The universal UI and tool suite for managing vector databases at scale.
|
|
|
Docs
|
Hosted Instance
Quick! Can you tell me exactly what information is embedded in your Pinecone or Chroma vector database? I bet you can't. While those teams are focusing on building the underlying architecture we made it easier for you to manage vector data without the headaches and API calls.
We call it VectorAdmin and we want to be the best universal GUI for vector database management.
VectorAdmin aims to be a full-stack application that gives you total control over your otherwise unwieldy vector data that you are embedding via an API or using tools like LangChain, which don't show you want you just saved into your database.
VectorAdmin is a full capable multi-user product that you can run locally via Docker as well as host remotely and manage multiple vector databases at once.
VectorAdmin is more than a single tool. VectorAdmin is a suite of tools that make interacting with and understanding vectorized text easy without compromise for the controls you would expect from a traditional database management system.
Some cool features of VectorAdmin
- Multi-user instance support and oversight
- Atomically view, update, and delete singular text chunks of embeddings.
- Copy entire documents or even whole namespaces and embeddings without paying to re-embed.
- Upload & embed new documents directly into the vector database.
- Migrate an entire existing vector database to another type or instance. still in progress
- Manage multiple concurrent vector databases at once.
- Permission data and access to data
- 100% Cloud deployment ready.
- Automated regression testing that run as namespaces or collections are updated with new documents to ensure response quality. still in progress
- Full API, Javascript, and Python standalone client and LangChain integration. still in progress
- Extremely efficient cost-saving measures for managing very large documents. You'll never pay to embed a massive document or transcript more than once.
This monorepo consists of three main sections:
document-processor
: Flask app to digest, parse, and embed documents easily.frontend
: A viteJS + React frontend that you can run to easily create and manage all your content.backend
: A nodeJS + express server to handle all the interactions and do all the vectorDB management.workers
: An InngestJS instance to handle long-running processes background tasks for snappy performance.docker
: Run this entire arch in a single command as a docker instance recommended.
yarn
andnode
on your machinepython
3.9+ for running scripts indocument-processor/
.- access to an OpenAI API key if planning to update embeddings or upload new documents.
- a Pinecone.io free account or a running ChromaDB instance.
Get up and running in minutes with Docker
The below instructions will not work on Windows.
yarn dev:setup
from the project root directory.cd document-processor && python3.9 -m virtual-env v-env && source v-env/bin/activate && pip install -r requirements.txt
In separate terminal windows from project root:
yarn prisma:setup
to create DB migration and client and then runyarn dev:server
yarn dev:frontend
yarn dev:workers
cd document-processor && flask run --host '0.0.0.0' --port 8888
On first boot and visiting of the homepage, you will be automatically redirected to create your primary admin account, organization, and database connection.
- create issue
- create PR with branch name format of
<issue number>-<short name>
- yee haw let's merge
VectorAdmin by Mintplex Labs Inc contains a telemetry feature that collects anonymous usage information.
We use this information to help us understand how VectorAdmin is used, to help us prioritize work on new features and bug fixes, and to help us improve VectorAdmin's performance and stability.
Set DISABLE_TELEMETRY
in your server or docker .env settings to "true" to opt out of telemetry.
DISABLE_TELEMETRY="true"
We will only track usage details that help us make product and roadmap decisions, specifically:
- Server is started or booted up.
- Version of your installation.
- Type of job when executed.
You can verify these claims by finding all locations Telemetry.sendTelemetry
is called. Additionally these events are written to the output log so you can also see the specific data which was sent - if enabled. No IP or other identifying information is collected. The Telemetry provider is PostHog - an open-source telemetry collection service.
- Build docker image from source
git clone [email protected]:Mintplex-Labs/vector-admin.git
cd vector-admin
cd docker/
cp .env.example .env
.- 編輯
.env
檔案並更新變數。請更新以下所有內容:
JWT_SECRET="some-random-string"
DATABASE_CONNECTION_STRING="postgresql:https://vectoradmin:[email protected]:5433/vdbms" # Valid PG Connection string.
INNGEST_SIGNING_KEY="some-random-string"
docker-compose up -d --build vector-admin
- 要存取完整的應用程序,請在瀏覽器中存取 https://localhost:3001 。
- 連接到向量資料庫
- Pinecone
連接您的組織後,您將需要輸入 Pinecone 配置和金鑰。連接後,您可能會
Sync Pinecone Data
在主頁上看到一個按鈕。這表示您的向量資料庫中存在可以提取的現有資料。如果進行同步,同步時間取決於您在 Pinecone 中嵌入的文檔數量。否則,您只需建立一個工作區並透過 UI 新增文件。 - Chroma running locally
本機執行的Chroma 當嘗試連線到也在同一台電腦上執行的 Chroma 實例時,請使用
https://host.docker.internal:[CHROMA_PORT]
作為要連線的 URL。連接後,您可能會Sync Chroma Data在組織主頁上看到一個按鈕。這表示您的向量資料庫中存在可以提取的現有資料。如果進行同步,同步時間取決於您在 Chroma 中嵌入的文件數量。否則,您只需建立一個工作區並透過 UI 新增文件。