkg.com/react-tooltip/-/react-tooltip-5.25.2.tgz#efb51845ec2e863045812ad1dc1927573922d629" + integrity sha512-MwZ3S9xcHpojZaKqjr5mTs0yp/YBPpKFcayY7MaaIIBr2QskkeeyelpY2YdGLxIMyEj4sxl0rGoK6dQIKvNLlw== + dependencies: + "@floating-ui/dom" "^1.0.0" + classnames "^2.3.0" + react@^18.2.0: version "18.2.0" resolved "https://registry.yarnpkg.com/react/-/react-18.2.0.tgz#555bd98592883255fa00de14f1151a917b5d77d5" diff --git a/pull_request_template.md b/pull_request_template.md new file mode 100644 index 000000000..1167880b1 --- /dev/null +++ b/pull_request_template.md @@ -0,0 +1,36 @@ + + ### Pull Request Type + + + +- [ ] ✨ feat +- [ ] 🐛 fix +- [ ] ♻️ refactor +- [ ] 💄 style +- [ ] 🔨 chore +- [ ] 📝 docs + +### Relevant Issues + + + +resolves #xxx + + +### What is in this change? + +Describe the changes in this PR that are impactful to the repo. + + +### Additional Information + +Add any other context about the Pull Request here that was not captured above. + +### Developer Validations + + + +- [ ] I ran `yarn lint` from the root of the repo & committed changes +- [ ] Relevant documentation has been updated +- [ ] I have tested my code functionality +- [ ] Docker build succeeds locally diff --git a/server/.env.example b/server/.env.example index 5b159a03d..23e20bb13 100644 --- a/server/.env.example +++ b/server/.env.example @@ -37,6 +37,14 @@ JWT_SECRET="my-random-string-for-seeding" # Please generate random string at lea # OLLAMA_MODEL_PREF='llama2' # OLLAMA_MODEL_TOKEN_LIMIT=4096 +# LLM_PROVIDER='togetherai' +# TOGETHER_AI_API_KEY='my-together-ai-key' +# TOGETHER_AI_MODEL_PREF='mistralai/Mixtral-8x7B-Instruct-v0.1' + +# LLM_PROVIDER='mistral' +# MISTRAL_API_KEY='example-mistral-ai-api-key' +# MISTRAL_MODEL_PREF='mistral-tiny' + ########################################### ######## Embedding API SElECTION ########## ########################################### @@ -82,6 +90,16 @@ VECTOR_DB="lancedb" # QDRANT_ENDPOINT="http://localhost:6333" # QDRANT_API_KEY= +# Enable all below if you are using vector database: Milvus. +# VECTOR_DB="milvus" +# MILVUS_ADDRESS="http://localhost:19530" +# MILVUS_USERNAME= +# MILVUS_PASSWORD= + +# Enable all below if you are using vector database: Zilliz Cloud. +# VECTOR_DB="zilliz" +# ZILLIZ_ENDPOINT="https://sample.api.gcp-us-west1.zillizcloud.com" +# ZILLIZ_API_TOKEN=api-token-here # CLOUD DEPLOYMENT VARIRABLES ONLY # AUTH_TOKEN="hunter2" # This is the password to your application if remote hosting. diff --git a/server/.gitignore b/server/.gitignore index be4af591d..0913f9663 100644 --- a/server/.gitignore +++ b/server/.gitignore @@ -3,6 +3,7 @@ storage/assets/* !storage/assets/anything-llm.png storage/documents/* +storage/tmp/* storage/vector-cache/*.json storage/exports storage/imports diff --git a/server/endpoints/api/document/index.js b/server/endpoints/api/document/index.js index a813e2df6..817043526 100644 --- a/server/endpoints/api/document/index.js +++ b/server/endpoints/api/document/index.js @@ -5,8 +5,13 @@ const { checkProcessorAlive, acceptedFileTypes, processDocument, + processLink, } = require("../../../utils/files/documentProcessor"); -const { viewLocalFiles } = require("../../../utils/files"); +const { + viewLocalFiles, + findDocumentInDocuments, +} = require("../../../utils/files"); +const { reqBody } = require("../../../utils/http"); const { handleUploads } = setupMulter(); function apiDocumentEndpoints(app) { @@ -20,7 +25,6 @@ function apiDocumentEndpoints(app) { /* #swagger.tags = ['Documents'] #swagger.description = 'Upload a new file to AnythingLLM to be parsed and prepared for embedding.' - #swagger.requestBody = { description: 'File to be uploaded.', required: true, @@ -47,6 +51,21 @@ function apiDocumentEndpoints(app) { example: { success: true, error: null, + documents: [ + { + "location": "custom-documents/anythingllm.txt-6e8be64c-c162-4b43-9997-b068c0071e8b.json", + "name": "anythingllm.txt-6e8be64c-c162-4b43-9997-b068c0071e8b.json", + "url": "file:///Users/tim/Documents/anything-llm/collector/hotdir/anythingllm.txt", + "title": "anythingllm.txt", + "docAuthor": "Unknown", + "description": "Unknown", + "docSource": "a text file uploaded by the user.", + "chunkSource": "anythingllm.txt", + "published": "1/16/2024, 3:07:00 PM", + "wordCount": 93, + "token_count_estimate": 115, + } + ] } } } @@ -72,16 +91,113 @@ function apiDocumentEndpoints(app) { .end(); } - const { success, reason } = await processDocument(originalname); + const { success, reason, documents } = + await processDocument(originalname); if (!success) { - response.status(500).json({ success: false, error: reason }).end(); + response + .status(500) + .json({ success: false, error: reason, documents }) + .end(); + return; } console.log( `Document ${originalname} uploaded processed and successfully. It is now available in documents.` ); await Telemetry.sendTelemetry("document_uploaded"); - response.status(200).json({ success: true, error: null }); + response.status(200).json({ success: true, error: null, documents }); + } catch (e) { + console.log(e.message, e); + response.sendStatus(500).end(); + } + } + ); + + app.post( + "/v1/document/upload-link", + [validApiKey], + async (request, response) => { + /* + #swagger.tags = ['Documents'] + #swagger.description = 'Upload a valid URL for AnythingLLM to scrape and prepare for embedding.' + #swagger.requestBody = { + description: 'Link of web address to be scraped.', + required: true, + type: 'file', + content: { + "application/json": { + schema: { + type: 'object', + example: { + "link": "https://useanything.com" + } + } + } + } + } + #swagger.responses[200] = { + content: { + "application/json": { + schema: { + type: 'object', + example: { + success: true, + error: null, + documents: [ + { + "id": "c530dbe6-bff1-4b9e-b87f-710d539d20bc", + "url": "file://useanything_com.html", + "title": "useanything_com.html", + "docAuthor": "no author found", + "description": "No description found.", + "docSource": "URL link uploaded by the user.", + "chunkSource": "https:useanything.com.html", + "published": "1/16/2024, 3:46:33 PM", + "wordCount": 252, + "pageContent": "AnythingLLM is the best....", + "token_count_estimate": 447, + "location": "custom-documents/url-useanything_com-c530dbe6-bff1-4b9e-b87f-710d539d20bc.json" + } + ] + } + } + } + } + } + #swagger.responses[403] = { + schema: { + "$ref": "#/definitions/InvalidAPIKey" + } + } + */ + try { + const { link } = reqBody(request); + const processingOnline = await checkProcessorAlive(); + + if (!processingOnline) { + response + .status(500) + .json({ + success: false, + error: `Document processing API is not online. Link ${link} will not be processed automatically.`, + }) + .end(); + } + + const { success, reason, documents } = await processLink(link); + if (!success) { + response + .status(500) + .json({ success: false, error: reason, documents }) + .end(); + return; + } + + console.log( + `Link ${link} uploaded processed and successfully. It is now available in documents.` + ); + await Telemetry.sendTelemetry("document_uploaded"); + response.status(200).json({ success: true, error: null, documents }); } catch (e) { console.log(e.message, e); response.sendStatus(500).end(); @@ -133,6 +249,61 @@ function apiDocumentEndpoints(app) { } }); + app.get("/v1/document/:docName", [validApiKey], async (request, response) => { + /* + #swagger.tags = ['Documents'] + #swagger.description = 'Get a single document by its unique AnythingLLM document name' + #swagger.parameters['docName'] = { + in: 'path', + description: 'Unique document name to find (name in /documents)', + required: true, + type: 'string' + } + #swagger.responses[200] = { + content: { + "application/json": { + schema: { + type: 'object', + example: { + "localFiles": { + "name": "documents", + "type": "folder", + items: [ + { + "name": "my-stored-document.txt-uuid1234.json", + "type": "file", + "id": "bb07c334-4dab-4419-9462-9d00065a49a1", + "url": "file://my-stored-document.txt", + "title": "my-stored-document.txt", + "cached": false + }, + ] + } + } + } + } + } + } + #swagger.responses[403] = { + schema: { + "$ref": "#/definitions/InvalidAPIKey" + } + } + */ + try { + const { docName } = request.params; + const document = await findDocumentInDocuments(docName); + if (!document) { + response.sendStatus(404).end(); + return; + } + response.status(200).json({ document }); + } catch (e) { + console.log(e.message, e); + response.sendStatus(500).end(); + } + }); + app.get( "/v1/document/accepted-file-types", [validApiKey], diff --git a/server/endpoints/api/system/index.js b/server/endpoints/api/system/index.js index 3548c3068..b18019b14 100644 --- a/server/endpoints/api/system/index.js +++ b/server/endpoints/api/system/index.js @@ -139,7 +139,7 @@ function apiSystemEndpoints(app) { */ try { const body = reqBody(request); - const { newValues, error } = updateENV(body); + const { newValues, error } = await updateENV(body); if (process.env.NODE_ENV === "production") await dumpENV(); response.status(200).json({ newValues, error }); } catch (e) { diff --git a/server/endpoints/api/workspace/index.js b/server/endpoints/api/workspace/index.js index 032fe41c3..c1642ce4a 100644 --- a/server/endpoints/api/workspace/index.js +++ b/server/endpoints/api/workspace/index.js @@ -11,6 +11,11 @@ const { const { getVectorDbClass } = require("../../../utils/helpers"); const { multiUserMode, reqBody } = require("../../../utils/http"); const { validApiKey } = require("../../../utils/middleware/validApiKey"); +const { + streamChatWithWorkspace, + writeResponseChunk, + VALID_CHAT_MODE, +} = require("../../../utils/chats/stream"); function apiWorkspaceEndpoints(app) { if (!app) return; @@ -196,10 +201,11 @@ function apiWorkspaceEndpoints(app) { return; } - await WorkspaceChats.delete({ workspaceId: Number(workspace.id) }); - await DocumentVectors.deleteForWorkspace(Number(workspace.id)); - await Document.delete({ workspaceId: Number(workspace.id) }); - await Workspace.delete({ id: Number(workspace.id) }); + const workspaceId = Number(workspace.id); + await WorkspaceChats.delete({ workspaceId: workspaceId }); + await DocumentVectors.deleteForWorkspace(workspaceId); + await Document.delete({ workspaceId: workspaceId }); + await Workspace.delete({ id: workspaceId }); try { await VectorDb["delete-namespace"]({ namespace: slug }); } catch (e) { @@ -375,8 +381,8 @@ function apiWorkspaceEndpoints(app) { content: { "application/json": { example: { - adds: [], - deletes: ["custom-documents/anythingllm-hash.json"] + adds: ["custom-documents/my-pdf.pdf-hash.json"], + deletes: ["custom-documents/anythingllm.txt-hash.json"] } } } @@ -441,7 +447,7 @@ function apiWorkspaceEndpoints(app) { #swagger.tags = ['Workspaces'] #swagger.description = 'Execute a chat with a workspace' #swagger.requestBody = { - description: 'prompt to send to the workspace and the type of conversation (query or chat).', + description: 'Send a prompt to the workspace and the type of conversation (query or chat).+++ +` + + HTMLEncode(code) + "
Query: Will not use LLM unless there are relevant sources from vectorDB & does not recall chat history.
Chat: Uses LLM general knowledge w/custom embeddings to produce output, uses rolling chat history.', required: true, type: 'object', content: { @@ -482,7 +488,28 @@ function apiWorkspaceEndpoints(app) { const workspace = await Workspace.get({ slug }); if (!workspace) { - response.sendStatus(400).end(); + response.status(400).json({ + id: uuidv4(), + type: "abort", + textResponse: null, + sources: [], + close: true, + error: `Workspace ${slug} is not a valid workspace.`, + }); + return; + } + + if (!message?.length || !VALID_CHAT_MODE.includes(mode)) { + response.status(400).json({ + id: uuidv4(), + type: "abort", + textResponse: null, + sources: [], + close: true, + error: !message?.length + ? "message parameter cannot be empty." + : `${mode} is not a valid mode.`, + }); return; } @@ -505,6 +532,126 @@ function apiWorkspaceEndpoints(app) { } } ); + + app.post( + "/v1/workspace/:slug/stream-chat", + [validApiKey], + async (request, response) => { + /* + #swagger.tags = ['Workspaces'] + #swagger.description = 'Execute a streamable chat with a workspace' + #swagger.requestBody = { + description: 'Send a prompt to the workspace and the type of conversation (query or chat).
Query: Will not use LLM unless there are relevant sources from vectorDB & does not recall chat history.
Chat: Uses LLM general knowledge w/custom embeddings to produce output, uses rolling chat history.', + required: true, + type: 'object', + content: { + "application/json": { + example: { + message: "What is AnythingLLM?", + mode: "query | chat" + } + } + } + } + #swagger.responses[200] = { + content: { + "text/event-stream": { + schema: { + type: 'array', + example: [ + { + id: 'uuid-123', + type: "abort | textResponseChunk", + textResponse: "First chunk", + sources: [], + close: false, + error: "null | text string of the failure mode." + }, + { + id: 'uuid-123', + type: "abort | textResponseChunk", + textResponse: "chunk two", + sources: [], + close: false, + error: "null | text string of the failure mode." + }, + { + id: 'uuid-123', + type: "abort | textResponseChunk", + textResponse: "final chunk of LLM output!", + sources: [{title: "anythingllm.txt", chunk: "This is a context chunk used in the answer of the prompt by the LLM. This will only return in the final chunk."}], + close: true, + error: "null | text string of the failure mode." + } + ] + } + } + } + } + #swagger.responses[403] = { + schema: { + "$ref": "#/definitions/InvalidAPIKey" + } + } + */ + try { + const { slug } = request.params; + const { message, mode = "query" } = reqBody(request); + const workspace = await Workspace.get({ slug }); + + if (!workspace) { + response.status(400).json({ + id: uuidv4(), + type: "abort", + textResponse: null, + sources: [], + close: true, + error: `Workspace ${slug} is not a valid workspace.`, + }); + return; + } + + if (!message?.length || !VALID_CHAT_MODE.includes(mode)) { + response.status(400).json({ + id: uuidv4(), + type: "abort", + textResponse: null, + sources: [], + close: true, + error: !message?.length + ? "Message is empty" + : `${mode} is not a valid mode.`, + }); + return; + } + + response.setHeader("Cache-Control", "no-cache"); + response.setHeader("Content-Type", "text/event-stream"); + response.setHeader("Access-Control-Allow-Origin", "*"); + response.setHeader("Connection", "keep-alive"); + response.flushHeaders(); + + await streamChatWithWorkspace(response, workspace, message, mode); + await Telemetry.sendTelemetry("sent_chat", { + LLMSelection: process.env.LLM_PROVIDER || "openai", + Embedder: process.env.EMBEDDING_ENGINE || "inherit", + VectorDbSelection: process.env.VECTOR_DB || "pinecone", + }); + response.end(); + } catch (e) { + console.error(e); + writeResponseChunk(response, { + id: uuidv4(), + type: "abort", + textResponse: null, + sources: [], + close: true, + error: e.message, + }); + response.end(); + } + } + ); } module.exports = { apiWorkspaceEndpoints }; diff --git a/server/endpoints/chat.js b/server/endpoints/chat.js index d0a2923c5..adfec0ec3 100644 --- a/server/endpoints/chat.js +++ b/server/endpoints/chat.js @@ -1,7 +1,6 @@ const { v4: uuidv4 } = require("uuid"); const { reqBody, userFromSession, multiUserMode } = require("../utils/http"); const { Workspace } = require("../models/workspace"); -const { chatWithWorkspace } = require("../utils/chats"); const { validatedRequest } = require("../utils/middleware/validatedRequest"); const { WorkspaceChats } = require("../models/workspaceChats"); const { SystemSettings } = require("../models/systemSettings"); @@ -9,6 +8,7 @@ const { Telemetry } = require("../models/telemetry"); const { streamChatWithWorkspace, writeResponseChunk, + VALID_CHAT_MODE, } = require("../utils/chats/stream"); function chatEndpoints(app) { @@ -32,6 +32,20 @@ function chatEndpoints(app) { return; } + if (!message?.length || !VALID_CHAT_MODE.includes(mode)) { + response.status(400).json({ + id: uuidv4(), + type: "abort", + textResponse: null, + sources: [], + close: true, + error: !message?.length + ? "Message is empty." + : `${mode} is not a valid mode.`, + }); + return; + } + response.setHeader("Cache-Control", "no-cache"); response.setHeader("Content-Type", "text/event-stream"); response.setHeader("Access-Control-Allow-Origin", "*"); @@ -95,85 +109,6 @@ function chatEndpoints(app) { } } ); - - app.post( - "/workspace/:slug/chat", - [validatedRequest], - async (request, response) => { - try { - const user = await userFromSession(request, response); - const { slug } = request.params; - const { message, mode = "query" } = reqBody(request); - - const workspace = multiUserMode(response) - ? await Workspace.getWithUser(user, { slug }) - : await Workspace.get({ slug }); - - if (!workspace) { - response.sendStatus(400).end(); - return; - } - - if (multiUserMode(response) && user.role !== "admin") { - const limitMessagesSetting = await SystemSettings.get({ - label: "limit_user_messages", - }); - const limitMessages = limitMessagesSetting?.value === "true"; - - if (limitMessages) { - const messageLimitSetting = await SystemSettings.get({ - label: "message_limit", - }); - const systemLimit = Number(messageLimitSetting?.value); - - if (!!systemLimit) { - const currentChatCount = await WorkspaceChats.count({ - user_id: user.id, - createdAt: { - gte: new Date(new Date() - 24 * 60 * 60 * 1000), - }, - }); - - if (currentChatCount >= systemLimit) { - response.status(500).json({ - id: uuidv4(), - type: "abort", - textResponse: null, - sources: [], - close: true, - error: `You have met your maximum 24 hour chat quota of ${systemLimit} chats set by the instance administrators. Try again later.`, - }); - return; - } - } - } - } - - const result = await chatWithWorkspace(workspace, message, mode, user); - await Telemetry.sendTelemetry( - "sent_chat", - { - multiUserMode: multiUserMode(response), - LLMSelection: process.env.LLM_PROVIDER || "openai", - Embedder: process.env.EMBEDDING_ENGINE || "inherit", - VectorDbSelection: process.env.VECTOR_DB || "pinecone", - }, - user?.id - ); - response.status(200).json({ ...result }); - } catch (e) { - console.error(e); - response.status(500).json({ - id: uuidv4(), - type: "abort", - textResponse: null, - sources: [], - close: true, - error: e.message, - }); - } - } - ); } module.exports = { chatEndpoints }; diff --git a/server/endpoints/invite.js b/server/endpoints/invite.js index 08f9a14e9..4fd8d1545 100644 --- a/server/endpoints/invite.js +++ b/server/endpoints/invite.js @@ -33,7 +33,7 @@ function inviteEndpoints(app) { app.post("/invite/:code", async (request, response) => { try { const { code } = request.params; - const userParams = reqBody(request); + const { username, password } = reqBody(request); const invite = await Invite.get({ code }); if (!invite || invite.status !