Give your AI app
a searchable memory

A vector database does not store the text itself. It stores the semantic coordinates an AI embedding model produces from your text (or images), which is why it can search by meaning instead of by keyword. Any time you want AI to answer from your own data (a RAG knowledge base), semantic search, similarity recommendations or an AI support bot, you need a vector database. Qdrant is one of the most popular open-source options, and n8n, Dify and LangChain all integrate with it out of the box.

Qdrant is free and open source, but self-hosting means renting a VPS (from about NT$160 a month), running Docker yourself, setting up the API key and HTTPS, watching memory usage and keeping up with releases. RoamerHost starts at NT$299/mo, provisions about 60 seconds after payment, and comes with its own API key, HTTPS endpoint and resource isolation. Take the managed route if you want it handled; self-host if you know DevOps and want full control. Either way you can snapshot your data out at any time, so you are never locked in.

Qdrant Cloud is usage-billed: the smallest paid cluster runs about US$114 a month (roughly NT$3,600) and needs a USD credit card. The comparable Pro plan on RoamerHost is NT$899 a month, at a flat price, billed in TWD, with Taiwan-based support in English or Chinese. Qdrant Cloud wins on multi-region deployment and much larger clusters; if your application stays under a few million vectors, the managed plan is better value.

Estimating with a common 768-dimension embedding (OpenAI text-embedding-3-small, multilingual E5 and the like): Lite holds roughly 100,000 vectors, Standard about 250,000 and Pro about 600,000. A single document usually becomes dozens to hundreds of passage vectors, so 100,000 vectors covers several hundred to a couple of thousand documents — plenty for a typical company knowledge base. Real capacity varies with dimension count and payload size.

n8n ships a Qdrant Vector Store node, and Dify knowledge bases can point at an external Qdrant. Once your instance is live you get a dedicated endpoint (https://qdr-xxxx.roamerhost.com) and an API key; fill those two values into n8n or Dify and you are done. If your n8n also runs on RoamerHost, the two talk over the same internal network with very low latency.

Each instance is its own Docker container with its own data directory and isolated memory and CPU quotas. All connections go over HTTPS and API key authentication is always enforced — a request without a key is rejected. The key is only ever shown in your dashboard after you log in.

Yes. Qdrant has snapshots built in, so you can export an entire collection at any time and take it to your own server or to Qdrant Cloud. There is no proprietary format. For a plan upgrade, contact support and we will raise your resource quota with your data intact.

All plans are monthly with no contract: Lite NT$299/mo (0.25 vCPU / 4 GB), Standard NT$499/mo (0.5 vCPU / 10 GB), Pro NT$899/mo (1 vCPU / 20 GB). They differ in compute and storage; every plan includes your own API key and HTTPS endpoint, unlimited collections, live 60 seconds after payment, with no setup fee.

In your own Qdrant instance: each instance is an isolated Docker container on its own network, never shared with other customers, with HTTPS on every connection. Every request must carry your API key. If your n8n or Dify also runs on RoamerHost, connections stay in the same data center with very low latency.

Yes. Monthly billing with no contract: cancel in one click from the dashboard and billing stops at the end of the current term; fees already paid for that term are not refunded. After expiry the instance is paused and your data is kept for 3 months, so renewing within that window restores everything as it was; after 3 months it is deleted. To move out, export your collections with a snapshot first.

Yes. At checkout, choose "Company invoice" under E-invoice and enter your company name and Taiwan Tax ID (統一編號) to receive a triplicate company invoice; individuals can choose "Personal receipt".

Yes, with n8n: its built-in Qdrant Vector Store node takes your endpoint and API key, splits documents, turns them into vectors, stores them in Qdrant and powers an AI answer bot — see "RAG with n8n + Qdrant" in the Support Center for the steps. Calling the API from your own code takes some programming background.

Yes. On the Free Trial page, pick Qdrant and leave your contact details — no credit card, nothing to install. Our team sets up a trial environment so you can try it before you decide.

Start with the dashboard: you can restart or rebuild the instance yourself, and a restart fixes most connection problems. For how-to questions, ask Roamer, the chat assistant in the bottom-right corner, or use Roamer AI support in your dashboard. For a human, email [email protected] or message our LINE official account @818kotud — we reply within 24 hours on business days, from Taiwan.