Name: LLM Wiki v2 (CROW)
Availability: InStock
Author: onepush

Deploy and Host self hosted LLM Wiki v2 (CROW) on Railway

This template deploys LLM Wiki v2 (crow-hq/llm-wiki-v2, the CROW project) as one Railway service behind Caddy basic auth. It suits people who want a readable, self-hosted wiki they can keep in git.

About Hosting LLM Wiki v2 open-source software on Railway (self hosted LLM Wiki v2 template)

This template runs the upstream crow-hq/llm-wiki-v2 v1.0.0 image as one Railway service, with a Caddy basic-auth proxy in front because the upstream server has no login of its own.

Why Deploy LLM Wiki v2, the self-hosted LLM-maintained Markdown wiki on Railway (Railway Free Trial)

Obsidian is a strong local Markdown app, but it keeps notes on one device and leaves the filing to you. Railway billing is usage-based, so watch the cost after you deploy.

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying LLM Wiki v2 on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.

Railway vs Other Hosting Providers and VPS for LLM Wiki v2 self hosting

Provider	What You Get with Railway	What You Get with the Other Provider
DigitalOcean	A container with Caddy basic auth, a /data volume and a health check, set up from the template.	A virtual machine you configure yourself, including the proxy, authentication and backups.
AWS	The same one-container template with a generated WIKI_PASSWORD and no instance sizing.	Broad control over compute and storage, with the instance, disk and access controls left to you.
Hetzner	The upstream v1.0.0 image behind Caddy basic auth, with the wiki kept on a volume.	Low-cost servers where you install the package or Docker and maintain the host.

Common Use Cases for hosted LLM Wiki v2

Filing a stack of documents: Drop text, Markdown, PDF, Word, PowerPoint, Excel or OpenDocument files onto the page and each is filed as a plain Markdown note in the right folder, merged with existing knowledge.
Asking questions over your notes: Type a question and the answer cites the notes it used, so you can click through and check the source.
Browsing the wiki as a map: A live view of folders, notes and links shows how the knowledge is organised.
Giving an MCP client a memory: Turn on Memory from the page and paste the connect line, so Claude Code, Codex or another MCP client can recall from the wiki and file new notes back into it.
Choosing how notes get filed: Run CROW mode, where a small typed classifier routes notes into folders, or classic mode, where the model decides placement itself.

Dependencies for LLM Wiki v2 Docker hosted on Railway

This template runs one container holding the upstream LLM Wiki v2 server behind a Caddy basic-auth proxy. The wiki is plain Markdown on a volume, so no separate database is needed.

Deployment Dependencies for Managed LLM Wiki v2 Service (one container with Caddy basic auth and SQLite-free Markdown on a volume)

The template deploys one service, llm-wiki. The volume holds the wiki Markdown, settings and usage log, so keep it across redeploys. The LLM key is optional at deploy time and can be added later on the Settings page.

Implementation Details for LLM Wiki v2 (unmodified llm-wiki-v2 v1.0.0 image behind a Caddy basic-auth proxy)

The image ghcr.io/onepush-dev/llm-wiki-v2:1.0.0-r2 sits on ghcr.io/onepush-dev/llm-wiki-v2-base:1.0.0, the upstream Dockerfile at tag v1.0.0 built unmodified, plus Caddy and a start script. The one template variable is WIKI_PASSWORD, generated per deploy; Caddy re-hashes it at each start.

How does LLM Wiki v2 compare against other tools

LLM Wiki v2 vs Obsidian (Obsidian Alternative)

Where Obsidian wins: Obsidian is a local Markdown knowledge app you use directly on your own machine, with a polished editing experience built around your personal notes.
Where LLM Wiki v2 differs: an LLM files the documents you drop in and keeps the wiki organised, and answers questions with citations to the notes it used.
Hosting model: this template runs the wiki as a password-protected service on Railway, with the Markdown kept on a volume, rather than as a desktop app.

LLM Wiki v2 vs DokuWiki (DokuWiki Alternative)

Where DokuWiki wins: DokuWiki is a classic, long-established self-hosted wiki with a mature editing model that people edit by hand.
Where LLM Wiki v2 differs: the notes are written and merged by an LLM from your sources, and questions are answered from them with citations.
Hosting model: LLM Wiki v2 is a single container here, with Caddy basic auth in front of the upstream server, which has no login of its own.