== "pending") { response @@ -42,7 +42,11 @@ function inviteEndpoints(app) { return; } - const { user, error } = await User.create(userParams); + const { user, error } = await User.create({ + username, + password, + role: "default", + }); if (!user) { console.error("Accepting invite:", error); response diff --git a/server/endpoints/system.js b/server/endpoints/system.js index decde3249..27f6c95cb 100644 --- a/server/endpoints/system.js +++ b/server/endpoints/system.js @@ -1,13 +1,14 @@ const path = require("path"); +const fs = require("fs"); process.env.NODE_ENV === "development" ? require("dotenv").config({ path: `.env.${process.env.NODE_ENV}` }) : require("dotenv").config({ - path: process.env.STORAGE_DIR - ? path.resolve(process.env.STORAGE_DIR, ".env") - : path.resolve(__dirname, ".env"), - }); + path: process.env.STORAGE_DIR + ? path.resolve(process.env.STORAGE_DIR, ".env") + : path.resolve(__dirname, ".env"), + }); -const { viewLocalFiles } = require("../utils/files"); +const { viewLocalFiles, normalizePath } = require("../utils/files"); const { exportData, unpackAndOverwriteImport } = require("../utils/files/data"); const { checkProcessorAlive, @@ -21,6 +22,7 @@ const { makeJWT, userFromSession, multiUserMode, + queryParams, } = require("../utils/http"); const { setupDataImports, @@ -34,7 +36,6 @@ const { validatedRequest } = require("../utils/middleware/validatedRequest"); const { handleImports } = setupDataImports(); const { handleLogoUploads } = setupLogoUploads(); const { handlePfpUploads } = setupPfpUploads(); -const fs = require("fs"); const { getDefaultFilename, determineLogoFilepath, @@ -111,6 +112,8 @@ function systemEndpoints(app) { app.post("/request-token", async (request, response) => { try { + const bcrypt = require("bcrypt"); + if (await SystemSettings.isMultiUserMode()) { const { username, password } = reqBody(request); const existingUser = await User.get({ username }); @@ -125,7 +128,6 @@ function systemEndpoints(app) { return; } - const bcrypt = require("bcrypt"); if (!bcrypt.compareSync(password, existingUser.password)) { response.status(200).json({ user: null, @@ -163,7 +165,12 @@ function systemEndpoints(app) { return; } else { const { password } = reqBody(request); - if (password !== process.env.AUTH_TOKEN) { + if ( + !bcrypt.compareSync( + password, + bcrypt.hashSync(process.env.AUTH_TOKEN, 10) + ) + ) { response.status(401).json({ valid: false, token: null, @@ -185,16 +192,23 @@ function systemEndpoints(app) { } }); - app.get("/system/system-vectors", [validatedRequest], async (_, response) => { - try { - const VectorDb = getVectorDbClass(); - const vectorCount = await VectorDb.totalVectors(); - response.status(200).json({ vectorCount }); - } catch (e) { - console.log(e.message, e); - response.sendStatus(500).end(); + app.get( + "/system/system-vectors", + [validatedRequest], + async (request, response) => { + try { + const query = queryParams(request); + const VectorDb = getVectorDbClass(); + const vectorCount = !!query.slug + ? await VectorDb.namespaceCount(query.slug) + : await VectorDb.totalVectors(); + response.status(200).json({ vectorCount }); + } catch (e) { + console.log(e.message, e); + response.sendStatus(500).end(); + } } - }); + ); app.delete( "/system/remove-document", @@ -274,8 +288,14 @@ function systemEndpoints(app) { [validatedRequest, flexUserRoleValid], async (request, response) => { try { + const user = await userFromSession(request, response); + if (!!user && user.role !== "admin") { + response.sendStatus(401).end(); + return; + } + const body = reqBody(request); - const { newValues, error } = updateENV(body); + const { newValues, error } = await updateENV(body); if (process.env.NODE_ENV === "production") await dumpENV(); response.status(200).json({ newValues, error }); } catch (e) { @@ -297,7 +317,7 @@ function systemEndpoints(app) { } const { usePassword, newPassword } = reqBody(request); - const { error } = updateENV( + const { error } = await updateENV( { AuthToken: usePassword ? newPassword : "", JWTSecret: usePassword ? v4() : "", @@ -340,7 +360,7 @@ function systemEndpoints(app) { message_limit: 25, }); - updateENV( + await updateENV( { AuthToken: "", JWTSecret: process.env.JWT_SECRET || v4(), @@ -374,21 +394,23 @@ function systemEndpoints(app) { } }); - app.get("/system/data-export", [validatedRequest], async (_, response) => { - try { - const { filename, error } = await exportData(); - response.status(200).json({ filename, error }); - } catch (e) { - console.log(e.message, e); - response.sendStatus(500).end(); + app.get( + "/system/data-export", + [validatedRequest, flexUserRoleValid], + async (_, response) => { + try { + const { filename, error } = await exportData(); + response.status(200).json({ filename, error }); + } catch (e) { + console.log(e.message, e); + response.sendStatus(500).end(); + } } - }); + ); app.get("/system/data-exports/:filename", (request, response) => { const exportLocation = __dirname + "/../storage/exports/"; - const sanitized = path - .normalize(request.params.filename) - .replace(/^(\.\.(\/|\\|$))+/, ""); + const sanitized = normalizePath(request.params.filename); const finalDestination = path.join(exportLocation, sanitized); if (!fs.existsSync(finalDestination)) { @@ -489,7 +511,8 @@ function systemEndpoints(app) { } const userRecord = await User.get({ id: user.id }); - const oldPfpFilename = userRecord.pfpFilename; + const oldPfpFilename = normalizePath(userRecord.pfpFilename); + console.log("oldPfpFilename", oldPfpFilename); if (oldPfpFilename) { const oldPfpPath = path.join( @@ -523,7 +546,7 @@ function systemEndpoints(app) { try { const user = await userFromSession(request, response); const userRecord = await User.get({ id: user.id }); - const oldPfpFilename = userRecord.pfpFilename; + const oldPfpFilename = normalizePath(userRecord.pfpFilename); console.log("oldPfpFilename", oldPfpFilename); if (oldPfpFilename) { const oldPfpPath = path.join( diff --git a/server/models/systemSettings.js b/server/models/systemSettings.js index ac5c3fc11..95b362ca8 100644 --- a/server/models/systemSettings.js +++ b/server/models/systemSettings.js @@ -60,6 +60,19 @@ const SystemSettings = { QdrantApiKey: process.env.QDRANT_API_KEY, } : {}), + ...(vectorDB === "milvus" + ? { + MilvusAddress: process.env.MILVUS_ADDRESS, + MilvusUsername: process.env.MILVUS_USERNAME, + MilvusPassword: !!process.env.MILVUS_PASSWORD, + } + : {}), + ...(vectorDB === "zilliz" + ? { + ZillizEndpoint: process.env.ZILLIZ_ENDPOINT, + ZillizApiToken: process.env.ZILLIZ_API_TOKEN, + } + : {}), LLMProvider: llmProvider, ...(llmProvider === "openai" ? { @@ -144,9 +157,40 @@ const SystemSettings = { AzureOpenAiEmbeddingModelPref: process.env.EMBEDDING_MODEL_PREF, } : {}), + ...(llmProvider === "togetherai" + ? { + TogetherAiApiKey: !!process.env.TOGETHER_AI_API_KEY, + TogetherAiModelPref: process.env.TOGETHER_AI_MODEL_PREF, + + // For embedding credentials when ollama is selected. + OpenAiKey: !!process.env.OPEN_AI_KEY, + AzureOpenAiEndpoint: process.env.AZURE_OPENAI_ENDPOINT, + AzureOpenAiKey: !!process.env.AZURE_OPENAI_KEY, + AzureOpenAiEmbeddingModelPref: process.env.EMBEDDING_MODEL_PREF, + } + : {}), + ...(llmProvider === "mistral" + ? { + MistralApiKey: !!process.env.MISTRAL_API_KEY, + MistralModelPref: process.env.MISTRAL_MODEL_PREF, + + // For embedding credentials when mistral is selected. + OpenAiKey: !!process.env.OPEN_AI_KEY, + AzureOpenAiEndpoint: process.env.AZURE_OPENAI_ENDPOINT, + AzureOpenAiKey: !!process.env.AZURE_OPENAI_KEY, + AzureOpenAiEmbeddingModelPref: process.env.EMBEDDING_MODEL_PREF, + } + : {}), ...(llmProvider === "native" ? { NativeLLMModelPref: process.env.NATIVE_LLM_MODEL_PREF, + NativeLLMTokenLimit: process.env.NATIVE_LLM_MODEL_TOKEN_LIMIT, + + // For embedding credentials when ollama is selected. + OpenAiKey: !!process.env.OPEN_AI_KEY, + AzureOpenAiEndpoint: process.env.AZURE_OPENAI_ENDPOINT, + AzureOpenAiKey: !!process.env.AZURE_OPENAI_KEY, + AzureOpenAiEmbeddingModelPref: process.env.EMBEDDING_MODEL_PREF, } : {}), }; diff --git a/server/models/welcomeMessages.js b/server/models/welcomeMessages.js index 43e2d3f96..88393f36c 100644 --- a/server/models/welcomeMessages.js +++ b/server/models/welcomeMessages.js @@ -31,7 +31,10 @@ const WelcomeMessages = { await prisma.welcome_messages.deleteMany({}); // Delete all existing messages // Create new messages + // We create each message individually because prisma + // with sqlite does not support createMany() for (const [index, message] of messages.entries()) { + if (!message.response) continue; await prisma.welcome_messages.create({ data: { user: message.user, diff --git a/server/models/workspace.js b/server/models/workspace.js index 9139c25e9..6de8053e9 100644 --- a/server/models/workspace.js +++ b/server/models/workspace.js @@ -14,6 +14,7 @@ const Workspace = { "lastUpdatedAt", "openAiPrompt", "similarityThreshold", + "chatModel", ], new: async function (name = null, creatorId = null) { @@ -191,6 +192,20 @@ const Workspace = { return { success: false, error: error.message }; } }, + + resetWorkspaceChatModels: async () => { + try { + await prisma.workspaces.updateMany({ + data: { + chatModel: null, + }, + }); + return { success: true, error: null }; + } catch (error) { + console.error("Error resetting workspace chat models:", error.message); + return { success: false, error: error.message }; + } + }, }; module.exports = { Workspace }; diff --git a/server/package.json b/server/package.json index 0e2d909c8..9761125a4 100644 --- a/server/package.json +++ b/server/package.json @@ -27,7 +27,8 @@ "@pinecone-database/pinecone": "^0.1.6", "@prisma/client": "5.3.0", "@qdrant/js-client-rest": "^1.4.0", - "@xenova/transformers": "^2.10.0", + "@xenova/transformers": "^2.14.0", + "@zilliz/milvus2-sdk-node": "^2.3.5", "archiver": "^5.3.1", "bcrypt": "^5.1.0", "body-parser": "^1.20.2", diff --git a/server/prisma/migrations/20240113013409_init/migration.sql b/server/prisma/migrations/20240113013409_init/migration.sql new file mode 100644 index 000000000..09b9448ec --- /dev/null +++ b/server/prisma/migrations/20240113013409_init/migration.sql @@ -0,0 +1,2 @@ +-- AlterTable +ALTER TABLE "workspaces" ADD COLUMN "chatModel" TEXT; diff --git a/server/prisma/schema.prisma b/server/prisma/schema.prisma index 579859d27..8d4d13721 100644 --- a/server/prisma/schema.prisma +++ b/server/prisma/schema.prisma @@ -93,6 +93,7 @@ model workspaces { lastUpdatedAt DateTime @default(now()) openAiPrompt String? similarityThreshold Float? @default(0.25) + chatModel String? workspace_users workspace_users[] documents workspace_documents[] } diff --git a/server/swagger/init.js b/server/swagger/init.js index c84daf323..b68e3249c 100644 --- a/server/swagger/init.js +++ b/server/swagger/init.js @@ -1,4 +1,6 @@ const swaggerAutogen = require('swagger-autogen')({ openapi: '3.0.0' }); +const fs = require('fs') +const path = require('path') const doc = { info: { @@ -6,6 +8,8 @@ const doc = { title: 'AnythingLLM Developer API', description: 'API endpoints that enable programmatic reading, writing, and updating of your AnythingLLM instance. UI supplied by Swagger.io.', }, + // Swagger-autogen does not allow us to use relative paths as these will resolve to + // http:///api in the openapi.json file, so we need to monkey-patch this post-generation. host: '/api', schemes: ['http'], securityDefinitions: { @@ -25,7 +29,7 @@ const doc = { } }; -const outputFile = './openapi.json'; +const outputFile = path.resolve(__dirname, './openapi.json'); const endpointsFiles = [ '../endpoints/api/auth/index.js', '../endpoints/api/admin/index.js', @@ -34,4 +38,14 @@ const endpointsFiles = [ '../endpoints/api/system/index.js', ]; -swaggerAutogen(outputFile, endpointsFiles, doc) \ No newline at end of file +swaggerAutogen(outputFile, endpointsFiles, doc) + .then(({ data }) => { + const openApiSpec = { + ...data, + servers: [{ + url: "/api" + }] + } + fs.writeFileSync(outputFile, JSON.stringify(openApiSpec, null, 2), { encoding: 'utf-8', flag: 'w' }); + console.log(`Swagger-autogen: \x1b[32mPatched servers.url ✔\x1b[0m`) + }) \ No newline at end of file diff --git a/server/swagger/openapi.json b/server/swagger/openapi.json index cb065522e..c7532059d 100644 --- a/server/swagger/openapi.json +++ b/server/swagger/openapi.json @@ -7,7 +7,7 @@ }, "servers": [ { - "url": "http:///api/" + "url": "/api" } ], "paths": { @@ -845,7 +845,22 @@ "type": "object", "example": { "success": true, - "error": null + "error": null, + "documents": [ + { + "location": "custom-documents/anythingllm.txt-6e8be64c-c162-4b43-9997-b068c0071e8b.json", + "name": "anythingllm.txt-6e8be64c-c162-4b43-9997-b068c0071e8b.json", + "url": "file://Users/tim/Documents/anything-llm/collector/hotdir/anythingllm.txt", + "title": "anythingllm.txt", + "docAuthor": "Unknown", + "description": "Unknown", + "docSource": "a text file uploaded by the user.", + "chunkSource": "anythingllm.txt", + "published": "1/16/2024, 3:07:00 PM", + "wordCount": 93, + "token_count_estimate": 115 + } + ] } } } @@ -890,6 +905,88 @@ } } }, + "/v1/document/upload-link": { + "post": { + "tags": [ + "Documents" + ], + "description": "Upload a valid URL for AnythingLLM to scrape and prepare for embedding.", + "parameters": [ + { + "name": "Authorization", + "in": "header", + "schema": { + "type": "string" + } + } + ], + "responses": { + "200": { + "description": "OK", + "content": { + "application/json": { + "schema": { + "type": "object", + "example": { + "success": true, + "error": null, + "documents": [ + { + "id": "c530dbe6-bff1-4b9e-b87f-710d539d20bc", + "url": "file://useanything_com.html", + "title": "useanything_com.html", + "docAuthor": "no author found", + "description": "No description found.", + "docSource": "URL link uploaded by the user.", + "chunkSource": "https:useanything.com.html", + "published": "1/16/2024, 3:46:33 PM", + "wordCount": 252, + "pageContent": "AnythingLLM is the best....", + "token_count_estimate": 447, + "location": "custom-documents/url-useanything_com-c530dbe6-bff1-4b9e-b87f-710d539d20bc.json" + } + ] + } + } + } + } + }, + "403": { + "description": "Forbidden", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/InvalidAPIKey" + } + }, + "application/xml": { + "schema": { + "$ref": "#/components/schemas/InvalidAPIKey" + } + } + } + }, + "500": { + "description": "Internal Server Error" + } + }, + "requestBody": { + "description": "Link of web address to be scraped.", + "required": true, + "type": "file", + "content": { + "application/json": { + "schema": { + "type": "object", + "example": { + "link": "https://useanything.com" + } + } + } + } + } + } + }, "/v1/documents": { "get": { "tags": [ @@ -953,6 +1050,81 @@ } } }, + "/v1/document/{docName}": { + "get": { + "tags": [ + "Documents" + ], + "description": "Get a single document by its unique AnythingLLM document name", + "parameters": [ + { + "name": "docName", + "in": "path", + "required": true, + "schema": { + "type": "string" + }, + "description": "Unique document name to find (name in /documents)" + }, + { + "name": "Authorization", + "in": "header", + "schema": { + "type": "string" + } + } + ], + "responses": { + "200": { + "description": "OK", + "content": { + "application/json": { + "schema": { + "type": "object", + "example": { + "localFiles": { + "name": "documents", + "type": "folder", + "items": [ + { + "name": "my-stored-document.txt-uuid1234.json", + "type": "file", + "id": "bb07c334-4dab-4419-9462-9d00065a49a1", + "url": "file://my-stored-document.txt", + "title": "my-stored-document.txt", + "cached": false + } + ] + } + } + } + } + } + }, + "403": { + "description": "Forbidden", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/InvalidAPIKey" + } + }, + "application/xml": { + "schema": { + "$ref": "#/components/schemas/InvalidAPIKey" + } + } + } + }, + "404": { + "description": "Not Found" + }, + "500": { + "description": "Internal Server Error" + } + } + } + }, "/v1/document/accepted-file-types": { "get": { "tags": [ @@ -1518,9 +1690,11 @@ "content": { "application/json": { "example": { - "adds": [], + "adds": [ + "custom-documents/my-pdf.pdf-hash.json" + ], "deletes": [ - "custom-documents/anythingllm-hash.json" + "custom-documents/anythingllm.txt-hash.json" ] } } @@ -1598,7 +1772,106 @@ } }, "requestBody": { - "description": "prompt to send to the workspace and the type of conversation (query or chat).", + "description": "Send a prompt to the workspace and the type of conversation (query or chat).
Query: Will not use LLM unless there are relevant sources from vectorDB & does not recall chat history.
Chat: Uses LLM general knowledge w/custom embeddings to produce output, uses rolling chat history.", + "required": true, + "type": "object", + "content": { + "application/json": { + "example": { + "message": "What is AnythingLLM?", + "mode": "query | chat" + } + } + } + } + } + }, + "/v1/workspace/{slug}/stream-chat": { + "post": { + "tags": [ + "Workspaces" + ], + "description": "Execute a streamable chat with a workspace", + "parameters": [ + { + "name": "slug", + "in": "path", + "required": true, + "schema": { + "type": "string" + } + }, + { + "name": "Authorization", + "in": "header", + "schema": { + "type": "string" + } + } + ], + "responses": { + "200": { + "content": { + "text/event-stream": { + "schema": { + "type": "array", + "example": [ + { + "id": "uuid-123", + "type": "abort | textResponseChunk", + "textResponse": "First chunk", + "sources": [], + "close": false, + "error": "null | text string of the failure mode." + }, + { + "id": "uuid-123", + "type": "abort | textResponseChunk", + "textResponse": "chunk two", + "sources": [], + "close": false, + "error": "null | text string of the failure mode." + }, + { + "id": "uuid-123", + "type": "abort | textResponseChunk", + "textResponse": "final chunk of LLM output!", + "sources": [ + { + "title": "anythingllm.txt", + "chunk": "This is a context chunk used in the answer of the prompt by the LLM. This will only return in the final chunk." + } + ], + "close": true, + "error": "null | text string of the failure mode." + } + ] + } + } + }, + "description": "OK" + }, + "400": { + "description": "Bad Request" + }, + "403": { + "description": "Forbidden", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/InvalidAPIKey" + } + }, + "application/xml": { + "schema": { + "$ref": "#/components/schemas/InvalidAPIKey" + } + } + } + } + }, + "requestBody": { + "description": "Send a prompt to the workspace and the type of conversation (query or chat).
Query: Will not use LLM unless there are relevant sources from vectorDB & does not recall chat history.
