The Docker stack can include MongoDB, MeiliSearch, RAG API and vector storage. Feature-rich self-hosting is the reason LibreChat itself suggests 2 GB for smoother operation.
Users enabling RAG, search and many integrations.LibreChatのハードウェア要件：CPU、RAM、Docker、ローカルAI
LibreChatの1 GiB RAM、1 vCPU、Docker、MongoDB、RAG、ローカルモデルの要件を確認し、適したZimaOSハードウェアを選びましょう。
LibreChat requirements at a glance
LibreChat publishes a clear application baseline: 1 GiB RAM and 1 vCPU, with 2 GB RAM suggested for smoother operation when everything is enabled. Those figures describe LibreChat and its deployment stack, not the hardware required by a local LLM. If LibreChat connects to Ollama, vLLM or another local model server, that inference hardware must be sized separately.
- CPU
- Official minimum: 1 vCPU for LibreChat.
- RAM
- Official minimum: 1 GiB. LibreChat suggests 2 GB when everything is enabled for smoother operation.
- Docker stack
- The recommended Docker Compose setup bundles dependencies such as MongoDB, MeiliSearch, RAG API and vector-database components.
- GPU
- LibreChat itself does not require a GPU. A GPU is only needed when a connected local inference backend independently requires it.
- Storage
- Persistent chat/application data, MongoDB, search indexes, RAG/vector data and uploaded files need more storage than the frontend alone.
- Best Zima starting point
- ZimaBoard 2 832 is already well above the 1 GiB/1-vCPU minimum. Creator Pack is only justified if the same server also runs a verified local GPU model backend.
From official requirements to the right setup
LibreChat sizing starts by separating the chat platform from the model provider.
-
Official requirements
Meet the official LibreChat floor first: at least 1 vCPU and 1 GiB RAM; 2 GB is the safer application baseline for more features.
-
Confirm your needs
Decide whether models are remote APIs or local. Remote OpenAI-compatible/provider APIs keep inference off the ZimaOS host; local Ollama/vLLM moves the much larger model-memory/GPU requirement onto the server.
-
Leave room to grow
Account for bundled services. Docker Compose can include MongoDB, MeiliSearch, RAG API and vector storage, so a feature-rich deployment uses more RAM/storage than the UI alone.
-
Run it on ZimaOS
Install LibreChat from ZimaOS, enable the real providers/RAG/search features, then monitor the application stack separately from any local model server before upgrading hardware.
Check every playback client
- Remote APIs versus local model backend
- 1 GiB minimum / 2 GB smoother LibreChat memory
- MongoDB
- MeiliSearch
- RAG API / vector database
- Uploaded documents and search indexes
- Number of users/conversations
- Local Ollama/vLLM model memory if co-hosted
Official minimum requirements
LibreChat's current remote-hosting documentation publishes explicit minimum requirements.
Use 1 vCPU + 1 GiB as the official LibreChat floor and 2 GB as the project's smoother all-features suggestion. Do not add local-LLM VRAM to these numbers unless the model backend is actually hosted on the same machine.
| Requirement | Official minimum | What this supports |
|---|---|---|
| CPU | 1 vCPU minimum | Official LibreChat deployment minimum. |
| RAM | 1 GiB minimum | Official LibreChat deployment minimum. |
| All-features RAM guidance | 2 GB | LibreChat suggests increasing RAM to 2 GB for smoother operation with everything enabled. |
| Docker prerequisites | Git + Docker | Docker Compose is the recommended installation method for most scenarios. |
| Default Web port | 3080 | Current Docker quick start opens LibreChat at localhost:3080. |
| GPU | Not required by LibreChat itself | Local model backends have independent GPU/RAM requirements. |
When to upgrade your hardware
Upgrade LibreChat for more bundled services/users or because you also host local AI models.
RAG/search/database services make 1 GiB too tight
Many users/conversations increase database and search work
A local Ollama/vLLM backend is moved onto the same Zima server
More users, conversation history and indexes increase MongoDB/MeiliSearch activity even when the LLM is remote.