LLM Wiki v2 vs Notion (Notion Alternative)

Where Notion wins: Notion is a hosted workspace with collaboration features and no server for you to run.
Where LLM Wiki v2 differs: the wiki is plain Markdown you can read, edit and keep in git, and it runs on infrastructure you control.
Cost model: LLM Wiki v2 is free AGPL-3.0-or-later software; you pay for the hosting and for any model provider you connect.

LLM Wiki v2 with Docker vs Railway (Docker and Railway Comparison)

Docker on your own machine: you run the image yourself, set the variables in a file, and manage the volume and the public address.
Railway: the template creates the service, the volume at /data and the generated WIKI_PASSWORD for you, and the model key can be added later from the Settings page.
Shared behaviour: both run the same upstream v1.0.0 image behind the same basic-auth proxy, so the wiki works the same way in either place.

How to use LLM Wiki v2 on Railway?

Open the public URL for your llm-wiki service. Your browser will ask for basic auth, so sign in as admin with the WIKI_PASSWORD value from the llm-wiki Variables tab.
Open Settings on the page and paste a model key. An OpenRouter key also serves the default CROW classifier. Otherwise set OPENAI_API_KEY or GEMINI_API_KEY with OKF_LLM_PROVIDER, or switch to classic mode.
If you skip the key, the page still loads, but ingest or ask will return a message saying a key is needed.
Optionally, turn on MCP memory at /mcp from the page. Clients such as Claude Code or Codex must send the same basic-auth header.
Check that the service is healthy with the endpoint below. /health is left open for the health check and does not expose the wiki.

https://your-domain/health

How to self host LLM Wiki v2 on other VPS Services (LLM Wiki v2 self hosting guide)

Clone the Repository

Clone the upstream project from GitHub, then move into its folder. The repository is crow-hq/llm-wiki-v2, and the Docker files live at the top level.

git clone https://github.com/crow-hq/llm-wiki-v2
cd llm-wiki-v2

Install Dependencies

Use Python 3.11 or later. Install the package with pip or uv, which also fetches Python for you.

pip install llmwiki2
# or
uv tool install llmwiki2

Configure Environment Variables

Choose a provider and paste its key on the settings page the first time you run the app. For Docker, copy the example file and set your key there.

cp .env.example .env
# set OKF_LLM_PROVIDER and its key, for example OPENROUTER_API_KEY

Start the LLM Wiki v2 Application

Run llmwiki2, which opens a browser so you can pick a provider and paste a key. With Docker, run docker compose up -d --build and open http://localhost:8000.

Pricing and License of LLM Wiki v2 (LLM Wiki v2 pricing)

LLM Wiki v2 (CROW) is open-source software released under the AGPL-3.0-or-later license.

LLM Wiki v2 hosted vs self hosted comparison (Pricing, features, costs, and more)

LLM Wiki v2 is free AGPL-3.0-or-later software, and this template describes no hosted plan. Crow desktop is a separate upstream app.

Monthly cost of self hosting LLM Wiki v2 on Railway

Railway billing is usage-based and depends on RAM, CPU, volume size and services. This template runs one service with one volume, so cost grows with wiki size and query volume. Model provider charges are separate.

System Requirements for Hosting LLM Wiki v2 on a VPS

Running the upstream server on a VPS needs Python 3.11 or later. Put an authenticating proxy in front if you are not using this template's Caddy basic auth, because the upstream server has no login of its own.

Frequently Asked Questions (FAQs)

How do I sign in to my LLM Wiki v2 deployment?

Open the public URL and sign in as admin with the WIKI_PASSWORD value from the service Variables tab. The wiki itself has no login of its own, so the template adds Caddy basic auth in front of it.

Do I need a model key at deploy time?

No. You can add a key later on the Settings page, or set OPENROUTER_API_KEY, OPENAI_API_KEY or GEMINI_API_KEY as a service variable. Without a key the page still loads, but ingest and ask will tell you a key is needed.

What is CROW mode?

CROW is the default mode, where a small classifier decides which folder each note goes into. If you do not have a classifier key, switch to classic mode, where the language model makes those decisions itself.

Is this the same as the LLM Wiki template already on Railway?

No. That is a different project. This template deploys crow-hq/llm-wiki-v2 (CROW), which is not that other project.

Can MCP clients connect to my wiki?

Yes. Turn Memory on from the page, then send the same basic-auth header from your MCP client. The /mcp endpoint is not left open.

What should I back up?

Back up the /data volume, especially /data/wiki (your Markdown notes) and /data/config.json (your saved settings).