Chat: Uses LLM general knowledge w/custom embeddings to produce output, uses rolling chat history.", "required": true, "type": "object", "content": { diff --git a/server/utils/AiProviders/anthropic/index.js b/server/utils/AiProviders/anthropic/index.js index 709333231..56d3a80f0 100644 --- a/server/utils/AiProviders/anthropic/index.js +++ b/server/utils/AiProviders/anthropic/index.js @@ -2,7 +2,7 @@ const { v4 } = require("uuid"); const { chatPrompt } = require("../../chats"); class AnthropicLLM { - constructor(embedder = null) { + constructor(embedder = null, modelPreference = null) { if (!process.env.ANTHROPIC_API_KEY) throw new Error("No Anthropic API key was set."); @@ -12,7 +12,8 @@ class AnthropicLLM { apiKey: process.env.ANTHROPIC_API_KEY, }); this.anthropic = anthropic; - this.model = process.env.ANTHROPIC_MODEL_PREF || "claude-2"; + this.model = + modelPreference || process.env.ANTHROPIC_MODEL_PREF || "claude-2"; this.limits = { history: this.promptWindowLimit() * 0.15, system: this.promptWindowLimit() * 0.15, @@ -25,6 +26,7 @@ class AnthropicLLM { ); this.embedder = embedder; this.answerKey = v4().split("-")[0]; + this.defaultTemp = 0.7; } streamingEnabled() { diff --git a/server/utils/AiProviders/azureOpenAi/index.js b/server/utils/AiProviders/azureOpenAi/index.js index 185dac021..639ac102e 100644 --- a/server/utils/AiProviders/azureOpenAi/index.js +++ b/server/utils/AiProviders/azureOpenAi/index.js @@ -2,7 +2,7 @@ const { AzureOpenAiEmbedder } = require("../../EmbeddingEngines/azureOpenAi"); const { chatPrompt } = require("../../chats"); class AzureOpenAiLLM { - constructor(embedder = null) { + constructor(embedder = null, _modelPreference = null) { const { OpenAIClient, AzureKeyCredential } = require("@azure/openai"); if (!process.env.AZURE_OPENAI_ENDPOINT) throw new Error("No Azure API endpoint was set."); @@ -25,6 +25,7 @@ class AzureOpenAiLLM { "No embedding provider defined for AzureOpenAiLLM - falling back to AzureOpenAiEmbedder for embedding!" ); this.embedder = !embedder ? new AzureOpenAiEmbedder() : embedder; + this.defaultTemp = 0.7; } #appendContext(contextTexts = []) { @@ -93,7 +94,7 @@ class AzureOpenAiLLM { ); const textResponse = await this.openai .getChatCompletions(this.model, messages, { - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, }) .then((res) => { @@ -130,7 +131,7 @@ class AzureOpenAiLLM { this.model, messages, { - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, } ); diff --git a/server/utils/AiProviders/gemini/index.js b/server/utils/AiProviders/gemini/index.js index 03388e3e2..63549fb8d 100644 --- a/server/utils/AiProviders/gemini/index.js +++ b/server/utils/AiProviders/gemini/index.js @@ -1,14 +1,15 @@ const { chatPrompt } = require("../../chats"); class GeminiLLM { - constructor(embedder = null) { + constructor(embedder = null, modelPreference = null) { if (!process.env.GEMINI_API_KEY) throw new Error("No Gemini API key was set."); // Docs: https://ai.google.dev/tutorials/node_quickstart const { GoogleGenerativeAI } = require("@google/generative-ai"); const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY); - this.model = process.env.GEMINI_LLM_MODEL_PREF || "gemini-pro"; + this.model = + modelPreference || process.env.GEMINI_LLM_MODEL_PREF || "gemini-pro"; this.gemini = genAI.getGenerativeModel({ model: this.model }); this.limits = { history: this.promptWindowLimit() * 0.15, @@ -21,6 +22,7 @@ class GeminiLLM { "INVALID GEMINI LLM SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use Gemini as your LLM." ); this.embedder = embedder; + this.defaultTemp = 0.7; // not used for Gemini } #appendContext(contextTexts = []) { diff --git a/server/utils/AiProviders/lmStudio/index.js b/server/utils/AiProviders/lmStudio/index.js index 28c107df0..08950a7b9 100644 --- a/server/utils/AiProviders/lmStudio/index.js +++ b/server/utils/AiProviders/lmStudio/index.js @@ -2,7 +2,7 @@ const { chatPrompt } = require("../../chats"); // hybrid of openAi LLM chat completion for LMStudio class LMStudioLLM { - constructor(embedder = null) { + constructor(embedder = null, _modelPreference = null) { if (!process.env.LMSTUDIO_BASE_PATH) throw new Error("No LMStudio API Base Path was set."); @@ -12,7 +12,7 @@ class LMStudioLLM { }); this.lmstudio = new OpenAIApi(config); // When using LMStudios inference server - the model param is not required so - // we can stub it here. + // we can stub it here. LMStudio can only run one model at a time. this.model = "model-placeholder"; this.limits = { history: this.promptWindowLimit() * 0.15, @@ -25,6 +25,7 @@ class LMStudioLLM { "INVALID LM STUDIO SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use LMStudio as your LLM." ); this.embedder = embedder; + this.defaultTemp = 0.7; } #appendContext(contextTexts = []) { @@ -85,7 +86,7 @@ class LMStudioLLM { const textResponse = await this.lmstudio .createChatCompletion({ model: this.model, - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, messages: await this.compressMessages( { @@ -122,7 +123,7 @@ class LMStudioLLM { const streamRequest = await this.lmstudio.createChatCompletion( { model: this.model, - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, stream: true, messages: await this.compressMessages( diff --git a/server/utils/AiProviders/localAi/index.js b/server/utils/AiProviders/localAi/index.js index 84954c994..6d265cf82 100644 --- a/server/utils/AiProviders/localAi/index.js +++ b/server/utils/AiProviders/localAi/index.js @@ -1,7 +1,7 @@ const { chatPrompt } = require("../../chats"); class LocalAiLLM { - constructor(embedder = null) { + constructor(embedder = null, modelPreference = null) { if (!process.env.LOCAL_AI_BASE_PATH) throw new Error("No LocalAI Base Path was set."); @@ -15,7 +15,7 @@ class LocalAiLLM { : {}), }); this.openai = new OpenAIApi(config); - this.model = process.env.LOCAL_AI_MODEL_PREF; + this.model = modelPreference || process.env.LOCAL_AI_MODEL_PREF; this.limits = { history: this.promptWindowLimit() * 0.15, system: this.promptWindowLimit() * 0.15, @@ -27,6 +27,7 @@ class LocalAiLLM { "INVALID LOCAL AI SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use LocalAI as your LLM." ); this.embedder = embedder; + this.defaultTemp = 0.7; } #appendContext(contextTexts = []) { @@ -85,7 +86,7 @@ class LocalAiLLM { const textResponse = await this.openai .createChatCompletion({ model: this.model, - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, messages: await this.compressMessages( { @@ -123,7 +124,7 @@ class LocalAiLLM { { model: this.model, stream: true, - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, messages: await this.compressMessages( { diff --git a/server/utils/AiProviders/mistral/index.js b/server/utils/AiProviders/mistral/index.js new file mode 100644 index 000000000..a25185c76 --- /dev/null +++ b/server/utils/AiProviders/mistral/index.js @@ -0,0 +1,184 @@ +const { chatPrompt } = require("../../chats"); + +class MistralLLM { + constructor(embedder = null, modelPreference = null) { + const { Configuration, OpenAIApi } = require("openai"); + if (!process.env.MISTRAL_API_KEY) + throw new Error("No Mistral API key was set."); + + const config = new Configuration({ + basePath: "https://api.mistral.ai/v1", + apiKey: process.env.MISTRAL_API_KEY, + }); + this.openai = new OpenAIApi(config); + this.model = + modelPreference || process.env.MISTRAL_MODEL_PREF || "mistral-tiny"; + this.limits = { + history: this.promptWindowLimit() * 0.15, + system: this.promptWindowLimit() * 0.15, + user: this.promptWindowLimit() * 0.7, + }; + + if (!embedder) + console.warn( + "No embedding provider defined for MistralLLM - falling back to OpenAiEmbedder for embedding!" + ); + this.embedder = embedder; + this.defaultTemp = 0.0; + } + + #appendContext(contextTexts = []) { + if (!contextTexts || !contextTexts.length) return ""; + return ( + "\nContext:\n" + + contextTexts + .map((text, i) => { + return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`; + }) + .join("") + ); + } + + streamingEnabled() { + return "streamChat" in this && "streamGetChatCompletion" in this; + } + + promptWindowLimit() { + return 32000; + } + + async isValidChatCompletionModel(modelName = "") { + return true; + } + + constructPrompt({ + systemPrompt = "", + contextTexts = [], + chatHistory = [], + userPrompt = "", + }) { + const prompt = { + role: "system", + content: `${systemPrompt}${this.#appendContext(contextTexts)}`, + }; + return [prompt, ...chatHistory, { role: "user", content: userPrompt }]; + } + + async isSafe(_ = "") { + return { safe: true, reasons: [] }; + } + + async sendChat(chatHistory = [], prompt, workspace = {}, rawHistory = []) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `Mistral chat: ${this.model} is not valid for chat completion!` + ); + + const textResponse = await this.openai + .createChatCompletion({ + model: this.model, + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), + messages: await this.compressMessages( + { + systemPrompt: chatPrompt(workspace), + userPrompt: prompt, + chatHistory, + }, + rawHistory + ), + }) + .then((json) => { + const res = json.data; + if (!res.hasOwnProperty("choices")) + throw new Error("Mistral chat: No results!"); + if (res.choices.length === 0) + throw new Error("Mistral chat: No results length!"); + return res.choices[0].message.content; + }) + .catch((error) => { + throw new Error( + `Mistral::createChatCompletion failed with: ${error.message}` + ); + }); + + return textResponse; + } + + async streamChat(chatHistory = [], prompt, workspace = {}, rawHistory = []) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `Mistral chat: ${this.model} is not valid for chat completion!` + ); + + const streamRequest = await this.openai.createChatCompletion( + { + model: this.model, + stream: true, + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), + messages: await this.compressMessages( + { + systemPrompt: chatPrompt(workspace), + userPrompt: prompt, + chatHistory, + }, + rawHistory + ), + }, + { responseType: "stream" } + ); + + return streamRequest; + } + + async getChatCompletion(messages = null, { temperature = 0.7 }) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `Mistral chat: ${this.model} is not valid for chat completion!` + ); + + const { data } = await this.openai.createChatCompletion({ + model: this.model, + messages, + temperature, + }); + + if (!data.hasOwnProperty("choices")) return null; + return data.choices[0].message.content; + } + + async streamGetChatCompletion(messages = null, { temperature = 0.7 }) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `Mistral chat: ${this.model} is not valid for chat completion!` + ); + + const streamRequest = await this.openai.createChatCompletion( + { + model: this.model, + stream: true, + messages, + temperature, + }, + { responseType: "stream" } + ); + return streamRequest; + } + + // Simple wrapper for dynamic embedder & normalize interface for all LLM implementations + async embedTextInput(textInput) { + return await this.embedder.embedTextInput(textInput); + } + async embedChunks(textChunks = []) { + return await this.embedder.embedChunks(textChunks); + } + + async compressMessages(promptArgs = {}, rawHistory = []) { + const { messageArrayCompressor } = require("../../helpers/chat"); + const messageArray = this.constructPrompt(promptArgs); + return await messageArrayCompressor(this, messageArray, rawHistory); + } +} + +module.exports = { + MistralLLM, +}; diff --git a/server/utils/AiProviders/native/index.js b/server/utils/AiProviders/native/index.js index faac4fa03..de1a97f3d 100644 --- a/server/utils/AiProviders/native/index.js +++ b/server/utils/AiProviders/native/index.js @@ -10,11 +10,11 @@ const ChatLlamaCpp = (...args) => ); class NativeLLM { - constructor(embedder = null) { + constructor(embedder = null, modelPreference = null) { if (!process.env.NATIVE_LLM_MODEL_PREF) throw new Error("No local Llama model was set."); - this.model = process.env.NATIVE_LLM_MODEL_PREF || null; + this.model = modelPreference || process.env.NATIVE_LLM_MODEL_PREF || null; this.limits = { history: this.promptWindowLimit() * 0.15, system: this.promptWindowLimit() * 0.15, @@ -29,6 +29,7 @@ class NativeLLM { // Make directory when it does not exist in existing installations if (!fs.existsSync(this.cacheDir)) fs.mkdirSync(this.cacheDir); + this.defaultTemp = 0.7; } async #initializeLlamaModel(temperature = 0.7) { @@ -93,8 +94,6 @@ class NativeLLM { } // Ensure the user set a value for the token limit - // and if undefined - assume 4096 window. - // DEV: Currently this ENV is not configurable. promptWindowLimit() { const limit = process.env.NATIVE_LLM_MODEL_TOKEN_LIMIT || 4096; if (!limit || isNaN(Number(limit))) @@ -132,7 +131,7 @@ class NativeLLM { ); const model = await this.#llamaClient({ - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), }); const response = await model.call(messages); return response.content; @@ -145,7 +144,7 @@ class NativeLLM { async streamChat(chatHistory = [], prompt, workspace = {}, rawHistory = []) { const model = await this.#llamaClient({ - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), }); const messages = await this.compressMessages( { diff --git a/server/utils/AiProviders/ollama/index.js b/server/utils/AiProviders/ollama/index.js index 55205c23d..af7fe8210 100644 --- a/server/utils/AiProviders/ollama/index.js +++ b/server/utils/AiProviders/ollama/index.js @@ -3,12 +3,12 @@ const { StringOutputParser } = require("langchain/schema/output_parser"); // Docs: https://github.com/jmorganca/ollama/blob/main/docs/api.md class OllamaAILLM { - constructor(embedder = null) { + constructor(embedder = null, modelPreference = null) { if (!process.env.OLLAMA_BASE_PATH) throw new Error("No Ollama Base Path was set."); this.basePath = process.env.OLLAMA_BASE_PATH; - this.model = process.env.OLLAMA_MODEL_PREF; + this.model = modelPreference || process.env.OLLAMA_MODEL_PREF; this.limits = { history: this.promptWindowLimit() * 0.15, system: this.promptWindowLimit() * 0.15, @@ -20,6 +20,7 @@ class OllamaAILLM { "INVALID OLLAMA SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use Ollama as your LLM." ); this.embedder = embedder; + this.defaultTemp = 0.7; } #ollamaClient({ temperature = 0.07 }) { @@ -113,7 +114,7 @@ class OllamaAILLM { ); const model = this.#ollamaClient({ - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), }); const textResponse = await model .pipe(new StringOutputParser()) @@ -136,7 +137,7 @@ class OllamaAILLM { ); const model = this.#ollamaClient({ - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), }); const stream = await model .pipe(new StringOutputParser()) diff --git a/server/utils/AiProviders/openAi/index.js b/server/utils/AiProviders/openAi/index.js index ccc7ba0e9..582bc054d 100644 --- a/server/utils/AiProviders/openAi/index.js +++ b/server/utils/AiProviders/openAi/index.js @@ -2,7 +2,7 @@ const { OpenAiEmbedder } = require("../../EmbeddingEngines/openAi"); const { chatPrompt } = require("../../chats"); class OpenAiLLM { - constructor(embedder = null) { + constructor(embedder = null, modelPreference = null) { const { Configuration, OpenAIApi } = require("openai"); if (!process.env.OPEN_AI_KEY) throw new Error("No OpenAI API key was set."); @@ -10,7 +10,8 @@ class OpenAiLLM { apiKey: process.env.OPEN_AI_KEY, }); this.openai = new OpenAIApi(config); - this.model = process.env.OPEN_MODEL_PREF || "gpt-3.5-turbo"; + this.model = + modelPreference || process.env.OPEN_MODEL_PREF || "gpt-3.5-turbo"; this.limits = { history: this.promptWindowLimit() * 0.15, system: this.promptWindowLimit() * 0.15, @@ -22,6 +23,7 @@ class OpenAiLLM { "No embedding provider defined for OpenAiLLM - falling back to OpenAiEmbedder for embedding!" ); this.embedder = !embedder ? new OpenAiEmbedder() : embedder; + this.defaultTemp = 0.7; } #appendContext(contextTexts = []) { @@ -126,7 +128,7 @@ class OpenAiLLM { const textResponse = await this.openai .createChatCompletion({ model: this.model, - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, messages: await this.compressMessages( { @@ -164,7 +166,7 @@ class OpenAiLLM { { model: this.model, stream: true, - temperature: Number(workspace?.openAiTemp ?? 0.7), + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), n: 1, messages: await this.compressMessages( { diff --git a/server/utils/AiProviders/togetherAi/index.js b/server/utils/AiProviders/togetherAi/index.js new file mode 100644 index 000000000..341661f8d --- /dev/null +++ b/server/utils/AiProviders/togetherAi/index.js @@ -0,0 +1,199 @@ +const { chatPrompt } = require("../../chats"); + +function togetherAiModels() { + const { MODELS } = require("./models.js"); + return MODELS || {}; +} + +class TogetherAiLLM { + constructor(embedder = null, modelPreference = null) { + const { Configuration, OpenAIApi } = require("openai"); + if (!process.env.TOGETHER_AI_API_KEY) + throw new Error("No TogetherAI API key was set."); + + const config = new Configuration({ + basePath: "https://api.together.xyz/v1", + apiKey: process.env.TOGETHER_AI_API_KEY, + }); + this.openai = new OpenAIApi(config); + this.model = modelPreference || process.env.TOGETHER_AI_MODEL_PREF; + this.limits = { + history: this.promptWindowLimit() * 0.15, + system: this.promptWindowLimit() * 0.15, + user: this.promptWindowLimit() * 0.7, + }; + + if (!embedder) + throw new Error( + "INVALID TOGETHER AI SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use Together AI as your LLM." + ); + this.embedder = embedder; + this.defaultTemp = 0.7; + } + + #appendContext(contextTexts = []) { + if (!contextTexts || !contextTexts.length) return ""; + return ( + "\nContext:\n" + + contextTexts + .map((text, i) => { + return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`; + }) + .join("") + ); + } + + allModelInformation() { + return togetherAiModels(); + } + + streamingEnabled() { + return "streamChat" in this && "streamGetChatCompletion" in this; + } + + // Ensure the user set a value for the token limit + // and if undefined - assume 4096 window. + promptWindowLimit() { + const availableModels = this.allModelInformation(); + return availableModels[this.model]?.maxLength || 4096; + } + + async isValidChatCompletionModel(model = "") { + const availableModels = this.allModelInformation(); + return availableModels.hasOwnProperty(model); + } + + constructPrompt({ + systemPrompt = "", + contextTexts = [], + chatHistory = [], + userPrompt = "", + }) { + const prompt = { + role: "system", + content: `${systemPrompt}${this.