Family/team deployments rather than one personal account.This is the largest possible jump because model weights, context and GPU VRAM become a separate workload. Size the model server independently.
Users converting LibreChat from cloud APIs to fully local AI.Plan hardware growth with confidence
Scale LibreChat by keeping UI/orchestration resources separate from model inference.
Use remote model APIs when the host is small
LibreChat can provide the full chat interface while inference happens elsewhere, keeping the ZimaOS application footprint close to its normal 1–2 GB class.
ZimaBoard 2 832 is already ample for this architecture.Allocate extra RAM for MongoDB/MeiliSearch/RAG features
The Docker Compose bundle is more than one frontend container. Search and vector/document services add memory and storage.
ZimaBoard 2 1664 gives comfortable multi-service headroom.Keep uploaded/RAG data on persistent SSD storage
Conversation history, indexes, uploaded files and vector data should survive container recreation and can grow over time.
Use SSD/NVMe or larger multi-drive storage for document-heavy deployments.Add GPU hardware only for a local inference backend
LibreChat does not consume CUDA to render the chat UI. Ollama/vLLM/other inference software is the component that justifies VRAM.
Creator Pack is conditional on the selected local model fitting the verified GPU.Can it run on ZimaOS?
LibreChat is currently available in the ZimaOS App Store under AI.
Install LibreChat from the ZimaOS App Store
Use the packaged AI chat platform and connect the model providers you actually need.
Open LibreChat in the ZimaOS App StoreTreat bundled databases/search as part of the host workload
LibreChat's local Docker architecture bundles MongoDB, MeiliSearch, RAG API and vector DB components.
Read LibreChat local installation architectureSize local Ollama separately
A local model endpoint can live on the same or another server; its RAM/VRAM requirements are not included in LibreChat's 1 GiB minimum.
Read LibreChat custom endpoint guidanceChoose Zima hardware for LibreChat
LibreChat itself is light compared with local model inference. Pick higher Zima tiers only for RAG/database/storage features or a separately justified local AI backend.
Will LibreChat call remote APIs, or also run local models on the same Zima server?
Remote APIs / separate model server
LibreChat's own stack is modest.
- Normal LibreChat deploymentZimaBoard 2 832
- More RAG/search/users and containersZimaBoard 2 1664
- Document-heavy multi-drive AI workspaceZimaCube 2 Standard
Same-machine local GPU inference
Creator Pack becomes relevant only if its verified GPU VRAM fits the selected local model.
- Conditional local-GPU AI stackZimaCube 2 Creator Pack
The official 1 GiB/1-vCPU floor covers LibreChat, not a co-hosted model. User count, RAG/search services, document storage and local inference can change total server requirements.
| Zima hardware | Best for | Example workload | Core configuration | Recommended boundary | Next step |
|---|---|---|---|---|---|
| ZimaBoard 2 832 | LibreChat with remote/cloud model providers. | LibreChat, MongoDB/search and normal personal use. |
|
RAG-heavy or many-user deployments can outgrow the minimum even though remote inference stays off-host. | Get Now |
| ZimaBoard 2 1664 | Feature-rich LibreChat with more local services. | RAG/search, more conversations/users and companion AI tools. |
|
Local LLM model memory is not included in this 16 GB system-RAM budget. | Get Now |
| ZimaCube 2 Standard | A document-heavy LibreChat/RAG workspace with multi-drive storage. | Uploaded knowledge, indexes, backups and broader AI/self-hosted storage. |
|
8 GB is still system RAM; choose this tier primarily for integrated storage rather than application RAM. | Get Now |
| ZimaCube 2 Pro | A larger LibreChat/RAG and self-hosted AI control-plane server. | More databases, RAG services, storage and containers. |
|
16 GB does not guarantee a co-hosted local model. Add a supported GPU or use a separate inference server. | Get Now |
| ZimaCube 2 Creator Pack | LibreChat plus independently justified local NVIDIA inference. | Chat UI, RAG/storage and a compatible Ollama/vLLM-style local model backend. |
|
Extreme overkill for LibreChat alone. Verify exact GPU VRAM and model fit before promising local-model capability. | Get Now |
What the Press Says
Highlights from trusted reviewers worldwide.
