RAGFlow hosting

Deep document understanding for RAG — tables, layouts and scanned pages.

Agents & RAGApache-2.0from $36.99/mo

In short

RAGFlow hosting at AIHostBox means a dedicated, single-tenant RAGFlow instance on our infrastructure: we install it, pin it to a reviewed release, put it behind TLS on your domain, back it up nightly and keep it patched. Plans start at $36.99 per month and the admin account, the data and any API keys are yours.

Layout parsingTable extractionOCRChunk inspectionCitations

RAGFlow is an Apache-2.0 retrieval-augmented generation engine from InfiniFlow. Its parsing layer, DeepDoc, reads documents the way a person does — detecting layout, tables, headings and scanned text — and chunks them with a template chosen per knowledge base, such as general, Q&A, table, paper, book, laws or presentation. Every chunk is visible and editable, and answers carry citations back to the passage used. That ingestion-first design is what separates it from builders like Dify or Flowise, which treat parsing as a solved step.

It is also one of the heavier stacks in the catalogue. Beside the RAGFlow server, the official compose runs MySQL for metadata, Elasticsearch as the default document engine, MinIO for uploaded files and Valkey for task queues. Parsing a few hundred pages of scanned PDF is a sustained CPU job that should not tie up a laptop, and the knowledge bases, agents and API keys it produces must stay reachable for the applications that query them. Recent releases add an agent canvas, GraphRAG and MCP support on top of retrieval.

RAGFlow hosting plans

ProUpstream minimum (16 GB)

16 GB RAM · 6 vCPU · NVMe

$36.99/mo

Order

MaxLarge document sets

24 GB RAM · 8 vCPU · NVMe

$52.99/mo

Order

Every plan is a dedicated single-tenant instance with NVMe storage, nightly off-box backups, TLS and managed patching. Quarterly, semi-annual and annual terms are discounted at checkout. Model tokens are not included — you bring your own key.

How we run RAGFlow

The RAGFlow server image with its bundled web front-end, plus MySQL for metadata, Elasticsearch as document and vector engine, MinIO for uploaded files and Valkey for the task queue, as in the project’s docker compose.

What we configure for you

Every default password (ELASTIC_PASSWORD, MYSQL_PASSWORD, MINIO_PASSWORD, REDIS_PASSWORD) replaced — upstream ships them all as infini_rag_flow
DOC_ENGINE left on elasticsearch, the upstream default
REGISTER_ENABLED switched off once your accounts exist, so strangers cannot sign up
vm.max_map_count raised to 262144, which Elasticsearch requires
Your chat and embedding provider keys, or an Ollama endpoint, entered in model settings

Which plan fits

Upstream’s prerequisite is 4 cores and 16 GB, so that is where our plans start. Pro (16 GB, 6 vCPU) handles a team’s knowledge bases of reports and manuals; Max (24 GB, 8 vCPU) gives Elasticsearch and the DeepDoc parsers room for large document sets and several people querying at once.

The honest limits

DeepDoc parsing and OCR run on CPU here, so a long scanned PDF can take minutes to finish. Embedding and chat use your own API keys or a separate runtime, since the image no longer bundles models.

What people run RAGFlow for

Financial and legal documents

Keep table structure intact through ingestion.

Scanned archives

OCR and index paper documents into a searchable corpus.

Cited answers for support teams

Answer from product manuals with a pointer to the exact passage used.

Retrieval API for your own app

Query knowledge bases through RAGFlow’s HTTP API or Python SDK instead of building ingestion yourself.

Licence and version policy

Item	Position
Upstream licence	Apache-2.0
Sold as managed hosting	Yes
Tenancy	Single tenant — one customer per instance
Patching	Critical advisories within 24 hours, high severity within 72

Our full position on licences, including the applications we decline to sell, is on the licensing page.

Reading on RAGFlow

ComparisonDify vs Flowise vs Langflow

Works well with RAGFlow

More Agents & RAG we host

Questions

RAGFlow hosting: your questions

Which LLM and embedding providers does RAGFlow support?

You add providers under model settings: OpenAI, Anthropic, Gemini, DeepSeek, Mistral and local servers such as Ollama, among many others. Chat, embedding, reranking and image-to-text models are chosen separately, so you can embed with one provider and answer with another, all using your own keys.

Does RAGFlow still ship built-in embedding models?

No. Since v0.22.0 only the slim image is published, without bundled embedding models. You point RAGFlow at an embedding API with your key or at a local embedding server; upstream’s compose file includes an optional TEI service for this, defaulting to Qwen3-Embedding-0.6B.

How much memory does RAGFlow really need?

Upstream lists 4 cores, 16 GB RAM and 50 GB disk as prerequisites, and our Pro plan meets that floor. Max (24 GB) is for large document sets or heavy parallel parsing. If you expect tens of thousands of pages, tell us before ordering so we can size the disk.

Can I use RAGFlow from my own code?

Yes. Knowledge bases, chat assistants and agents are all reachable through RAGFlow’s HTTP API and Python SDK, authenticated with API keys generated in the web UI. You can upload documents, trigger parsing, retrieve chunks and run conversations programmatically, which makes it a retrieval back end for another front-end.

Which parts of RAGFlow’s data matter for backup?

State lives in four places: MySQL holds users, knowledge-base settings and document records, MinIO holds the original files, Elasticsearch holds the searchable chunks and vectors, and Redis holds queued tasks. Nightly off-box backups cover the persistent volumes; MySQL and MinIO are the irreplaceable part, since the index can be rebuilt by re-parsing.

Run RAGFlow without running a server

A dedicated RAGFlow instance, deployed and maintained by us. Month to month, cancel whenever.

Browse the catalogue Talk to a human