#appendContext(contextTexts)}`, + }; + return [prompt, ...chatHistory, { role: "user", content: userPrompt }]; + } + + async isSafe(_input = "") { + // Not implemented so must be stubbed + return { safe: true, reasons: [] }; + } + + async sendChat(chatHistory = [], prompt, workspace = {}, rawHistory = []) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `Together AI chat: ${this.model} is not valid for chat completion!` + ); + + const textResponse = await this.openai + .createChatCompletion({ + model: this.model, + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), + n: 1, + messages: await this.compressMessages( + { + systemPrompt: chatPrompt(workspace), + userPrompt: prompt, + chatHistory, + }, + rawHistory + ), + }) + .then((json) => { + const res = json.data; + if (!res.hasOwnProperty("choices")) + throw new Error("Together AI chat: No results!"); + if (res.choices.length === 0) + throw new Error("Together AI chat: No results length!"); + return res.choices[0].message.content; + }) + .catch((error) => { + throw new Error( + `TogetherAI::createChatCompletion failed with: ${error.message}` + ); + }); + + return textResponse; + } + + async streamChat(chatHistory = [], prompt, workspace = {}, rawHistory = []) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `TogetherAI chat: ${this.model} is not valid for chat completion!` + ); + + const streamRequest = await this.openai.createChatCompletion( + { + model: this.model, + stream: true, + temperature: Number(workspace?.openAiTemp ?? this.defaultTemp), + n: 1, + messages: await this.compressMessages( + { + systemPrompt: chatPrompt(workspace), + userPrompt: prompt, + chatHistory, + }, + rawHistory + ), + }, + { responseType: "stream" } + ); + return { type: "togetherAiStream", stream: streamRequest }; + } + + async getChatCompletion(messages = null, { temperature = 0.7 }) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `TogetherAI chat: ${this.model} is not valid for chat completion!` + ); + + const { data } = await this.openai.createChatCompletion({ + model: this.model, + messages, + temperature, + }); + + if (!data.hasOwnProperty("choices")) return null; + return data.choices[0].message.content; + } + + async streamGetChatCompletion(messages = null, { temperature = 0.7 }) { + if (!(await this.isValidChatCompletionModel(this.model))) + throw new Error( + `TogetherAI chat: ${this.model} is not valid for chat completion!` + ); + + const streamRequest = await this.openai.createChatCompletion( + { + model: this.model, + stream: true, + messages, + temperature, + }, + { responseType: "stream" } + ); + return { type: "togetherAiStream", stream: streamRequest }; + } + + // Simple wrapper for dynamic embedder & normalize interface for all LLM implementations + async embedTextInput(textInput) { + return await this.embedder.embedTextInput(textInput); + } + async embedChunks(textChunks = []) { + return await this.embedder.embedChunks(textChunks); + } + + async compressMessages(promptArgs = {}, rawHistory = []) { + const { messageArrayCompressor } = require("../../helpers/chat"); + const messageArray = this.constructPrompt(promptArgs); + return await messageArrayCompressor(this, messageArray, rawHistory); + } +} + +module.exports = { + TogetherAiLLM, + togetherAiModels, +}; diff --git a/server/utils/AiProviders/togetherAi/models.js b/server/utils/AiProviders/togetherAi/models.js new file mode 100644 index 000000000..ad940bc39 --- /dev/null +++ b/server/utils/AiProviders/togetherAi/models.js @@ -0,0 +1,226 @@ +const MODELS = { + "togethercomputer/alpaca-7b": { + id: "togethercomputer/alpaca-7b", + organization: "Stanford", + name: "Alpaca (7B)", + maxLength: 2048, + }, + "Austism/chronos-hermes-13b": { + id: "Austism/chronos-hermes-13b", + organization: "Austism", + name: "Chronos Hermes (13B)", + maxLength: 2048, + }, + "togethercomputer/CodeLlama-13b-Instruct": { + id: "togethercomputer/CodeLlama-13b-Instruct", + organization: "Meta", + name: "Code Llama Instruct (13B)", + maxLength: 8192, + }, + "togethercomputer/CodeLlama-34b-Instruct": { + id: "togethercomputer/CodeLlama-34b-Instruct", + organization: "Meta", + name: "Code Llama Instruct (34B)", + maxLength: 8192, + }, + "togethercomputer/CodeLlama-7b-Instruct": { + id: "togethercomputer/CodeLlama-7b-Instruct", + organization: "Meta", + name: "Code Llama Instruct (7B)", + maxLength: 8192, + }, + "DiscoResearch/DiscoLM-mixtral-8x7b-v2": { + id: "DiscoResearch/DiscoLM-mixtral-8x7b-v2", + organization: "DiscoResearch", + name: "DiscoLM Mixtral 8x7b", + maxLength: 32768, + }, + "togethercomputer/falcon-40b-instruct": { + id: "togethercomputer/falcon-40b-instruct", + organization: "TII UAE", + name: "Falcon Instruct (40B)", + maxLength: 2048, + }, + "togethercomputer/falcon-7b-instruct": { + id: "togethercomputer/falcon-7b-instruct", + organization: "TII UAE", + name: "Falcon Instruct (7B)", + maxLength: 2048, + }, + "togethercomputer/GPT-NeoXT-Chat-Base-20B": { + id: "togethercomputer/GPT-NeoXT-Chat-Base-20B", + organization: "Together", + name: "GPT-NeoXT-Chat-Base (20B)", + maxLength: 2048, + }, + "togethercomputer/llama-2-13b-chat": { + id: "togethercomputer/llama-2-13b-chat", + organization: "Meta", + name: "LLaMA-2 Chat (13B)", + maxLength: 4096, + }, + "togethercomputer/llama-2-70b-chat": { + id: "togethercomputer/llama-2-70b-chat", + organization: "Meta", + name: "LLaMA-2 Chat (70B)", + maxLength: 4096, + }, + "togethercomputer/llama-2-7b-chat": { + id: "togethercomputer/llama-2-7b-chat", + organization: "Meta", + name: "LLaMA-2 Chat (7B)", + maxLength: 4096, + }, + "togethercomputer/Llama-2-7B-32K-Instruct": { + id: "togethercomputer/Llama-2-7B-32K-Instruct", + organization: "Together", + name: "LLaMA-2-7B-32K-Instruct (7B)", + maxLength: 32768, + }, + "mistralai/Mistral-7B-Instruct-v0.1": { + id: "mistralai/Mistral-7B-Instruct-v0.1", + organization: "MistralAI", + name: "Mistral (7B) Instruct v0.1", + maxLength: 4096, + }, + "mistralai/Mistral-7B-Instruct-v0.2": { + id: "mistralai/Mistral-7B-Instruct-v0.2", + organization: "MistralAI", + name: "Mistral (7B) Instruct v0.2", + maxLength: 32768, + }, + "mistralai/Mixtral-8x7B-Instruct-v0.1": { + id: "mistralai/Mixtral-8x7B-Instruct-v0.1", + organization: "MistralAI", + name: "Mixtral-8x7B Instruct", + maxLength: 32768, + }, + "Gryphe/MythoMax-L2-13b": { + id: "Gryphe/MythoMax-L2-13b", + organization: "Gryphe", + name: "MythoMax-L2 (13B)", + maxLength: 4096, + }, + "NousResearch/Nous-Hermes-llama-2-7b": { + id: "NousResearch/Nous-Hermes-llama-2-7b", + organization: "NousResearch", + name: "Nous Hermes LLaMA-2 (7B)", + maxLength: 4096, + }, + "NousResearch/Nous-Hermes-Llama2-13b": { + id: "NousResearch/Nous-Hermes-Llama2-13b", + organization: "NousResearch", + name: "Nous Hermes Llama-2 (13B)", + maxLength: 4096, + }, + "NousResearch/Nous-Hermes-Llama2-70b": { + id: "NousResearch/Nous-Hermes-Llama2-70b", + organization: "NousResearch", + name: "Nous Hermes Llama-2 (70B)", + maxLength: 4096, + }, + "NousResearch/Nous-Hermes-2-Yi-34B": { + id: "NousResearch/Nous-Hermes-2-Yi-34B", + organization: "NousResearch", + name: "Nous Hermes-2 Yi (34B)", + maxLength: 4096, + }, + "NousResearch/Nous-Capybara-7B-V1p9": { + id: "NousResearch/Nous-Capybara-7B-V1p9", + organization: "NousResearch", + name: "Nous Capybara v1.9 (7B)", + maxLength: 8192, + }, + "openchat/openchat-3.5-1210": { + id: "openchat/openchat-3.5-1210", + organization: "OpenChat", + name: "OpenChat 3.5 1210 (7B)", + maxLength: 8192, + }, + "teknium/OpenHermes-2-Mistral-7B": { + id: "teknium/OpenHermes-2-Mistral-7B", + organization: "teknium", + name: "OpenHermes-2-Mistral (7B)", + maxLength: 4096, + }, + "teknium/OpenHermes-2p5-Mistral-7B": { + id: "teknium/OpenHermes-2p5-Mistral-7B", + organization: "teknium", + name: "OpenHermes-2.5-Mistral (7B)", + maxLength: 4096, + }, + "Open-Orca/Mistral-7B-OpenOrca": { + id: "Open-Orca/Mistral-7B-OpenOrca", + organization: "OpenOrca", + name: "OpenOrca Mistral (7B) 8K", + maxLength: 8192, + }, + "garage-bAInd/Platypus2-70B-instruct": { + id: "garage-bAInd/Platypus2-70B-instruct", + organization: "garage-bAInd", + name: "Platypus2 Instruct (70B)", + maxLength: 4096, + }, + "togethercomputer/Pythia-Chat-Base-7B-v0.16": { + id: "togethercomputer/Pythia-Chat-Base-7B-v0.16", + organization: "Together", + name: "Pythia-Chat-Base (7B)", + maxLength: 2048, + }, + "togethercomputer/Qwen-7B-Chat": { + id: "togethercomputer/Qwen-7B-Chat", + organization: "Qwen", + name: "Qwen-Chat (7B)", + maxLength: 8192, + }, + "togethercomputer/RedPajama-INCITE-Chat-3B-v1": { + id: "togethercomputer/RedPajama-INCITE-Chat-3B-v1", + organization: "Together", + name: "RedPajama-INCITE Chat (3B)", + maxLength: 2048, + }, + "togethercomputer/RedPajama-INCITE-7B-Chat": { + id: "togethercomputer/RedPajama-INCITE-7B-Chat", + organization: "Together", + name: "RedPajama-INCITE Chat (7B)", + maxLength: 2048, + }, + "upstage/SOLAR-0-70b-16bit": { + id: "upstage/SOLAR-0-70b-16bit", + organization: "Upstage", + name: "SOLAR v0 (70B)", + maxLength: 4096, + }, + "togethercomputer/StripedHyena-Nous-7B": { + id: "togethercomputer/StripedHyena-Nous-7B", + organization: "Together", + name: "StripedHyena Nous (7B)", + maxLength: 32768, + }, + "lmsys/vicuna-7b-v1.5": { + id: "lmsys/vicuna-7b-v1.5", + organization: "LM Sys", + name: "Vicuna v1.5 (7B)", + maxLength: 4096, + }, + "lmsys/vicuna-13b-v1.5": { + id: "lmsys/vicuna-13b-v1.5", + organization: "LM Sys", + name: "Vicuna v1.5 (13B)", + maxLength: 4096, + }, + "lmsys/vicuna-13b-v1.5-16k": { + id: "lmsys/vicuna-13b-v1.5-16k", + organization: "LM Sys", + name: "Vicuna v1.5 16K (13B)", + maxLength: 16384, + }, + "zero-one-ai/Yi-34B-Chat": { + id: "zero-one-ai/Yi-34B-Chat", + organization: "01.AI", + name: "01-ai Yi Chat (34B)", + maxLength: 4096, + }, +}; + +module.exports.MODELS = MODELS; diff --git a/server/utils/AiProviders/togetherAi/scripts/.gitignore b/server/utils/AiProviders/togetherAi/scripts/.gitignore new file mode 100644 index 000000000..94a2dd146 --- /dev/null +++ b/server/utils/AiProviders/togetherAi/scripts/.gitignore @@ -0,0 +1 @@ +*.json \ No newline at end of file diff --git a/server/utils/AiProviders/togetherAi/scripts/chat_models.txt b/server/utils/AiProviders/togetherAi/scripts/chat_models.txt new file mode 100644 index 000000000..81c23bf4a --- /dev/null +++ b/server/utils/AiProviders/togetherAi/scripts/chat_models.txt @@ -0,0 +1,39 @@ +| Organization | Model Name | Model String for API | Max Seq Length | +| ------------- | ---------------------------- | -------------------------------------------- | -------------- | +| Stanford | Alpaca (7B) | togethercomputer/alpaca-7b | 2048 | +| Austism | Chronos Hermes (13B) | Austism/chronos-hermes-13b | 2048 | +| Meta | Code Llama Instruct (13B) | togethercomputer/CodeLlama-13b-Instruct | 8192 | +| Meta | Code Llama Instruct (34B) | togethercomputer/CodeLlama-34b-Instruct | 8192 | +| Meta | Code Llama Instruct (7B) | togethercomputer/CodeLlama-7b-Instruct | 8192 | +| DiscoResearch | DiscoLM Mixtral 8x7b | DiscoResearch/DiscoLM-mixtral-8x7b-v2 | 32768 | +| TII UAE | Falcon Instruct (40B) | togethercomputer/falcon-40b-instruct | 2048 | +| TII UAE | Falcon Instruct (7B) | togethercomputer/falcon-7b-instruct | 2048 | +| Together | GPT-NeoXT-Chat-Base (20B) | togethercomputer/GPT-NeoXT-Chat-Base-20B | 2048 | +| Meta | LLaMA-2 Chat (13B) | togethercomputer/llama-2-13b-chat | 4096 | +| Meta | LLaMA-2 Chat (70B) | togethercomputer/llama-2-70b-chat | 4096 | +| Meta | LLaMA-2 Chat (7B) | togethercomputer/llama-2-7b-chat | 4096 | +| Together | LLaMA-2-7B-32K-Instruct (7B) | togethercomputer/Llama-2-7B-32K-Instruct | 32768 | +| MistralAI | Mistral (7B) Instruct v0.1 | mistralai/Mistral-7B-Instruct-v0.1 | 4096 | +| MistralAI | Mistral (7B) Instruct v0.2 | mistralai/Mistral-7B-Instruct-v0.2 | 32768 | +| MistralAI | Mixtral-8x7B Instruct | mistralai/Mixtral-8x7B-Instruct-v0.1 | 32768 | +| Gryphe | MythoMax-L2 (13B) | Gryphe/MythoMax-L2-13b | 4096 | +| NousResearch | Nous Hermes LLaMA-2 (7B) | NousResearch/Nous-Hermes-llama-2-7b | 4096 | +| NousResearch | Nous Hermes Llama-2 (13B) | NousResearch/Nous-Hermes-Llama2-13b | 4096 | +| NousResearch | Nous Hermes Llama-2 (70B) | NousResearch/Nous-Hermes-Llama2-70b | 4096 | +| NousResearch | Nous Hermes-2 Yi (34B) | NousResearch/Nous-Hermes-2-Yi-34B | 4096 | +| NousResearch | Nous Capybara v1.9 (7B) | NousResearch/Nous-Capybara-7B-V1p9 | 8192 | +| OpenChat | OpenChat 3.5 1210 (7B) | openchat/openchat-3.5-1210 | 8192 | +| teknium | OpenHermes-2-Mistral (7B) | teknium/OpenHermes-2-Mistral-7B | 4096 | +| teknium | OpenHermes-2.5-Mistral (7B) | teknium/OpenHermes-2p5-Mistral-7B | 4096 | +| OpenOrca | OpenOrca Mistral (7B) 8K | Open-Orca/Mistral-7B-OpenOrca | 8192 | +| garage-bAInd | Platypus2 Instruct (70B) | garage-bAInd/Platypus2-70B-instruct | 4096 | +| Together | Pythia-Chat-Base (7B) | togethercomputer/Pythia-Chat-Base-7B-v0.16 | 2048 | +| Qwen | Qwen-Chat (7B) | togethercomputer/Qwen-7B-Chat | 8192 | +| Together | RedPajama-INCITE Chat (3B) | togethercomputer/RedPajama-INCITE-Chat-3B-v1 | 2048 | +| Together | RedPajama-INCITE Chat (7B) | togethercomputer/RedPajama-INCITE-7B-Chat | 2048 | +| Upstage | SOLAR v0 (70B) | upstage/SOLAR-0-70b-16bit | 4096 | +| Together | StripedHyena Nous (7B) | togethercomputer/StripedHyena-Nous-7B | 32768 | +| LM Sys | Vicuna v1.5 (7B) | lmsys/vicuna-7b-v1.5 | 4096 | +| LM Sys | Vicuna v1.5 (13B) | lmsys/vicuna-13b-v1.5 | 4096 | +| LM Sys | Vicuna v1.5 16K (13B) | lmsys/vicuna-13b-v1.5-16k | 16384 | +| 01.AI | 01-ai Yi Chat (34B) | zero-one-ai/Yi-34B-Chat | 4096 | \ No newline at end of file diff --git a/server/utils/AiProviders/togetherAi/scripts/parse.mjs b/server/utils/AiProviders/togetherAi/scripts/parse.mjs new file mode 100644 index 000000000..b96d40ab1 --- /dev/null +++ b/server/utils/AiProviders/togetherAi/scripts/parse.mjs @@ -0,0 +1,41 @@ +// Together AI does not provide a simple REST API to get models, +// so we have a table which we copy from their documentation +// https://docs.together.ai/edit/inference-models that we can +// then parse and get all models from in a format that makes sense +// Why this does not exist is so bizarre, but whatever. + +// To run, cd into this directory and run `node parse.mjs` +// copy outputs into the export in ../models.js + +// Update the date below if you run this again because TogetherAI added new models. +// Last Collected: Jan 10, 2023 + +import fs from "fs"; + +function parseChatModels() { + const fixed = {}; + const tableString = fs.readFileSync("chat_models.txt", { encoding: "utf-8" }); + const rows = tableString.split("\n").slice(2); + + rows.forEach((row) => { + const [provider, name, id, maxLength] = row.split("|").slice(1, -1); + const data = { + provider: provider.trim(), + name: name.trim(), + id: id.trim(), + maxLength: Number(maxLength.trim()), + }; + + fixed[data.id] = { + id: data.id, + organization: data.provider, + name: data.name, + maxLength: data.maxLength, + }; + }); + + fs.writeFileSync("chat_models.json", JSON.stringify(fixed, null, 2), "utf-8"); + return fixed; +} + +parseChatModels(); diff --git a/server/utils/EmbeddingEngines/native/index.js b/server/utils/EmbeddingEngines/native/index.js index 69e13a9e3..fc933e1b8 100644 --- a/server/utils/EmbeddingEngines/native/index.js +++ b/server/utils/EmbeddingEngines/native/index.js @@ -1,6 +1,7 @@ const path = require("path"); const fs = require("fs"); const { toChunks } = require("../../helpers"); +const { v4 } = require("uuid"); class NativeEmbedder { constructor() { @@ -14,13 +15,30 @@ class NativeEmbedder { this.modelPath = path.resolve(this.cacheDir, "Xenova", "all-MiniLM-L6-v2"); // Limit of how many strings we can process in a single pass to stay with resource or network limits - this.maxConcurrentChunks = 50; + this.maxConcurrentChunks = 25; this.embeddingMaxChunkLength = 1_000; // Make directory when it does not exist in existing installations if (!fs.existsSync(this.cacheDir)) fs.mkdirSync(this.cacheDir); } + #tempfilePath() { + const filename = `${v4()}.tmp`; + const tmpPath = process.env.STORAGE_DIR + ? path.resolve(process.env.STORAGE_DIR, "tmp") + : path.resolve(__dirname, `../../../storage/tmp`); + if (!fs.existsSync(tmpPath)) fs.mkdirSync(tmpPath, { recursive: true }); + return path.resolve(tmpPath, filename); + } + + async #writeToTempfile(filePath, data) { + try { + await fs.promises.appendFile(filePath, data, { encoding: "utf8" }); + } catch (e) { + console.error(`Error writing to tempfile: ${e}`); + } + } + async embedderClient() { if (!fs.existsSync(this.modelPath)) { console.log( @@ -61,18 +79,51 @@ class NativeEmbedder { return result?.[0] || []; } + // If you are thinking you want to edit this function - you probably don't. + // This process was benchmarked heavily on a t3.small (2GB RAM 1vCPU) + // and without careful memory management for the V8 garbage collector + // this function will likely result in an OOM on any resource-constrained deployment. + // To help manage very large documents we run a concurrent write-log each iteration + // to keep the embedding result out of memory. The `maxConcurrentChunk` is set to 25, + // as 50 seems to overflow no matter what. Given the above, memory use hovers around ~30% + // during a very large document (>100K words) but can spike up to 70% before gc. + // This seems repeatable for all document sizes. + // While this does take a while, it is zero set up and is 100% free and on-instance. async embedChunks(textChunks = []) { - const Embedder = await this.embedderClient(); - const embeddingResults = []; - for (const chunk of toChunks(textChunks, this.maxConcurrentChunks)) { - const output = await Embedder(chunk, { + const tmpFilePath = this.