“ZimaCube 2: Not just another NAS, tested with 25TB storage, local AI agents, 4K transcoding, and real homelab workflows.”Read full review
“The ZimaBoard 2 is a compact x86 server board that can be turned into a mini NAS, home server, media box, or self-hosting hub.”Read full review
“ZimaCube 2: A modern, high-performance NAS with plenty of room to grow—built for users who want more than basic storage.”Read full review
“Coverage focused on ZimaCube 2's open hardware foundation, no monthly fee, and self-hosting flexibility.”Read full review
Loved by the Community
Stories and reviews from people who build with Zima every day.
★★★★★
Zima Blade Little yet Powerful
Maybe I am not digital natives but I live with PCs since 12 years old in 1984 when IBM PC clone come to my home. Many years have passed and many operating system I've tried. For me Zima blade and CasaOS was a quantum leap for home PC enthusiast and server lab machine to make me stay curious and relevant for this era.
★★★★★
Very good!!
I use ZimaCube Pro as 5th Proxmox cluster node. It runs several VMs and containers, including a VM with GPU passthrough to run a self-hosted LLM. A specific LXC container runs a Samba server for NAS capabilities using four of six RAID 6 SATA HDDs with ZFS.
★★★★★
Great innovation for mini server!
It is very useful and makes a powerful mini server for many purposes, including university and college students in engineering and electronics. Thank you so much for making this server.
★★★★★
Avaliação ZimaBoard 2
Construí um servidor de uso pessoal. O desempenho está muito bom e funciona perfeitamente onde quer que eu esteja. A surpresa é não dependermos de grandes estruturas para termos nosso próprio servidor de dados. Como iniciante, estou gostando bastante do ZimaOS, pois ele é simples e eficiente.
Frequently asked questions
FAQ topics follow query fan-out around 1 GB vs 2 GB RAM, MongoDB/MeiliSearch, RAG overhead, GPU need and LibreChat versus the local model backend. Reddit discussions highlight confusion about the multi-service Docker stack.
How much RAM does LibreChat need?
LibreChat officially lists 1 GiB minimum and suggests 2 GB for smoother operation with everything enabled.
Why can LibreChat use more than 1 GB in Docker?
The Docker deployment is a stack, not only one UI process. MongoDB, MeiliSearch, RAG API and vector/database services can all contribute memory and storage.
Does LibreChat need a GPU?
No. LibreChat itself is a chat/orchestration platform. A GPU is needed only if a connected local model server independently requires GPU acceleration.
Does using Ollama inside LibreChat change the hardware requirement?
Yes if Ollama runs on the same machine. Then model weights, VRAM/RAM, context and concurrency become an additional workload that must be sized using Ollama/model requirements.
Can LibreChat run on ZimaBoard 2 832?
Yes. Its 4-core Intel N150 and 8 GB RAM are comfortably above LibreChat's 1-vCPU/1-GiB application minimum and leave room for normal bundled services.
Why does LibreChat use MongoDB and MeiliSearch?
MongoDB stores application/chat data and MeiliSearch supports search. Reddit comparisons with Open WebUI frequently focus on these extra services because they make LibreChat more operationally complex than a single-container frontend.
How much storage should LibreChat have for RAG?
There is no universal fixed number. Uploaded documents, embeddings/vector data, search indexes and conversation history can outgrow the base application quickly; size persistent SSD storage from the document corpus.
When is ZimaCube 2 Creator Pack justified for LibreChat?
Only when the same server also runs a compatible local GPU model backend. Remote API users gain no meaningful LibreChat benefit from a dedicated RTX GPU.
What sources and further reading informed this LibreChat hardware guide?
LibreChat's current remote deployment guide is the authority for the 1 GiB/1-vCPU minimum and 2 GB smoother-use note. Docker/local docs establish MongoDB, MeiliSearch, RAG and vector services. Reddit was used for query fan-out around LibreChat's heavier multi-service setup versus simpler UIs. ZimaOS confirms the current AI app.