#tempfilePath(); + const chunks = toChunks(textChunks, this.maxConcurrentChunks); + const chunkLen = chunks.length; + + for (let [idx, chunk] of chunks.entries()) { + if (idx === 0) await this.#writeToTempfile(tmpFilePath, "["); + let data; + let pipeline = await this.embedderClient(); + let output = await pipeline(chunk, { pooling: "mean", normalize: true, }); - if (output.length === 0) continue; - embeddingResults.push(output.tolist()); + + if (output.length === 0) { + pipeline = null; + output = null; + data = null; + continue; + } + + data = JSON.stringify(output.tolist()); + await this.#writeToTempfile(tmpFilePath, data); + console.log(`\x1b[34m[Embedded Chunk ${idx + 1} of ${chunkLen}]\x1b[0m`); + if (chunkLen - 1 !== idx) await this.#writeToTempfile(tmpFilePath, ","); + if (chunkLen - 1 === idx) await this.#writeToTempfile(tmpFilePath, "]"); + pipeline = null; + output = null; + data = null; } + const embeddingResults = JSON.parse( + fs.readFileSync(tmpFilePath, { encoding: "utf-8" }) + ); + fs.rmSync(tmpFilePath, { force: true }); return embeddingResults.length > 0 ? embeddingResults.flat() : null; } } diff --git a/server/utils/chats/index.js b/server/utils/chats/index.js index 7e9be6e5b..764c7795a 100644 --- a/server/utils/chats/index.js +++ b/server/utils/chats/index.js @@ -71,7 +71,7 @@ async function chatWithWorkspace( return await VALID_COMMANDS[command](workspace, message, uuid, user); } - const LLMConnector = getLLMProvider(); + const LLMConnector = getLLMProvider(workspace?.chatModel); const VectorDb = getVectorDbClass(); const { safe, reasons = [] } = await LLMConnector.isSafe(message); if (!safe) { @@ -91,6 +91,18 @@ async function chatWithWorkspace( const hasVectorizedSpace = await VectorDb.hasNamespace(workspace.slug); const embeddingsCount = await VectorDb.namespaceCount(workspace.slug); if (!hasVectorizedSpace || embeddingsCount === 0) { + if (chatMode === "query") { + return { + id: uuid, + type: "textResponse", + sources: [], + close: true, + error: null, + textResponse: + "There is no relevant information in this workspace to answer your query.", + }; + } + // If there are no embeddings - chat like a normal LLM chat interface. return await emptyEmbeddingChat({ uuid, @@ -131,6 +143,20 @@ async function chatWithWorkspace( }; } + // If in query mode and no sources are found, do not + // let the LLM try to hallucinate a response or use general knowledge + if (chatMode === "query" && sources.length === 0) { + return { + id: uuid, + type: "textResponse", + sources: [], + close: true, + error: null, + textResponse: + "There is no relevant information in this workspace to answer your query.", + }; + } + // Compress message to ensure prompt passes token limit with room for response // and build system messages based on inputs and history. const messages = await LLMConnector.compressMessages( @@ -145,7 +171,7 @@ async function chatWithWorkspace( // Send the text completion. const textResponse = await LLMConnector.getChatCompletion(messages, { - temperature: workspace?.openAiTemp ?? 0.7, + temperature: workspace?.openAiTemp ?? LLMConnector.defaultTemp, }); if (!textResponse) { diff --git a/server/utils/chats/stream.js b/server/utils/chats/stream.js index a6ade1819..cff565ed6 100644 --- a/server/utils/chats/stream.js +++ b/server/utils/chats/stream.js @@ -8,6 +8,7 @@ const { chatPrompt, } = require("."); +const VALID_CHAT_MODE = ["chat", "query"]; function writeResponseChunk(response, data) { response.write(`data: ${JSON.stringify(data)}\n\n`); return; @@ -29,7 +30,7 @@ async function streamChatWithWorkspace( return; } - const LLMConnector = getLLMProvider(); + const LLMConnector = getLLMProvider(workspace?.chatModel); const VectorDb = getVectorDbClass(); const { safe, reasons = [] } = await LLMConnector.isSafe(message); if (!safe) { @@ -50,6 +51,19 @@ async function streamChatWithWorkspace( const hasVectorizedSpace = await VectorDb.hasNamespace(workspace.slug); const embeddingsCount = await VectorDb.namespaceCount(workspace.slug); if (!hasVectorizedSpace || embeddingsCount === 0) { + if (chatMode === "query") { + writeResponseChunk(response, { + id: uuid, + type: "textResponse", + textResponse: + "There is no relevant information in this workspace to answer your query.", + sources: [], + close: true, + error: null, + }); + return; + } + // If there are no embeddings - chat like a normal LLM chat interface. return await streamEmptyEmbeddingChat({ response, @@ -93,6 +107,21 @@ async function streamChatWithWorkspace( return; } + // If in query mode and no sources are found, do not + // let the LLM try to hallucinate a response or use general knowledge + if (chatMode === "query" && sources.length === 0) { + writeResponseChunk(response, { + id: uuid, + type: "textResponse", + textResponse: + "There is no relevant information in this workspace to answer your query.", + sources: [], + close: true, + error: null, + }); + return; + } + // Compress message to ensure prompt passes token limit with room for response // and build system messages based on inputs and history. const messages = await LLMConnector.compressMessages( @@ -112,7 +141,7 @@ async function streamChatWithWorkspace( `\x1b[31m[STREAMING DISABLED]\x1b[0m Streaming is not available for ${LLMConnector.constructor.name}. Will use regular chat method.` ); completeText = await LLMConnector.getChatCompletion(messages, { - temperature: workspace?.openAiTemp ?? 0.7, + temperature: workspace?.openAiTemp ?? LLMConnector.defaultTemp, }); writeResponseChunk(response, { uuid, @@ -124,7 +153,7 @@ async function streamChatWithWorkspace( }); } else { const stream = await LLMConnector.streamGetChatCompletion(messages, { - temperature: workspace?.openAiTemp ?? 0.7, + temperature: workspace?.openAiTemp ?? LLMConnector.defaultTemp, }); completeText = await handleStreamResponses(response, stream, { uuid, @@ -262,6 +291,96 @@ function handleStreamResponses(response, stream, responseProps) { }); } + if (stream.type === "togetherAiStream") { + return new Promise((resolve) => { + let fullText = ""; + let chunk = ""; + stream.stream.data.on("data", (data) => { + const lines = data + ?.toString() + ?.split("\n") + .filter((line) => line.trim() !== ""); + + for (const line of lines) { + let validJSON = false; + const message = chunk + line.replace(/^data: /, ""); + + if (message !== "[DONE]") { + // JSON chunk is incomplete and has not ended yet + // so we need to stitch it together. You would think JSON + // chunks would only come complete - but they don't! + try { + JSON.parse(message); + validJSON = true; + } catch {} + + if (!validJSON) { + // It can be possible that the chunk decoding is running away + // and the message chunk fails to append due to string length. + // In this case abort the chunk and reset so we can continue. + // ref: https://github.com/Mintplex-Labs/anything-llm/issues/416 + try { + chunk += message; + } catch (e) { + console.error(`Chunk appending error`, e); + chunk = ""; + } + continue; + } else { + chunk = ""; + } + } + + if (message == "[DONE]") { + writeResponseChunk(response, { + uuid, + sources, + type: "textResponseChunk", + textResponse: "", + close: true, + error: false, + }); + resolve(fullText); + } else { + let finishReason = null; + let token = ""; + try { + const json = JSON.parse(message); + token = json?.choices?.[0]?.delta?.content; + finishReason = json?.choices?.[0]?.finish_reason || null; + } catch { + continue; + } + + if (token) { + fullText += token; + writeResponseChunk(response, { + uuid, + sources: [], + type: "textResponseChunk", + textResponse: token, + close: false, + error: false, + }); + } + + if (finishReason !== null) { + writeResponseChunk(response, { + uuid, + sources, + type: "textResponseChunk", + textResponse: "", + close: true, + error: false, + }); + resolve(fullText); + } + } + } + }); + }); + } + // If stream is not a regular OpenAI Stream (like if using native model, Ollama, or most LangChain interfaces) // we can just iterate the stream content instead. if (!stream.hasOwnProperty("data")) { @@ -385,6 +504,7 @@ function handleStreamResponses(response, stream, responseProps) { } module.exports = { + VALID_CHAT_MODE, streamChatWithWorkspace, writeResponseChunk, }; diff --git a/server/utils/files/documentProcessor.js b/server/utils/files/documentProcessor.js index 5239a8708..27d0f5f2b 100644 --- a/server/utils/files/documentProcessor.js +++ b/server/utils/files/documentProcessor.js @@ -35,7 +35,7 @@ async function processDocument(filename = "") { .then((res) => res) .catch((e) => { console.log(e.message); - return { success: false, reason: e.message }; + return { success: false, reason: e.message, documents: [] }; }); } @@ -55,7 +55,7 @@ async function processLink(link = "") { .then((res) => res) .catch((e) => { console.log(e.message); - return { success: false, reason: e.message }; + return { success: false, reason: e.message, documents: [] }; }); } diff --git a/server/utils/files/index.js b/server/utils/files/index.js index b6c7a3070..e713a318a 100644 --- a/server/utils/files/index.js +++ b/server/utils/files/index.js @@ -2,32 +2,6 @@ const fs = require("fs"); const path = require("path"); const { v5: uuidv5 } = require("uuid"); -async function collectDocumentData(folderName = null) { - if (!folderName) throw new Error("No docPath provided in request"); - const folder = - process.env.NODE_ENV === "development" - ? path.resolve(__dirname, `../../storage/documents/${folderName}`) - : path.resolve(process.env.STORAGE_DIR, `documents/${folderName}`); - - const dirExists = fs.existsSync(folder); - if (!dirExists) - throw new Error( - `No documents folder for ${folderName} - did you run collector/main.py for this element?` - ); - - const files = fs.readdirSync(folder); - const fileData = []; - files.forEach((file) => { - if (path.extname(file) === ".json") { - const filePath = path.join(folder, file); - const data = fs.readFileSync(filePath, "utf8"); - console.log(`Parsing document: ${file}`); - fileData.push(JSON.parse(data)); - } - }); - return fileData; -} - // Should take in a folder that is a subfolder of documents // eg: youtube-subject/video-123.json async function fileData(filePath = null) { @@ -35,8 +9,15 @@ async function fileData(filePath = null) { const fullPath = process.env.NODE_ENV === "development" - ? path.resolve(__dirname, `../../storage/documents/${filePath}`) - : path.resolve(process.env.STORAGE_DIR, `documents/${filePath}`); + ? path.resolve( + __dirname, + `../../storage/documents/${normalizePath(filePath)}` + ) + : path.resolve( + process.env.STORAGE_DIR, + `documents/${normalizePath(filePath)}` + ); + const fileExists = fs.existsSync(fullPath); if (!fileExists) return null; @@ -142,11 +123,18 @@ async function storeVectorResult(vectorData = [], filename = null) { async function purgeSourceDocument(filename = null) { if (!filename) return; console.log(`Purging source document of ${filename}.`); - const filePath = process.env.NODE_ENV === "development" - ? path.resolve(__dirname, `../../storage/documents`, filename) - : path.resolve(process.env.STORAGE_DIR, `documents`, filename); + ? path.resolve( + __dirname, + `../../storage/documents`, + normalizePath(filename) + ) + : path.resolve( + process.env.STORAGE_DIR, + `documents`, + normalizePath(filename) + ); if (!fs.existsSync(filePath)) return; fs.rmSync(filePath); @@ -169,12 +157,54 @@ async function purgeVectorCache(filename = null) { return; } +// Search for a specific document by its unique name in the entire `documents` +// folder via iteration of all folders and checking if the expected file exists. +async function findDocumentInDocuments(documentName = null) { + if (!documentName) return null; + const documentsFolder = + process.env.NODE_ENV === "development" + ? path.resolve(__dirname, `../../storage/documents`) + : path.resolve(process.env.STORAGE_DIR, `documents`); + + for (const folder of fs.readdirSync(documentsFolder)) { + const isFolder = fs + .lstatSync(path.join(documentsFolder, folder)) + .isDirectory(); + if (!isFolder) continue; + + const targetFilename = normalizePath(documentName); + const targetFileLocation = path.join( + documentsFolder, + folder, + targetFilename + ); + if (!fs.existsSync(targetFileLocation)) continue; + + const fileData = fs.readFileSync(targetFileLocation, "utf8"); + const cachefilename = `${folder}/${targetFilename}`; + const { pageContent, ...metadata } = JSON.parse(fileData); + return { + name: targetFilename, + type: "file", + ...metadata, + cached: await cachedVectorInformation(cachefilename, true), + }; + } + + return null; +} + +function normalizePath(filepath = "") { + return path.normalize(filepath).replace(/^(\.\.(\/|\\|$))+/, ""); +} + module.exports = { + findDocumentInDocuments, cachedVectorInformation, - collectDocumentData, viewLocalFiles, purgeSourceDocument, purgeVectorCache, storeVectorResult, fileData, + normalizePath, }; diff --git a/server/utils/files/pfp.js b/server/utils/files/pfp.js index 943aa595f..dd6ba0fe2 100644 --- a/server/utils/files/pfp.js +++ b/server/utils/files/pfp.js @@ -2,6 +2,7 @@ const path = require("path"); const fs = require("fs"); const { getType } = require("mime"); const { User } = require("../../models/user"); +const { normalizePath } = require("."); function fetchPfp(pfpPath) { if (!fs.existsSync(pfpPath)) { @@ -32,8 +33,7 @@ async function determinePfpFilepath(id) { const basePath = process.env.STORAGE_DIR ? path.join(process.env.STORAGE_DIR, "assets/pfp") : path.join(__dirname, "../../storage/assets/pfp"); - const pfpFilepath = path.join(basePath, pfpFilename); - + const pfpFilepath = path.join(basePath, normalizePath(pfpFilename)); if (!fs.existsSync(pfpFilepath)) return null; return pfpFilepath; } diff --git a/server/utils/files/purgeDocument.js b/server/utils/files/purgeDocument.js index 27fe14710..46e9d37da 100644 --- a/server/utils/files/purgeDocument.js +++ b/server/utils/files/purgeDocument.js @@ -1,7 +1,6 @@ const fs = require("fs"); const path = require("path"); - -const { purgeVectorCache, purgeSourceDocument } = require("."); +const { purgeVectorCache, purgeSourceDocument, normalizePath } = require("."); const { Document } = require("../../models/documents"); const { Workspace } = require("../../models/workspace"); @@ -22,10 +21,10 @@ async function purgeFolder(folderName) { ? path.resolve(__dirname, `../../storage/documents`) : path.resolve(process.env.STORAGE_DIR, `documents`); - const folderPath = path.resolve(documentsFolder, folderName); + const folderPath = path.resolve(documentsFolder, normalizePath(folderName)); const filenames = fs .readdirSync(folderPath) - .map((file) => path.join(folderName, file)); + .map((file) => path.join(folderPath, file)); const workspaces = await Workspace.where(); const purgePromises = []; diff --git a/server/utils/helpers/customModels.js b/server/utils/helpers/customModels.js index 5bd7b299e..53c641e75 100644 --- a/server/utils/helpers/customModels.js +++ b/server/utils/helpers/customModels.js @@ -1,4 +1,12 @@ -const SUPPORT_CUSTOM_MODELS = ["openai", "localai", "ollama", "native-llm"]; +const { togetherAiModels } = require("../AiProviders/togetherAi"); +const SUPPORT_CUSTOM_MODELS = [ + "openai", + "localai", + "ollama", + "native-llm", + "togetherai", + "mistral", +]; async function getCustomModels(provider = "", apiKey = null, basePath = null) { if (!SUPPORT_CUSTOM_MODELS.includes(provider)) @@ -10,7 +18,11 @@ async function getCustomModels(provider = "", apiKey = null, basePath = null) { case "localai": return await localAIModels(basePath, apiKey); case "ollama": - return await ollamaAIModels(basePath, apiKey); + return await ollamaAIModels(basePath); + case "togetherai": + return await getTogetherAiModels(); + case "mistral": + return await getMistralModels(apiKey); case "native-llm": return nativeLLMModels(); default: @@ -44,7 +56,7 @@ async function openAiModels(apiKey = null) { async function localAIModels(basePath = null, apiKey = null) { const { Configuration, OpenAIApi } = require("openai"); const config = new Configuration({ - basePath, + basePath: basePath || process.env.LOCAL_AI_BASE_PATH, apiKey: apiKey || process.env.LOCAL_AI_API_KEY, }); const openai = new OpenAIApi(config); @@ -61,13 +73,14 @@ async function localAIModels(basePath = null, apiKey = null) { return { models, error: null }; } -async function ollamaAIModels(basePath = null, _apiKey = null) { +async function ollamaAIModels(basePath = null) { let url; try { - new URL(basePath); - if (basePath.split("").slice(-1)?.[0] === "/") + let urlPath = basePath ?? process.env.OLLAMA_BASE_PATH; + new URL(urlPath); + if (urlPath.split("").slice(-1)?.[0] === "/") throw new Error("BasePath Cannot end in /!"); - url = basePath; + url = urlPath; } catch { return { models: [], error: "Not a valid URL." }; } @@ -92,6 +105,41 @@ async function ollamaAIModels(basePath = null, _apiKey = null) { return { models, error: null }; } +async function getTogetherAiModels() { + const knownModels = togetherAiModels(); + if (!Object.keys(knownModels).length === 0) + return { models: [], error: null }; + + const models = Object.values(knownModels).map((model) => { + return { + id: model.id, + organization: model.organization, + name: model.name, + }; + }); + return { models, error: null }; +} + +async function getMistralModels(apiKey = null) { + const { Configuration, OpenAIApi } = require("openai"); + const config = new Configuration({ + apiKey: apiKey || process.env.MISTRAL_API_KEY, + basePath: "https://api.mistral.ai/v1", + }); + const openai = new OpenAIApi(config); + const models = await openai + .listModels() + .then((res) => res.data.data.filter((model) => !model.id.includes("embed"))) + .catch((e) => { + console.error(`Mistral:listModels`, e.message); + return []; + }); + + // Api Key was successful so lets save it for future uses + if (models.length > 0 && !!apiKey) process.env.MISTRAL_API_KEY = apiKey; + return { models, error: null }; +} + function nativeLLMModels() { const fs = require("fs"); const path = require("path"); diff --git a/server/utils/helpers/index.js b/server/utils/helpers/index.js index bde5e8a0a..b72bb7977 100644 --- a/server/utils/helpers/index.js +++ b/server/utils/helpers/index.js @@ -16,39 +16,51 @@ function getVectorDbClass() { case "qdrant": const { QDrant } = require("../vectorDbProviders/qdrant"); return QDrant; + case "milvus": + const { Milvus } = require("../vectorDbProviders/milvus"); + return Milvus; + case "zilliz": + const { Zilliz } = require("../vectorDbProviders/zilliz"); + return Zilliz; default: throw new Error("ENV: No VECTOR_DB value found in environment!"); } } -function getLLMProvider() { +function getLLMProvider(modelPreference = null) { const vectorSelection = process.env.LLM_PROVIDER || "openai"; const embedder = getEmbeddingEngineSelection(); switch (vectorSelection) { case "openai": const { OpenAiLLM } = require("../AiProviders/openAi"); - return new OpenAiLLM(embedder); + return new OpenAiLLM(embedder, modelPreference); case "azure": const { AzureOpenAiLLM } = require("../AiProviders/azureOpenAi"); - return new AzureOpenAiLLM(embedder); + return new AzureOpenAiLLM(embedder, modelPreference); case "anthropic": const { AnthropicLLM } = require("../AiProviders/anthropic"); - return new AnthropicLLM(embedder); + return new AnthropicLLM(embedder, modelPreference); case "gemini": const { GeminiLLM } = require("../AiProviders/gemini"); - return new GeminiLLM(embedder); + return new GeminiLLM(embedder, modelPreference); case "lmstudio": const { LMStudioLLM } = require("../AiProviders/lmStudio"); - return new LMStudioLLM(embedder); + return new LMStudioLLM(embedder, modelPreference); case "localai": const { LocalAiLLM } = require("../AiProviders/localAi"); - return new LocalAiLLM(embedder); + return new LocalAiLLM(embedder, modelPreference); case "ollama": const { OllamaAILLM } = require("../AiProviders/ollama"); - return new OllamaAILLM(embedder); + return new OllamaAILLM(embedder, modelPreference); + case "togetherai": + const { TogetherAiLLM } = require("../AiProviders/togetherAi"); + return new TogetherAiLLM(embedder, modelPreference); + case "mistral": + const { MistralLLM } = require("../AiProviders/mistral"); + return new MistralLLM(embedder, modelPreference); case "native": const { NativeLLM } = require("../AiProviders/native"); - return new NativeLLM(embedder); + return new NativeLLM(embedder, modelPreference); default: throw new Error("ENV: No LLM_PROVIDER value found in environment!"); } @@ -70,6 +82,7 @@ function getEmbeddingEngineSelection() { return new LocalAiEmbedder(); case "native": const { NativeEmbedder } = require("../EmbeddingEngines/native"); + console.log("\x1b[34m[INFO]\x1b[0m Using Native Embedder"); return new NativeEmbedder(); default: return null; diff --git a/server/utils/helpers/updateENV.js b/server/utils/helpers/updateENV.js index ebaadba73..45f006f6f 100644 --- a/server/utils/helpers/updateENV.js +++ b/server/utils/helpers/updateENV.js @@ -2,6 +2,7 @@ const KEY_MAPPING = { LLMProvider: { envKey: "LLM_PROVIDER", checks: [isNotEmpty, supportedLLM], + postUpdate: [wipeWorkspaceModelPreference], }, // OpenAI Settings OpenAiKey: { @@ -94,12 +95,26 @@ const KEY_MAPPING = { checks: [nonZero], }, + MistralApiKey: { + envKey: "MISTRAL_API_KEY", + checks: [isNotEmpty], + }, + MistralModelPref: { + envKey: "MISTRAL_MODEL_PREF", + checks: [isNotEmpty], + }, + // Native LLM Settings NativeLLMModelPref: { envKey: "NATIVE_LLM_MODEL_PREF", checks: [isDownloadedModel], }, + NativeLLMTokenLimit: { + envKey: "NATIVE_LLM_MODEL_TOKEN_LIMIT", + checks: [nonZero], + }, + EmbeddingEngine: { envKey: "EMBEDDING_ENGINE", checks: [supportedEmbeddingModel], @@ -170,6 +185,40 @@ const KEY_MAPPING = { checks: [], }, + // Milvus Options + MilvusAddress: { + envKey: "MILVUS_ADDRESS", + checks: [isValidURL, validDockerizedUrl], + }, + MilvusUsername: { + envKey: "MILVUS_USERNAME", + checks: [isNotEmpty], + }, + MilvusPassword: { + envKey: "MILVUS_PASSWORD", + checks: [isNotEmpty], + }, + + // Zilliz Cloud Options + ZillizEndpoint: { + envKey: "ZILLIZ_ENDPOINT", + checks: [isValidURL], + }, + ZillizApiToken: { + envKey: "ZILLIZ_API_TOKEN", + checks: [isNotEmpty], + }, + + // Together Ai Options + TogetherAiApiKey: { + envKey: "TOGETHER_AI_API_KEY", + checks: [isNotEmpty], + }, + TogetherAiModelPref: { + envKey: "TOGETHER_AI_MODEL_PREF", + checks: [isNotEmpty], + }, + // System Settings AuthToken: { envKey: "AUTH_TOKEN", @@ -233,7 +282,7 @@ function validOllamaLLMBasePath(input = "") { } function supportedLLM(input = "") { - return [ + const validSelection = [ "openai", "azure", "anthropic", @@ -242,7 +291,10 @@ function supportedLLM(input = "") { "localai", "ollama", "native", + "togetherai", + "mistral", ].includes(input); + return validSelection ? null : `${input} is not a valid LLM provider.`; } function validGeminiModel(input = "") { @@ -267,7 +319,15 @@ function supportedEmbeddingModel(input = "") { } function supportedVectorDB(input = "") { - const supported = ["chroma", "pinecone", "lancedb", "weaviate", "qdrant"]; + const supported = [ + "chroma", + "pinecone", + "lancedb", + "weaviate", + "qdrant", + "milvus", + "zilliz", + ]; return supported.includes(input) ? null : `Invalid VectorDB type. Must be one of ${supported.join(", ")}.`; @@ -329,11 +389,20 @@ function validDockerizedUrl(input = "") { return null; } +// If the LLMProvider has changed we need to reset all workspace model preferences to +// null since the provider<>model name combination will be invalid for whatever the new +// provider is. +async function wipeWorkspaceModelPreference(key, prev, next) { + if (prev === next) return; + const { Workspace } = require("../../models/workspace"); + await Workspace.resetWorkspaceChatModels(); +} + // This will force update .env variables which for any which reason were not able to be parsed or // read from an ENV file as this seems to be a complicating step for many so allowing people to write // to the process will at least alleviate that issue. It does not perform comprehensive validity checks or sanity checks // and is simply for debugging when the .env not found issue many come across. -function updateENV(newENVs = {}, force = false) { +async function updateENV(newENVs = {}, force = false) { let error = ""; const validKeys = Object.keys(KEY_MAPPING); const ENV_KEYS = Object.keys(newENVs).filter( @@ -341,21 +410,25 @@ function updateENV(newENVs = {}, force = false) { ); const newValues = {}; - ENV_KEYS.forEach((key) => { - const { envKey, checks } = KEY_MAPPING[key]; - const value = newENVs[key]; + for (const key of ENV_KEYS) { + const { envKey, checks, postUpdate = [] } = KEY_MAPPING[key]; + const prevValue = process.env[envKey]; + const nextValue = newENVs[key]; const errors = checks - .map((validityCheck) => validityCheck(value, force)) + .map((validityCheck) => validityCheck(nextValue, force)) .filter((err) => typeof err === "string"); if (errors.length > 0) { error += errors.join("\n"); - return; + break; } - newValues[key] = value; - process.env[envKey] = value; - }); + newValues[key] = nextValue; + process.env[envKey] = nextValue; + + for (const postUpdateFunc of postUpdate) + await postUpdateFunc(key, prevValue, nextValue); + } return { newValues, error: error?.length > 0 ? error : false }; } diff --git a/server/utils/http/index.js b/server/utils/http/index.js index d9d9d0485..d4551dae6 100644 --- a/server/utils/http/index.js +++ b/server/utils/http/index.js @@ -25,6 +25,8 @@ function makeJWT(info = {}, expiry = "30d") { return JWT.sign(info, process.env.JWT_SECRET, { expiresIn: expiry }); } +// Note: Only valid for finding users in multi-user mode +// as single-user mode with password is not a "user" async function userFromSession(request, response = null) { if (!!response && !!response.locals?.user) { return response.locals.user; diff --git a/server/utils/middleware/validatedRequest.js b/server/utils/middleware/validatedRequest.js index 275522bb9..6f3df26da 100644 --- a/server/utils/middleware/validatedRequest.js +++ b/server/utils/middleware/validatedRequest.js @@ -36,8 +36,9 @@ async function validatedRequest(request, response, next) { return; } + const bcrypt = require("bcrypt"); const { p } = decodeJWT(token); - if (p !== process.env.AUTH_TOKEN) { + if (!bcrypt.compareSync(p, bcrypt.hashSync(process.env.AUTH_TOKEN, 10))) { response.status(401).json({ error: "Invalid auth token found.", }); diff --git a/server/utils/vectorDbProviders/milvus/MILVUS_SETUP.md b/server/utils/vectorDbProviders/milvus/MILVUS_SETUP.md new file mode 100644 index 000000000..6bd9b8150 --- /dev/null +++ b/server/utils/vectorDbProviders/milvus/MILVUS_SETUP.md @@ -0,0 +1,40 @@ +# How to setup a local (or remote) Milvus Vector Database + +[Official Milvus Docs](https://milvus.io/docs/example_code.md) for reference. + +### How to get started + +**Requirements** + +Choose one of the following + +- Cloud + + - [Cloud account](https://cloud.zilliz.com/) + +- Local + - Docker + - `git` available in your CLI/terminal + +**Instructions** + +- Cloud + + - Create a Cluster on your cloud account + - Get connect Public Endpoint and Token + - Set .env.development variable in server + +- Local + - Download yaml file `wget https://github.com/milvus-io/milvus/releases/download/v2.3.4/milvus-standalone-docker-compose.yml -O docker-compose.yml` + - Start Milvus `sudo docker compose up -d` + - Check the containers are up and running `sudo docker compose ps` + - Get port number and set .env.development variable in server + +eg: `server/.env.development` + +``` +VECTOR_DB="milvus" +MILVUS_ADDRESS="http://localhost:19530" +MILVUS_USERNAME=minioadmin # Whatever your username and password are +MILVUS_PASSWORD=minioadmin +``` diff --git a/server/utils/vectorDbProviders/milvus/index.js b/server/utils/vectorDbProviders/milvus/index.js new file mode 100644 index 000000000..cc934a9a2 --- /dev/null +++ b/server/utils/vectorDbProviders/milvus/index.js @@ -0,0 +1,364 @@ +const { + DataType, + MetricType, + IndexType, + MilvusClient, +} = require("@zilliz/milvus2-sdk-node"); +const { RecursiveCharacterTextSplitter } = require("langchain/text_splitter"); +const { v4: uuidv4 } = require("uuid"); +const { storeVectorResult, cachedVectorInformation } = require("../../files"); +const { + toChunks, + getLLMProvider, + getEmbeddingEngineSelection, +} = require("../../helpers"); + +const Milvus = { + name: "Milvus", + connect: async function () { + if (process.env.VECTOR_DB !== "milvus") + throw new Error("Milvus::Invalid ENV settings"); + + const client = new MilvusClient({ + address: process.env.MILVUS_ADDRESS, + username: process.env.MILVUS_USERNAME, + password: process.env.MILVUS_PASSWORD, + }); + + const { isHealthy } = await client.checkHealth(); + if (!isHealthy) + throw new Error( + "MilvusDB::Invalid Heartbeat received - is the instance online?" + ); + + return { client }; + }, + heartbeat: async function () { + await this.connect(); + return { heartbeat: Number(new Date()) }; + }, + totalVectors: async function () { + const { client } = await this.connect(); + const { collection_names } = await client.listCollections(); + const total = collection_names.reduce(async (acc, collection_name) => { + const statistics = await client.getCollectionStatistics({ + collection_name, + }); + return Number(acc) + Number(statistics?.data?.row_count ?? 0); + }, 0); + return total; + }, + namespaceCount: async function (_namespace = null) { + const { client } = await this.connect(); + const statistics = await client.getCollectionStatistics({ + collection_name: _namespace, + }); + return Number(statistics?.data?.row_count ?? 0); + }, + namespace: async function (client, namespace = null) { + if (!namespace) throw new Error("No namespace value provided."); + const collection = await client + .getCollectionStatistics({ collection_name: namespace }) + .catch(() => null); + return collection; + }, + hasNamespace: async function (namespace = null) { + if (!namespace) return false; + const { client } = await this.connect(); + return await this.namespaceExists(client, namespace); + }, + namespaceExists: async function (client, namespace = null) { + if (!namespace) throw new Error("No namespace value provided."); + const { value } = await client + .hasCollection({ collection_name: namespace }) + .catch((e) => { + console.error("MilvusDB::namespaceExists", e.message); + return { value: false }; + }); + return value; + }, + deleteVectorsInNamespace: async function (client, namespace = null) { + await client.dropCollection({ collection_name: namespace }); + return true; + }, + // Milvus requires a dimension aspect for collection creation + // we pass this in from the first chunk to infer the dimensions like other + // providers do. + getOrCreateCollection: async function (client, namespace, dimensions = null) { + const isExists = await this.namespaceExists(client, namespace); + if (!isExists) { + if (!dimensions) + throw new Error( + `Milvus:getOrCreateCollection Unable to infer vector dimension from input. Open an issue on Github for support.` + ); + + await client.createCollection({ + collection_name: namespace, + fields: [ + { + name: "id", + description: "id", + data_type: DataType.VarChar, + max_length: 255, + is_primary_key: true, + }, + { + name: "vector", + description: "vector", + data_type: DataType.FloatVector, + dim: dimensions, + }, + { + name: "metadata", + decription: "metadata", + data_type: DataType.JSON, + }, + ], + }); + await client.createIndex({ + collection_name: namespace, + field_name: "vector", + index_type: IndexType.AUTOINDEX, + metric_type: MetricType.COSINE, + }); + await client.loadCollectionSync({ + collection_name: namespace, + }); + } + }, + addDocumentToNamespace: async function ( + namespace, + documentData = {}, + fullFilePath = null + ) { + const { DocumentVectors } = require("../../../models/vectors"); + try { + let vectorDimension = null; + const { pageContent, docId, ...metadata } = documentData; + if (!pageContent || pageContent.length == 0) return false; + + console.log("Adding new vectorized document into namespace", namespace); + const cacheResult = await cachedVectorInformation(fullFilePath); + if (cacheResult.exists) { + const { client } = await this.connect(); + const { chunks } = cacheResult; + const documentVectors = []; + vectorDimension = chunks[0][0].values.length || null; + + await this.getOrCreateCollection(client, namespace, vectorDimension); + for (const chunk of chunks) { + // Before sending to Pinecone and saving the records to our db + // we need to assign the id of each chunk that is stored in the cached file. + const newChunks = chunk.map((chunk) => { + const id = uuidv4(); + documentVectors.push({ docId, vectorId: id }); + return { id, vector: chunk.values, metadata: chunk.metadata }; + }); + const insertResult = await client.insert({ + collection_name: namespace, + data: newChunks, + }); + + if (insertResult?.status.error_code !== "Success") { + throw new Error( + `Error embedding into Milvus! Reason:${insertResult?.status.reason}` + ); + } + } + await DocumentVectors.bulkInsert(documentVectors); + await client.flushSync({ collection_names: [namespace] }); + return true; + } + + const textSplitter = new RecursiveCharacterTextSplitter({ + chunkSize: + getEmbeddingEngineSelection()?.embeddingMaxChunkLength || 1_000, + chunkOverlap: 20, + }); + const textChunks = await textSplitter.splitText(pageContent); + + console.log("Chunks created from document:", textChunks.length); + const LLMConnector = getLLMProvider(); + const documentVectors = []; + const vectors = []; + const vectorValues = await LLMConnector.embedChunks(textChunks); + + if (!!vectorValues && vectorValues.length > 0) { + for (const [i, vector] of vectorValues.entries()) { + if (!vectorDimension) vectorDimension = vector.length; + const vectorRecord = { + id: uuidv4(), + values: vector, + // [DO NOT REMOVE] + // LangChain will be unable to find your text if you embed manually and dont include the `text` key. + metadata: { ...metadata, text: textChunks[i] }, + }; + + vectors.push(vectorRecord); + documentVectors.push({ docId, vectorId: vectorRecord.id }); + } + } else { + throw new Error( + "Could not embed document chunks! This document will not be recorded." + ); + } + + if (vectors.length > 0) { + const chunks = []; + const { client } = await this.connect(); + await this.getOrCreateCollection(client, namespace, vectorDimension); + + console.log("Inserting vectorized chunks into Milvus."); + for (const chunk of toChunks(vectors, 100)) { + chunks.push(chunk); + const insertResult = await client.insert({ + collection_name: namespace, + data: chunk.map((item) => ({ + id: item.id, + vector: item.values, + metadata: chunk.metadata, + })), + }); + + if (insertResult?.status.error_code !== "Success") { + throw new Error( + `Error embedding into Milvus! Reason:${insertResult?.status.reason}` + ); + } + } + await storeVectorResult(chunks, fullFilePath); + await client.flushSync({ collection_names: [namespace] }); + } + + await DocumentVectors.bulkInsert(documentVectors); + return true; + } catch (e) { + console.error(e); + console.error("addDocumentToNamespace", e.message); + return false; + } + }, + deleteDocumentFromNamespace: async function (namespace, docId) { + const { DocumentVectors } = require("../../../models/vectors"); + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) return; + const knownDocuments = await DocumentVectors.where({ docId }); + if (knownDocuments.length === 0) return; + + const vectorIds = knownDocuments.map((doc) => doc.vectorId); + const queryIn = vectorIds.map((v) => `'${v}'`).join(","); + await client.deleteEntities({ + collection_name: namespace, + expr: `id in [${queryIn}]`, + }); + + const indexes = knownDocuments.map((doc) => doc.id); + await DocumentVectors.deleteIds(indexes); + + // Even after flushing Milvus can take some time to re-calc the count + // so all we can hope to do is flushSync so that the count can be correct + // on a later call. + await client.flushSync({ collection_names: [namespace] }); + return true; + }, + performSimilaritySearch: async function ({ + namespace = null, + input = "", + LLMConnector = null, + similarityThreshold = 0.25, + }) { + if (!namespace || !input || !LLMConnector) + throw new Error("Invalid request to performSimilaritySearch."); + + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) { + return { + contextTexts: [], + sources: [], + message: "Invalid query - no documents found for workspace!", + }; + } + + const queryVector = await LLMConnector.embedTextInput(input); + const { contextTexts, sourceDocuments } = await this.similarityResponse( + client, + namespace, + queryVector, + similarityThreshold + ); + + const sources = sourceDocuments.map((metadata, i) => { + return { ...metadata, text: contextTexts[i] }; + }); + return { + contextTexts, + sources: this.curateSources(sources), + message: false, + }; + }, + similarityResponse: async function ( + client, + namespace, + queryVector, + similarityThreshold = 0.25 + ) { + const result = { + contextTexts: [], + sourceDocuments: [], + scores: [], + }; + const response = await client.search({ + collection_name: namespace, + vectors: queryVector, + }); + response.results.forEach((match) => { + if (match.score < similarityThreshold) return; + result.contextTexts.push(match.metadata.text); + result.sourceDocuments.push(match); + result.scores.push(match.score); + }); + return result; + }, + "namespace-stats": async function (reqBody = {}) { + const { namespace = null } = reqBody; + if (!namespace) throw new Error("namespace required"); + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) + throw new Error("Namespace by that name does not exist."); + const stats = await this.namespace(client, namespace); + return stats + ? stats + : { message: "No stats were able to be fetched from DB for namespace" }; + }, + "delete-namespace": async function (reqBody = {}) { + const { namespace = null } = reqBody; + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) + throw new Error("Namespace by that name does not exist."); + + const statistics = await this.namespace(client, namespace); + await this.deleteVectorsInNamespace(client, namespace); + const vectorCount = Number(statistics?.data?.row_count ?? 0); + return { + message: `Namespace ${namespace} was deleted along with ${vectorCount} vectors.`, + }; + }, + curateSources: function (sources = []) { + const documents = []; + for (const source of sources) { + const { metadata = {} } = source; + if (Object.keys(metadata).length > 0) { + documents.push({ + ...metadata, + ...(source.hasOwnProperty("pageContent") + ? { text: source.pageContent } + : {}), + }); + } + } + + return documents; + }, +}; + +module.exports.Milvus = Milvus; diff --git a/server/utils/vectorDbProviders/qdrant/index.js b/server/utils/vectorDbProviders/qdrant/index.js index 49b25a3d6..2783cde93 100644 --- a/server/utils/vectorDbProviders/qdrant/index.js +++ b/server/utils/vectorDbProviders/qdrant/index.js @@ -108,13 +108,20 @@ const QDrant = { await client.deleteCollection(namespace); return true; }, - getOrCreateCollection: async function (client, namespace) { + // QDrant requires a dimension aspect for collection creation + // we pass this in from the first chunk to infer the dimensions like other + // providers do. + getOrCreateCollection: async function (client, namespace, dimensions = null) { if (await this.namespaceExists(client, namespace)) { return await client.getCollection(namespace); } + if (!dimensions) + throw new Error( + `Qdrant:getOrCreateCollection Unable to infer vector dimension from input. Open an issue on Github for support.` + ); await client.createCollection(namespace, { vectors: { - size: 1536, //TODO: Fixed to OpenAI models - when other embeddings exist make variable. + size: dimensions, distance: "Cosine", }, }); @@ -127,6 +134,7 @@ const QDrant = { ) { const { DocumentVectors } = require("../../../models/vectors"); try { + let vectorDimension = null; const { pageContent, docId, ...metadata } = documentData; if (!pageContent || pageContent.length == 0) return false; @@ -134,15 +142,20 @@ const QDrant = { const cacheResult = await cachedVectorInformation(fullFilePath); if (cacheResult.exists) { const { client } = await this.connect(); - const collection = await this.getOrCreateCollection(client, namespace); + const { chunks } = cacheResult; + const documentVectors = []; + vectorDimension = chunks[0][0].vector.length || null; + + const collection = await this.getOrCreateCollection( + client, + namespace, + vectorDimension + ); if (!collection) throw new Error("Failed to create new QDrant collection!", { namespace, }); - const { chunks } = cacheResult; - const documentVectors = []; - for (const chunk of chunks) { const submission = { ids: [], @@ -204,6 +217,7 @@ const QDrant = { if (!!vectorValues && vectorValues.length > 0) { for (const [i, vector] of vectorValues.entries()) { + if (!vectorDimension) vectorDimension = vector.length; const vectorRecord = { id: uuidv4(), vector: vector, @@ -227,7 +241,11 @@ const QDrant = { } const { client } = await this.connect(); - const collection = await this.getOrCreateCollection(client, namespace); + const collection = await this.getOrCreateCollection( + client, + namespace, + vectorDimension + ); if (!collection) throw new Error("Failed to create new QDrant collection!", { namespace, diff --git a/server/utils/vectorDbProviders/zilliz/index.js b/server/utils/vectorDbProviders/zilliz/index.js new file mode 100644 index 000000000..b8493e1c2 --- /dev/null +++ b/server/utils/vectorDbProviders/zilliz/index.js @@ -0,0 +1,365 @@ +const { + DataType, + MetricType, + IndexType, + MilvusClient, +} = require("@zilliz/milvus2-sdk-node"); +const { RecursiveCharacterTextSplitter } = require("langchain/text_splitter"); +const { v4: uuidv4 } = require("uuid"); +const { storeVectorResult, cachedVectorInformation } = require("../../files"); +const { + toChunks, + getLLMProvider, + getEmbeddingEngineSelection, +} = require("../../helpers"); + +// Zilliz is basically a copy of Milvus DB class with a different constructor +// to connect to the cloud +const Zilliz = { + name: "Zilliz", + connect: async function () { + if (process.env.VECTOR_DB !== "zilliz") + throw new Error("Zilliz::Invalid ENV settings"); + + const client = new MilvusClient({ + address: process.env.ZILLIZ_ENDPOINT, + token: process.env.ZILLIZ_API_TOKEN, + }); + + const { isHealthy } = await client.checkHealth(); + if (!isHealthy) + throw new Error( + "Zilliz::Invalid Heartbeat received - is the instance online?" + ); + + return { client }; + }, + heartbeat: async function () { + await this.connect(); + return { heartbeat: Number(new Date()) }; + }, + totalVectors: async function () { + const { client } = await this.connect(); + const { collection_names } = await client.listCollections(); + const total = collection_names.reduce(async (acc, collection_name) => { + const statistics = await client.getCollectionStatistics({ + collection_name, + }); + return Number(acc) + Number(statistics?.data?.row_count ?? 0); + }, 0); + return total; + }, + namespaceCount: async function (_namespace = null) { + const { client } = await this.connect(); + const statistics = await client.getCollectionStatistics({ + collection_name: _namespace, + }); + return Number(statistics?.data?.row_count ?? 0); + }, + namespace: async function (client, namespace = null) { + if (!namespace) throw new Error("No namespace value provided."); + const collection = await client + .getCollectionStatistics({ collection_name: namespace }) + .catch(() => null); + return collection; + }, + hasNamespace: async function (namespace = null) { + if (!namespace) return false; + const { client } = await this.connect(); + return await this.namespaceExists(client, namespace); + }, + namespaceExists: async function (client, namespace = null) { + if (!namespace) throw new Error("No namespace value provided."); + const { value } = await client + .hasCollection({ collection_name: namespace }) + .catch((e) => { + console.error("Zilliz::namespaceExists", e.message); + return { value: false }; + }); + return value; + }, + deleteVectorsInNamespace: async function (client, namespace = null) { + await client.dropCollection({ collection_name: namespace }); + return true; + }, + // Zilliz requires a dimension aspect for collection creation + // we pass this in from the first chunk to infer the dimensions like other + // providers do. + getOrCreateCollection: async function (client, namespace, dimensions = null) { + const isExists = await this.namespaceExists(client, namespace); + if (!isExists) { + if (!dimensions) + throw new Error( + `Zilliz:getOrCreateCollection Unable to infer vector dimension from input. Open an issue on Github for support.` + ); + + await client.createCollection({ + collection_name: namespace, + fields: [ + { + name: "id", + description: "id", + data_type: DataType.VarChar, + max_length: 255, + is_primary_key: true, + }, + { + name: "vector", + description: "vector", + data_type: DataType.FloatVector, + dim: dimensions, + }, + { + name: "metadata", + decription: "metadata", + data_type: DataType.JSON, + }, + ], + }); + await client.createIndex({ + collection_name: namespace, + field_name: "vector", + index_type: IndexType.AUTOINDEX, + metric_type: MetricType.COSINE, + }); + await client.loadCollectionSync({ + collection_name: namespace, + }); + } + }, + addDocumentToNamespace: async function ( + namespace, + documentData = {}, + fullFilePath = null + ) { + const { DocumentVectors } = require("../../../models/vectors"); + try { + let vectorDimension = null; + const { pageContent, docId, ...metadata } = documentData; + if (!pageContent || pageContent.length == 0) return false; + + console.log("Adding new vectorized document into namespace", namespace); + const cacheResult = await cachedVectorInformation(fullFilePath); + if (cacheResult.exists) { + const { client } = await this.connect(); + const { chunks } = cacheResult; + const documentVectors = []; + vectorDimension = chunks[0][0].values.length || null; + + await this.getOrCreateCollection(client, namespace, vectorDimension); + for (const chunk of chunks) { + // Before sending to Pinecone and saving the records to our db + // we need to assign the id of each chunk that is stored in the cached file. + const newChunks = chunk.map((chunk) => { + const id = uuidv4(); + documentVectors.push({ docId, vectorId: id }); + return { id, vector: chunk.values, metadata: chunk.metadata }; + }); + const insertResult = await client.insert({ + collection_name: namespace, + data: newChunks, + }); + + if (insertResult?.status.error_code !== "Success") { + throw new Error( + `Error embedding into Zilliz! Reason:${insertResult?.status.reason}` + ); + } + } + await DocumentVectors.bulkInsert(documentVectors); + await client.flushSync({ collection_names: [namespace] }); + return true; + } + + const textSplitter = new RecursiveCharacterTextSplitter({ + chunkSize: + getEmbeddingEngineSelection()?.embeddingMaxChunkLength || 1_000, + chunkOverlap: 20, + }); + const textChunks = await textSplitter.splitText(pageContent); + + console.log("Chunks created from document:", textChunks.length); + const LLMConnector = getLLMProvider(); + const documentVectors = []; + const vectors = []; + const vectorValues = await LLMConnector.embedChunks(textChunks); + + if (!!vectorValues && vectorValues.length > 0) { + for (const [i, vector] of vectorValues.entries()) { + if (!vectorDimension) vectorDimension = vector.length; + const vectorRecord = { + id: uuidv4(), + values: vector, + // [DO NOT REMOVE] + // LangChain will be unable to find your text if you embed manually and dont include the `text` key. + metadata: { ...metadata, text: textChunks[i] }, + }; + + vectors.push(vectorRecord); + documentVectors.push({ docId, vectorId: vectorRecord.id }); + } + } else { + throw new Error( + "Could not embed document chunks! This document will not be recorded." + ); + } + + if (vectors.length > 0) { + const chunks = []; + const { client } = await this.connect(); + await this.getOrCreateCollection(client, namespace, vectorDimension); + + console.log("Inserting vectorized chunks into Zilliz."); + for (const chunk of toChunks(vectors, 100)) { + chunks.push(chunk); + const insertResult = await client.insert({ + collection_name: namespace, + data: chunk.map((item) => ({ + id: item.id, + vector: item.values, + metadata: chunk.metadata, + })), + }); + + if (insertResult?.status.error_code !== "Success") { + throw new Error( + `Error embedding into Zilliz! Reason:${insertResult?.status.reason}` + ); + } + } + await storeVectorResult(chunks, fullFilePath); + await client.flushSync({ collection_names: [namespace] }); + } + + await DocumentVectors.bulkInsert(documentVectors); + return true; + } catch (e) { + console.error(e); + console.error("addDocumentToNamespace", e.message); + return false; + } + }, + deleteDocumentFromNamespace: async function (namespace, docId) { + const { DocumentVectors } = require("../../../models/vectors"); + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) return; + const knownDocuments = await DocumentVectors.where({ docId }); + if (knownDocuments.length === 0) return; + + const vectorIds = knownDocuments.map((doc) => doc.vectorId); + const queryIn = vectorIds.map((v) => `'${v}'`).join(","); + await client.deleteEntities({ + collection_name: namespace, + expr: `id in [${queryIn}]`, + }); + + const indexes = knownDocuments.map((doc) => doc.id); + await DocumentVectors.deleteIds(indexes); + + // Even after flushing Zilliz can take some time to re-calc the count + // so all we can hope to do is flushSync so that the count can be correct + // on a later call. + await client.flushSync({ collection_names: [namespace] }); + return true; + }, + performSimilaritySearch: async function ({ + namespace = null, + input = "", + LLMConnector = null, + similarityThreshold = 0.25, + }) { + if (!namespace || !input || !LLMConnector) + throw new Error("Invalid request to performSimilaritySearch."); + + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) { + return { + contextTexts: [], + sources: [], + message: "Invalid query - no documents found for workspace!", + }; + } + + const queryVector = await LLMConnector.embedTextInput(input); + const { contextTexts, sourceDocuments } = await this.similarityResponse( + client, + namespace, + queryVector, + similarityThreshold + ); + + const sources = sourceDocuments.map((metadata, i) => { + return { ...metadata, text: contextTexts[i] }; + }); + return { + contextTexts, + sources: this.curateSources(sources), + message: false, + }; + }, + similarityResponse: async function ( + client, + namespace, + queryVector, + similarityThreshold = 0.25 + ) { + const result = { + contextTexts: [], + sourceDocuments: [], + scores: [], + }; + const response = await client.search({ + collection_name: namespace, + vectors: queryVector, + }); + response.results.forEach((match) => { + if (match.score < similarityThreshold) return; + result.contextTexts.push(match.metadata.text); + result.sourceDocuments.push(match); + result.scores.push(match.score); + }); + return result; + }, + "namespace-stats": async function (reqBody = {}) { + const { namespace = null } = reqBody; + if (!namespace) throw new Error("namespace required"); + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) + throw new Error("Namespace by that name does not exist."); + const stats = await this.namespace(client, namespace); + return stats + ? stats + : { message: "No stats were able to be fetched from DB for namespace" }; + }, + "delete-namespace": async function (reqBody = {}) { + const { namespace = null } = reqBody; + const { client } = await this.connect(); + if (!(await this.namespaceExists(client, namespace))) + throw new Error("Namespace by that name does not exist."); + + const statistics = await this.namespace(client, namespace); + await this.deleteVectorsInNamespace(client, namespace); + const vectorCount = Number(statistics?.data?.row_count ?? 0); + return { + message: `Namespace ${namespace} was deleted along with ${vectorCount} vectors.`, + }; + }, + curateSources: function (sources = []) { + const documents = []; + for (const source of sources) { + const { metadata = {} } = source; + if (Object.keys(metadata).length > 0) { + documents.push({ + ...metadata, + ...(source.hasOwnProperty("pageContent") + ? { text: source.pageContent } + : {}), + }); + } + } + + return documents; + }, +}; + +module.exports.Zilliz = Zilliz; diff --git a/server/yarn.lock b/server/yarn.lock index 6215bf01f..cc129dfe9 100644 --- a/server/yarn.lock +++ b/server/yarn.lock @@ -160,6 +160,20 @@ "@azure/logger" "^1.0.3" tslib "^2.4.0" +"@colors/colors@1.6.0", "@colors/colors@^1.6.0": + version "1.6.0" + resolved "https://registry.yarnpkg.com/@colors/colors/-/colors-1.6.0.tgz#ec6cd237440700bc23ca23087f513c75508958b0" + integrity sha512-Ir+AOibqzrIsL6ajt3Rz3LskB7OiMVHqltZmspbW/TJuTVuyOMirVqAkjfY6JISiLHgyNqicAC8AyHHGzNd/dA== + +"@dabh/diagnostics@^2.0.2": + version "2.0.3" + resolved "https://registry.yarnpkg.com/@dabh/diagnostics/-/diagnostics-2.0.3.tgz#7f7e97ee9a725dffc7808d93668cc984e1dc477a" + integrity sha512-hrlQOIi7hAfzsMqlGSFyVucrx38O+j6wiGOf//H2ecvIEqYN4ADBSS2iLMh5UFyDunCNniUIPk/q3riFv45xRA== + dependencies: + colorspace "1.1.x" + enabled "2.0.x" + kuler "^2.0.0" + "@eslint-community/eslint-utils@^4.2.0": version "4.4.0" resolved "https://registry.yarnpkg.com/@eslint-community/eslint-utils/-/eslint-utils-4.4.0.tgz#a23514e8fb9af1269d5f7788aa556798d61c6b59" @@ -214,6 +228,35 @@ resolved "https://registry.yarnpkg.com/@graphql-typed-document-node/core/-/core-3.2.0.tgz#5f3d96ec6b2354ad6d8a28bf216a1d97b5426861" integrity sha512-mB9oAsNCm9aM3/SOv4YtBMqZbYj10R7dkq8byBqxGY/ncFwhf2oQzMV+LCRlWoDSEBJ3COiR1yeDvMtsoOsuFQ== +"@grpc/grpc-js@1.8.17": + version "1.8.17" + resolved "https://registry.yarnpkg.com/@grpc/grpc-js/-/grpc-js-1.8.17.tgz#a3a2f826fc033eae7d2f5ee41e0ab39cee948838" + integrity sha512-DGuSbtMFbaRsyffMf+VEkVu8HkSXEUfO3UyGJNtqxW9ABdtTIA+2UXAJpwbJS+xfQxuwqLUeELmL6FuZkOqPxw== + dependencies: + "@grpc/proto-loader" "^0.7.0" + "@types/node" ">=12.12.47" + +"@grpc/proto-loader@0.7.7": + version "0.7.7" + resolved "https://registry.yarnpkg.com/@grpc/proto-loader/-/proto-loader-0.7.7.tgz#d33677a77eea8407f7c66e2abd97589b60eb4b21" + integrity sha512-1TIeXOi8TuSCQprPItwoMymZXxWT0CPxUhkrkeCUH+D8U7QDwQ6b7SUz2MaLuWM2llT+J/TVFLmQI5KtML3BhQ== + dependencies: + "@types/long" "^4.0.1" + lodash.camelcase "^4.3.0" + long "^4.0.0" + protobufjs "^7.0.0" + yargs "^17.7.2" + +"@grpc/proto-loader@^0.7.0": + version "0.7.10" + resolved "https://registry.yarnpkg.com/@grpc/proto-loader/-/proto-loader-0.7.10.tgz#6bf26742b1b54d0a473067743da5d3189d06d720" + integrity sha512-CAqDfoaQ8ykFd9zqBDn4k6iWT9loLAlc2ETmDFS9JCD70gDcnA4L3AFEo2iV7KyAtAAHFW9ftq1Fz+Vsgq80RQ== + dependencies: + lodash.camelcase "^4.3.0" + long "^5.0.0" + protobufjs "^7.2.4" + yargs "^17.7.2" + "@hapi/hoek@^9.0.0": version "9.3.0" resolved "https://registry.yarnpkg.com/@hapi/hoek/-/hoek-9.3.0.tgz#8368869dcb735be2e7f5cb7647de78e167a251fb" @@ -226,6 +269,11 @@ dependencies: "@hapi/hoek" "^9.0.0" +"@huggingface/jinja@^0.1.0": + version "0.1.2" + resolved "https://registry.yarnpkg.com/@huggingface/jinja/-/jinja-0.1.2.tgz#073fa0a68ef481a1806b0186bbafd8013e586fbe" + integrity sha512-x5mpbfJt1nKmVep5WNP5VjNsjWApWNj8pPYI+uYMkBWH9bWUJmQmHt2lbf0VCoQd54Oq3XuFEh/UyoVh7rPxmg== + "@humanwhocodes/config-array@^0.11.13": version "0.11.13" resolved "https://registry.yarnpkg.com/@humanwhocodes/config-array/-/config-array-0.11.13.tgz#075dc9684f40a531d9b26b0822153c1e832ee297" @@ -755,6 +803,13 @@ resolved "https://registry.yarnpkg.com/@types/node/-/node-18.14.5.tgz#4a13a6445862159303fc38586598a9396fc408b3" integrity sha512-CRT4tMK/DHYhw1fcCEBwME9CSaZNclxfzVMe7GsO6ULSwsttbj70wSiX6rZdIjGblu93sTJxLdhNIT85KKI7Qw== +"@types/node@>=12.12.47": + version "20.10.8" + resolved "https://registry.yarnpkg.com/@types/node/-/node-20.10.8.tgz#f1e223cbde9e25696661d167a5b93a9b2a5d57c7" + integrity sha512-f8nQs3cLxbAFc00vEU59yf9UyGUftkPaLGfvbVOIDdx2i1b8epBqj2aNGyP19fiyXWvlmZ7qC1XLjAzw/OKIeA== + dependencies: + undici-types "~5.26.4" + "@types/node@>=13.7.0": version "20.10.3" resolved "https://registry.yarnpkg.com/@types/node/-/node-20.10.3.tgz#4900adcc7fc189d5af5bb41da8f543cea6962030" @@ -779,6 +834,11 @@ resolved "https://registry.yarnpkg.com/@types/retry/-/retry-0.12.0.tgz#2b35eccfcee7d38cd72ad99232fbd58bffb3c84d" integrity sha512-wWKOClTTiizcZhXnPY4wikVAwmdYHp8q6DmC+EJUzAMsycb7HB32Kh9RN4+0gExjmPmZSAQjgURXIGATPegAvA== +"@types/triple-beam@^1.3.2": + version "1.3.5" + resolved "https://registry.yarnpkg.com/@types/triple-beam/-/triple-beam-1.3.5.tgz#74fef9ffbaa198eb8b588be029f38b00299caa2c" + integrity sha512-6WaYesThRMCl19iryMYP7/x2OVgCtbIVflDGFpWnb9irXI3UjYE4AzmYuiUKY1AJstGijoY+MgUszMgRxIYTYw== + "@types/uuid@^9.0.1": version "9.0.7" resolved "https://registry.yarnpkg.com/@types/uuid/-/uuid-9.0.7.tgz#b14cebc75455eeeb160d5fe23c2fcc0c64f724d8" @@ -796,16 +856,29 @@ resolved "https://registry.yarnpkg.com/@ungap/structured-clone/-/structured-clone-1.2.0.tgz#756641adb587851b5ccb3e095daf27ae581c8406" integrity sha512-zuVdFrMJiuCDQUMCzQaD6KL28MjnqqN8XnAqiEq9PNm/hCPTSGfrXCOfwj1ow4LFb/tNymJPwsNbVePc1xFqrQ== -"@xenova/transformers@^2.10.0": - version "2.10.0" - resolved "https://registry.yarnpkg.com/@xenova/transformers/-/transformers-2.10.0.tgz#ae97d724a3addf78de7314336a9f7b28ed96a140" - integrity sha512-Al9WKiOsimAC3mU9Ef434GkHF0izmeAM7mMMx5npdWsWLAYL8fmJXCrULj6uCfjomMQ7jyN9rDtKpp570hffiw== +"@xenova/transformers@^2.14.0": + version "2.14.0" + resolved "https://registry.yarnpkg.com/@xenova/transformers/-/transformers-2.14.0.tgz#6fe128957e64377ca4fca910e77f6092f3f3512a" + integrity sha512-rQ3O7SW5EM64b6XFZGx3XQ2cfiroefxUwU9ShfSpEZyhd082GvwNJJKndxgaukse1hZP1JUDoT0DfjDiq4IZiw== dependencies: + "@huggingface/jinja" "^0.1.0" onnxruntime-web "1.14.0" sharp "^0.32.0" optionalDependencies: onnxruntime-node "1.14.0" +"@zilliz/milvus2-sdk-node@^2.3.5": + version "2.3.5" + resolved "https://registry.yarnpkg.com/@zilliz/milvus2-sdk-node/-/milvus2-sdk-node-2.3.5.tgz#6540bc03ebb99ab35f63e4eca7a1fd3ede2cf38c" + integrity sha512-bWbQnhvu+7jZXoqI+qySycwph3vloy0LDV54TBY4wRmu6HhMlqIqyIiI8sQNeSJFs8M1jHg1PlmhE/dvckA1bA== + dependencies: + "@grpc/grpc-js" "1.8.17" + "@grpc/proto-loader" "0.7.7" + dayjs "^1.11.7" + lru-cache "^9.1.2" + protobufjs "7.2.4" + winston "^3.9.0" + abbrev@1: version "1.1.1" resolved "https://registry.yarnpkg.com/abbrev/-/abbrev-1.1.1.tgz#f8f2c887ad10bf67f634f005b6987fed3179aac8" @@ -1487,7 +1560,7 @@ cmake-js@^7.2.1: which "^2.0.2" yargs "^17.6.0" -color-convert@^1.9.0: +color-convert@^1.9.0, color-convert@^1.9.3: version "1.9.3" resolved "https://registry.yarnpkg.com/color-convert/-/color-convert-1.9.3.tgz#bb71850690e1f136567de629d2d5471deda4c1e8" integrity sha512-QfAUtd+vFdAtFQcC8CCyYt1fYWxSqAiK2cSD6zDB8N3cpsEBAvRxp9zOGg6G/SHHJYAT88/az/IuDGALsNVbGg== @@ -1511,7 +1584,7 @@ color-name@^1.0.0, color-name@~1.1.4: resolved "https://registry.yarnpkg.com/color-name/-/color-name-1.1.4.tgz#c2a09a87acbde69543de6f63fa3995c826c536a2" integrity sha512-dOy+3AuW3a2wNbZHIuMZpTcgjGuLU/uBL/ubcZF9OXbDo8ff4O8yVp5Bf0efS8uEoYo5q4Fx7dY9OgQGXgAsQA== -color-string@^1.9.0: +color-string@^1.6.0, color-string@^1.9.0: version "1.9.1" resolved "https://registry.yarnpkg.com/color-string/-/color-string-1.9.1.tgz#4467f9146f036f855b764dfb5bf8582bf342c7a4" integrity sha512-shrVawQFojnZv6xM40anx4CkoDP+fZsw/ZerEMsW/pyzsRbElpsL/DBVW7q3ExxwusdNXI3lXpuhEZkzs8p5Eg== @@ -1524,6 +1597,14 @@ color-support@^1.1.2, color-support@^1.1.3: resolved "https://registry.yarnpkg.com/color-support/-/color-support-1.1.3.tgz#93834379a1cc9a0c61f82f52f0d04322251bd5a2" integrity sha512-qiBjkpbMLO/HL68y+lh4q0/O1MZFj2RX6X/KmMa3+gJD3z+WwI1ZzDHysvqHGS3mP6mznPckpXmw1nI9cJjyRg== +color@^3.1.3: + version "3.2.1" + resolved "https://registry.yarnpkg.com/color/-/color-3.2.1.tgz#3544dc198caf4490c3ecc9a790b54fe9ff45e164" + integrity sha512-aBl7dZI9ENN6fUGC7mWpMTPNHmWUSNan9tuWN6ahh5ZLNk9baLJOnSMlrQkHcrfFgz2/RigjUVAjdx36VcemKA== + dependencies: + color-convert "^1.9.3" + color-string "^1.6.0" + color@^4.2.3: version "4.2.3" resolved "https://registry.yarnpkg.com/color/-/color-4.2.3.tgz#d781ecb5e57224ee43ea9627560107c0e0c6463a" @@ -1537,6 +1618,14 @@ colors@^1.4.0: resolved "https://registry.yarnpkg.com/colors/-/colors-1.4.0.tgz#c50491479d4c1bdaed2c9ced32cf7c7dc2360f78" integrity sha512-a+UqTh4kgZg/SlGvfbzDHpgRu7AAQOmmqRHJnxhRZICKFUT91brVhNNt58CMWU9PsBbv3PDCZUHbVxuDiH2mtA== +colorspace@1.1.x: + version "1.1.4" + resolved "https://registry.yarnpkg.com/colorspace/-/colorspace-1.1.4.tgz#8d442d1186152f60453bf8070cd66eb364e59243" + integrity sha512-BgvKJiuVu1igBUF2kEjRCZXol6wiiGbY5ipL/oVPwm0BL9sIpMIzM8IK7vwuxIIzOXMV3Ey5w+vxhm0rR/TN8w== + dependencies: + color "^3.1.3" + text-hex "1.0.x" + combined-stream@^1.0.8: version "1.0.8" resolved "https://registry.yarnpkg.com/combined-stream/-/combined-stream-1.0.8.tgz#c3d45a8b34fd730631a110a8a2520682b31d5a7f" @@ -1680,6 +1769,11 @@ crypt@0.0.2: resolved "https://registry.yarnpkg.com/crypt/-/crypt-0.0.2.tgz#88d7ff7ec0dfb86f713dc87bbb42d044d3e6c41b" integrity sha512-mCxBlsHFYh9C+HVpiEacem8FEBnMXgU9gy4zmNC+SXAZNB/1idgp/aulFJ4FgCi7GPEVbfyng092GqL2k2rmow== +dayjs@^1.11.7: + version "1.11.10" + resolved "https://registry.yarnpkg.com/dayjs/-/dayjs-1.11.10.tgz#68acea85317a6e164457d6d6947564029a6a16a0" + integrity sha512-vjAczensTgRcqDERK0SR2XMwsF/tSvnvlv6VcF2GIhg6Sx4yOIt/irsr1RDJsKiIyBzJDpCoXiWWq28MqH2cnQ== + debug@2.6.9: version "2.6.9" resolved "https://registry.yarnpkg.com/debug/-/debug-2.6.9.tgz#5d128515df134ff327e90a4c93f4e077a536341f" @@ -1835,6 +1929,11 @@ emoji-regex@^8.0.0: resolved "https://registry.yarnpkg.com/emoji-regex/-/emoji-regex-8.0.0.tgz#e818fd69ce5ccfcb404594f842963bf53164cc37" integrity sha512-MSjYzcWNOA0ewAHpz0MxpYFvwg6yjy1NG3xteoqz644VCo/RPgnr1/GGt+ic3iJTzQ8Eu3TdM14SawnVUmGE6A== +enabled@2.0.x: + version "2.0.0" + resolved "https://registry.yarnpkg.com/enabled/-/enabled-2.0.0.tgz#f9dd92ec2d6f4bbc0d5d1e64e21d61cd4665e7c2" + integrity sha512-AKrN98kuwOzMIdAizXGI86UFBoo26CL21UM763y1h/GMSJ4/OHU9k2YlsmBpyScFo/wbLzWQJBMCW4+IO3/+OQ== + encode32@^1.1.0: version "1.1.0" resolved "https://registry.yarnpkg.com/encode32/-/encode32-1.1.0.tgz#0c54b45fb314ad5502e3c230cb95acdc5e5cd1dd" @@ -2254,6 +2353,11 @@ fd-slicer@~1.1.0: dependencies: pend "~1.2.0" +fecha@^4.2.0: + version "4.2.3" + resolved "https://registry.yarnpkg.com/fecha/-/fecha-4.2.3.tgz#4d9ccdbc61e8629b259fdca67e65891448d569fd" + integrity sha512-OP2IUU6HeYKJi3i0z4A19kHMQoLVs4Hc+DPqqxI2h/DPZHTm/vjsfC6P0b4jCMy14XizLBqvndQ+UilD7707Jw== + file-entry-cache@^6.0.1: version "6.0.1" resolved "https://registry.yarnpkg.com/file-entry-cache/-/file-entry-cache-6.0.1.tgz#211b2dd9659cb0394b073e7323ac3c933d522027" @@ -2339,6 +2443,11 @@ flow-remove-types@^2.217.1: pirates "^3.0.2" vlq "^0.2.1" +fn.name@1.x.x: + version "1.1.0" + resolved "https://registry.yarnpkg.com/fn.name/-/fn.name-1.1.0.tgz#26cad8017967aea8731bc42961d04a3d5988accc" + integrity sha512-GRnmB5gPyJpAhTQdSZTSp9uaPSvl09KoYcMQtsB9rQoOmzs9dH6ffeccH+Z+cv6P68Hu5bC6JjRh4Ah/mHSNRw== + follow-redirects@^1.14.8, follow-redirects@^1.14.9: version "1.15.2" resolved "https://registry.yarnpkg.com/follow-redirects/-/follow-redirects-1.15.2.tgz#b460864144ba63f2681096f274c4e57026da2c13" @@ -3344,6 +3453,11 @@ keyv@^4.5.3: dependencies: json-buffer "3.0.1" +kuler@^2.0.0: + version "2.0.0" + resolved "https://registry.yarnpkg.com/kuler/-/kuler-2.0.0.tgz#e2c570a3800388fb44407e851531c1d670b061b3" + integrity sha512-Xq9nH7KlWZmXAtodXDDRE7vs6DU1gTU8zYDHDiWLSip45Egwq3plLHzPn27NgvzL2r1LMPC1vdqh98sQxtqj4A== + ky@^0.33.1: version "0.33.3" resolved "https://registry.yarnpkg.com/ky/-/ky-0.33.3.tgz#bf1ad322a3f2c3428c13cfa4b3af95e6c4a2f543" @@ -3500,11 +3614,28 @@ log-symbols@^5.1.0: chalk "^5.0.0" is-unicode-supported "^1.1.0" +logform@^2.3.2, logform@^2.4.0: + version "2.6.0" + resolved "https://registry.yarnpkg.com/logform/-/logform-2.6.0.tgz#8c82a983f05d6eaeb2d75e3decae7a768b2bf9b5" + integrity sha512-1ulHeNPp6k/LD8H91o7VYFBng5i1BDE7HoKxVbZiGFidS1Rj65qcywLxX+pVfAPoQJEjRdvKcusKwOupHCVOVQ== + dependencies: + "@colors/colors" "1.6.0" + "@types/triple-beam" "^1.3.2" + fecha "^4.2.0" + ms "^2.1.1" + safe-stable-stringify "^2.3.1" + triple-beam "^1.3.0" + long@^4.0.0: version "4.0.0" resolved "https://registry.yarnpkg.com/long/-/long-4.0.0.tgz#9a7b71cfb7d361a194ea555241c92f7468d5bf28" integrity sha512-XsP+KhQif4bjX1kbuSiySJFNAehNxgLb6hPRGJ9QsUr8ajHkuXGdrHmFUTUUXhDwVX2R5bY4JNZEwbUiMhV+MA== +long@^5.0.0: + version "5.2.3" + resolved "https://registry.yarnpkg.com/long/-/long-5.2.3.tgz#a3ba97f3877cf1d778eccbcb048525ebb77499e1" + integrity sha512-lcHwpNoggQTObv5apGNCTdJrO69eHOZMi4BNC+rTLER8iHAqGrUVeLh/irVIM7zTw2bOXA8T6uNPeujwOLg/2Q== + loose-envify@^1.4.0: version "1.4.0" resolved "https://registry.yarnpkg.com/loose-envify/-/loose-envify-1.4.0.tgz#71ee51fa7be4caec1a63839f7e682d8132d30caf" @@ -3524,6 +3655,11 @@ lru-cache@^6.0.0: dependencies: yallist "^4.0.0" +lru-cache@^9.1.2: + version "9.1.2" + resolved "https://registry.yarnpkg.com/lru-cache/-/lru-cache-9.1.2.tgz#255fdbc14b75589d6d0e73644ca167a8db506835" + integrity sha512-ERJq3FOzJTxBbFjZ7iDs+NiK4VI9Wz+RdrrAB8dio1oV+YvdPzUEE4QNiT2VD51DkIbCYRUUzCRkssXCHqSnKQ== + make-dir@^3.1.0: version "3.1.0" resolved "https://registry.yarnpkg.com/make-dir/-/make-dir-3.1.0.tgz#415e967046b3a7f1d185277d84aa58203726a13f" @@ -4042,6 +4178,13 @@ once@^1.3.0, once@^1.3.1, once@^1.4.0: dependencies: wrappy "1" +one-time@^1.0.0: + version "1.0.0" + resolved "https://registry.yarnpkg.com/one-time/-/one-time-1.0.0.tgz#e06bc174aed214ed58edede573b433bbf827cb45" + integrity sha512-5DXOiRKwuSEcQ/l0kGCF6Q3jcADFv5tSmRaJck/OqkVFcOzutB134KRSfF0xDrL39MNnqxbHBbUUcjZIhTgb2g== + dependencies: + fn.name "1.x.x" + onetime@^5.1.0: version "5.1.2" resolved "https://registry.yarnpkg.com/onetime/-/onetime-5.1.2.tgz#d0e96ebb56b07476df1dd9c4806e5237985ca45e" @@ -4334,6 +4477,24 @@ prop-types@^15.8.1: object-assign "^4.1.1" react-is "^16.13.1" +protobufjs@7.2.4: + version "7.2.4" + resolved "https://registry.yarnpkg.com/protobufjs/-/protobufjs-7.2.4.tgz#3fc1ec0cdc89dd91aef9ba6037ba07408485c3ae" + integrity sha512-AT+RJgD2sH8phPmCf7OUZR8xGdcJRga4+1cOaXJ64hvcSkVhNcRHOwIxUatPH15+nj59WAGTDv3LSGZPEQbJaQ== + dependencies: + "@protobufjs/aspromise" "^1.1.2" + "@protobufjs/base64" "^1.1.2" + "@protobufjs/codegen" "^2.0.4" + "@protobufjs/eventemitter" "^1.1.0" + "@protobufjs/fetch" "^1.1.0" + "@protobufjs/float" "^1.0.2" + "@protobufjs/inquire" "^1.1.0" + "@protobufjs/path" "^1.1.2" + "@protobufjs/pool" "^1.1.0" + "@protobufjs/utf8" "^1.1.0" + "@types/node" ">=13.7.0" + long "^5.0.0" + protobufjs@^6.8.8: version "6.11.4" resolved "https://registry.yarnpkg.com/protobufjs/-/protobufjs-6.11.4.tgz#29a412c38bf70d89e537b6d02d904a6f448173aa" @@ -4353,6 +4514,24 @@ protobufjs@^6.8.8: "@types/node" ">=13.7.0" long "^4.0.0" +protobufjs@^7.0.0, protobufjs@^7.2.4: + version "7.2.5" + resolved "https://registry.yarnpkg.com/protobufjs/-/protobufjs-7.2.5.tgz#45d5c57387a6d29a17aab6846dcc283f9b8e7f2d" + integrity sha512-gGXRSXvxQ7UiPgfw8gevrfRWcTlSbOFg+p/N+JVJEK5VhueL2miT6qTymqAmjr1Q5WbOCyJbyrk6JfWKwlFn6A== + dependencies: + "@protobufjs/aspromise" "^1.1.2" + "@protobufjs/base64" "^1.1.2" + "@protobufjs/codegen" "^2.0.4" + "@protobufjs/eventemitter" "^1.1.0" + "@protobufjs/fetch" "^1.1.0" + "@protobufjs/float" "^1.0.2" + "@protobufjs/inquire" "^1.1.0" + "@protobufjs/path" "^1.1.2" + "@protobufjs/pool" "^1.1.0" + "@protobufjs/utf8" "^1.1.0" + "@types/node" ">=13.7.0" + long "^5.0.0" + proxy-addr@~2.0.7: version "2.0.7" resolved "https://registry.yarnpkg.com/proxy-addr/-/proxy-addr-2.0.7.tgz#f19fe69ceab311eeb94b42e70e8c2070f9ba1025" @@ -4605,6 +4784,11 @@ safe-regex-test@^1.0.0: get-intrinsic "^1.1.3" is-regex "^1.1.4" +safe-stable-stringify@^2.3.1: + version "2.4.3" + resolved "https://registry.yarnpkg.com/safe-stable-stringify/-/safe-stable-stringify-2.4.3.tgz#138c84b6f6edb3db5f8ef3ef7115b8f55ccbf886" + integrity sha512-e2bDA2WJT0wxseVd4lsDP4+3ONX6HpMXQa1ZhFQ7SU+GjvORCmShbCMltrtIDfkYhVHrOcPtj+KhmDBdPdZD1g== + "safer-buffer@>= 2.1.2 < 3", "safer-buffer@>= 2.1.2 < 3.0.0": version "2.1.2" resolved "https://registry.yarnpkg.com/safer-buffer/-/safer-buffer-2.1.2.tgz#44fa161b0187b9549dd84bb91802f9bd8385cd6a" @@ -4835,6 +5019,11 @@ ssri@^8.0.0, ssri@^8.0.1: dependencies: minipass "^3.1.1" +stack-trace@0.0.x: + version "0.0.10" + resolved "https://registry.yarnpkg.com/stack-trace/-/stack-trace-0.0.10.tgz#547c70b347e8d32b4e108ea1a2a159e5fdde19c0" + integrity sha512-KGzahc7puUKkzyMt+IqAep+TVNbKP+k2Lmwhub39m1AsTSkaDutx56aDCo+HLDzf/D26BIHTJWNiTG1KAJiQCg== + statuses@2.0.1: version "2.0.1" resolved "https://registry.yarnpkg.com/statuses/-/statuses-2.0.1.tgz#55cb000ccf1d48728bd23c685a063998cf1a1b63" @@ -5078,6 +5267,11 @@ tar@^6.0.2, tar@^6.1.11, tar@^6.1.2: mkdirp "^1.0.3" yallist "^4.0.0" +text-hex@1.0.x: + version "1.0.0" + resolved "https://registry.yarnpkg.com/text-hex/-/text-hex-1.0.0.tgz#69dc9c1b17446ee79a92bf5b884bb4b9127506f5" + integrity sha512-uuVGNWzgJ4yhRaNSiubPY7OjISw4sw4E5Uv0wbjp+OzcbmVU/rsT8ujgcXJhn9ypzsgr5vlzpPqP+MBBKcGvbg== + text-table@^0.2.0: version "0.2.0" resolved "https://registry.yarnpkg.com/text-table/-/text-table-0.2.0.tgz#7f5ee823ae805207c00af2df4a84ec3fcfa570b4" @@ -5107,6 +5301,11 @@ tr46@~0.0.3: resolved "https://registry.yarnpkg.com/tr46/-/tr46-0.0.3.tgz#8184fd347dac9cdc185992f3a6622e14b9d9ab6a" integrity sha512-N3WMsuqV66lT30CrXNbEjx4GEwlow3v6rr4mCcv6prnfwhS01rkgyFdjPNBYd9br7LpXV1+Emh01fHnq2Gdgrw== +triple-beam@^1.3.0: + version "1.4.1" + resolved "https://registry.yarnpkg.com/triple-beam/-/triple-beam-1.4.1.tgz#6fde70271dc6e5d73ca0c3b24e2d92afb7441984" + integrity sha512-aZbgViZrg1QNcG+LULa7nhZpJTZSLm/mXnHXnbAbjmN5aSa0y7V+wvv6+4WaBtpISJzThKy+PIPxc1Nq1EJ9mg== + tslib@^2.2.0, tslib@^2.4.0: version "2.6.1" resolved "https://registry.yarnpkg.com/tslib/-/tslib-2.6.1.tgz#fd8c9a0ff42590b25703c0acb3de3d3f4ede0410" @@ -5448,6 +5647,32 @@ wide-align@^1.1.2, wide-align@^1.1.5: dependencies: string-width "^1.0.2 || 2 || 3 || 4" +winston-transport@^4.5.0: + version "4.6.0" + resolved "https://registry.yarnpkg.com/winston-transport/-/winston-transport-4.6.0.tgz#f1c1a665ad1b366df72199e27892721832a19e1b" + integrity sha512-wbBA9PbPAHxKiygo7ub7BYRiKxms0tpfU2ljtWzb3SjRjv5yl6Ozuy/TkXf00HTAt+Uylo3gSkNwzc4ME0wiIg== + dependencies: + logform "^2.3.2" + readable-stream "^3.6.0" + triple-beam "^1.3.0" + +winston@^3.9.0: + version "3.11.0" + resolved "https://registry.yarnpkg.com/winston/-/winston-3.11.0.tgz#2d50b0a695a2758bb1c95279f0a88e858163ed91" + integrity sha512-L3yR6/MzZAOl0DsysUXHVjOwv8mKZ71TrA/41EIduGpOOV5LQVodqN+QdQ6BS6PJ/RdIshZhq84P/fStEZkk7g== + dependencies: + "@colors/colors" "^1.6.0" + "@dabh/diagnostics" "^2.0.2" + async "^3.2.3" + is-stream "^2.0.0" + logform "^2.4.0" + one-time "^1.0.0" + readable-stream "^3.4.0" + safe-stable-stringify "^2.3.1" + stack-trace "0.0.x" + triple-beam "^1.3.0" + winston-transport "^4.5.0" + wordwrapjs@^4.0.0: version "4.0.1" resolved "https://registry.yarnpkg.com/wordwrapjs/-/wordwrapjs-4.0.1.tgz#d9790bccfb110a0fc7836b5ebce0937b37a8b98f"