Deploy and Host Fabric on Railway

Fabric is an open-source framework for augmenting work with AI through reusable prompts called patterns. This template deploys Fabric as a hosted REST API, so apps, automations, and agents can call Fabric patterns, chat completions, model discovery, sessions, contexts, and helper tools over HTTP.

About Hosting Fabric

This Railway template builds Fabric from source using a custom Dockerfile, starts the Fabric REST API, and exposes it behind a Railway public domain. It includes bundled Fabric patterns and strategies, Swagger API docs, /health health checks, and API-key protection through the X-API-Key header. FABRIC_API_KEY is preconfigured by the template as ${{ secret(32) }}, so Railway generates a strong API secret automatically. For persistent usage, attach or keep the Railway volume at /home/appuser/.config/fabric; this keeps config, sessions, contexts, patterns, and strategies across redeploys. Before first use, set DEFAULT_VENDOR, DEFAULT_MODEL, and one provider API key such as OPENROUTER_API_KEY, OPENAI_API_KEY, ANTHROPIC_API_KEY, or GEMINI_API_KEY.

Common Use Cases

Run Fabric patterns from apps, agents, scripts, or workflow automations.
Host a reusable AI prompt API for summarization, extraction, analysis, and text transformation.
Deploy one Fabric endpoint that can use OpenRouter, OpenAI, Anthropic, Gemini, and many other providers.

Dependencies for Fabric Hosting

Railway account and project.
FABRIC_API_KEY, generated automatically by the template with ${{ secret(32) }}.
One AI provider API key matching DEFAULT_VENDOR.
Required user-selected variables: DEFAULT_VENDOR, DEFAULT_MODEL, and the matching provider key.
Railway volume mounted at /home/appuser/.config/fabric.
Public Railway domain for external API access.

Provider Variable Options

DEFAULT_VENDOR should use one of these provider names. Set only the matching credential variables for the provider you choose.

OpenAI-compatible providers:

`DEFAULT_VENDOR`	Required variables	Optional variables
`Abacus`	`ABACUS_API_KEY`	`ABACUS_API_BASE_URL`
`AIML`	`AIML_API_KEY`	`AIML_API_BASE_URL`
`Cerebras`	`CEREBRAS_API_KEY`	`CEREBRAS_API_BASE_URL`
`DeepSeek`	`DEEPSEEK_API_KEY`	`DEEPSEEK_API_BASE_URL`
`GitHub`	`GITHUB_API_KEY`	`GITHUB_API_BASE_URL`
`GrokAI`	`GROKAI_API_KEY`	`GROKAI_API_BASE_URL`
`Groq`	`GROQ_API_KEY`	`GROQ_API_BASE_URL`
`Infermatic`	`INFERMATIC_API_KEY`	`INFERMATIC_API_BASE_URL`
`Langdock`	`LANGDOCK_API_KEY`	`LANGDOCK_API_BASE_URL`, `LANGDOCK_REGION`
`LiteLLM`	`LITELLM_API_KEY`	`LITELLM_API_BASE_URL`
`Mammouth`	`MAMMOUTH_API_KEY`	`MAMMOUTH_API_BASE_URL`
`MiniMax`	`MINIMAX_API_KEY`	`MINIMAX_API_BASE_URL`
`Mistral`	`MISTRAL_API_KEY`	`MISTRAL_API_BASE_URL`
`Novita AI`	`NOVITA_AI_API_KEY`	`NOVITA_AI_API_BASE_URL`
`OpenRouter`	`OPENROUTER_API_KEY`	`OPENROUTER_API_BASE_URL`
`SiliconCloud`	`SILICONCLOUD_API_KEY`	`SILICONCLOUD_API_BASE_URL`
`Together`	`TOGETHER_API_KEY`	`TOGETHER_API_BASE_URL`
`Venice AI`	`VENICE_AI_API_KEY`	`VENICE_AI_API_BASE_URL`
`Z AI`	`Z_AI_API_KEY`	`Z_AI_API_BASE_URL`

Native and special providers:

`DEFAULT_VENDOR`	Required variables	Optional variables
`Anthropic`	`ANTHROPIC_API_KEY`	`ANTHROPIC_API_BASE_URL`
`Azure`	`AZURE_API_KEY`, `AZURE_API_BASE_URL`, `AZURE_DEPLOYMENTS`	`AZURE_API_VERSION`
`AzureAIGateway`	`AZUREAIGATEWAY_BACKEND`, `AZUREAIGATEWAY_GATEWAY_URL`, `AZUREAIGATEWAY_SUBSCRIPTION_KEY`	`AZUREAIGATEWAY_API_VERSION`
`AzureEntra`	`AZUREENTRA_API_BASE_URL`, `AZUREENTRA_DEPLOYMENTS`	`AZUREENTRA_API_VERSION`; also provide Azure identity env such as `AZURE_CLIENT_ID`, `AZURE_CLIENT_SECRET`, `AZURE_TENANT_ID`
`Bedrock`	`BEDROCK_AWS_REGION`	`BEDROCK_API_KEY`, `BEDROCK_AWS_ACCESS_KEY_ID`, `BEDROCK_AWS_SECRET_ACCESS_KEY`; AWS credential-chain env also works
`Codex`	`CODEX_REFRESH_TOKEN`, `CODEX_ACCOUNT_ID`	`CODEX_ACCESS_TOKEN`, `CODEX_BASE_URL`, `CODEX_AUTH_BASE_URL`
`Copilot`	`COPILOT_TENANT_ID`, `COPILOT_CLIENT_ID`, plus either `COPILOT_ACCESS_TOKEN` or `COPILOT_REFRESH_TOKEN` with `COPILOT_CLIENT_SECRET`	`COPILOT_API_BASE_URL`, `COPILOT_TIME_ZONE`
`DigitalOcean`	`DIGITALOCEAN_INFERENCE_KEY`	`DIGITALOCEAN_INFERENCE_BASE_URL`, `DIGITALOCEAN_TOKEN`
`Exolab`	`EXOLAB_API_BASE_URL`, `EXOLAB_MODELS`	`EXOLAB_API_KEY`
`Gemini`	`GEMINI_API_KEY`	None
`LM Studio`	`LM_STUDIO_API_URL`	`LM_STUDIO_API_KEY`
`Ollama`	`OLLAMA_API_URL`	`OLLAMA_API_KEY`, `OLLAMA_HTTP_TIMEOUT`
`OpenAI`	`OPENAI_API_KEY`	`OPENAI_API_BASE_URL`
`Perplexity`	`PERPLEXITY_API_KEY`	None
`VertexAI`	`VERTEXAI_PROJECT_ID`	`VERTEXAI_REGION`; also provide Google Application Default Credentials such as `GOOGLE_APPLICATION_CREDENTIALS`

Other Runtime Options

Variable	Default	Description
`DEFAULT_MODEL_CONTEXT_LENGTH`	empty	Optional context length override for the default model.
`PORT`	Railway-provided	HTTP port. Railway injects this automatically.
`FABRIC_CONFIG_DIR`	`/home/appuser/.config/fabric`	Runtime config directory. Keep this aligned with the Railway volume mount.
`FABRIC_APP_DATA_DIR`	`/usr/local/share/fabric`	Read-only bundled seed data directory inside the image.
`PATTERNS_LOADER_GIT_REPO_URL`	`https://github.com/danielmiessler/fabric.git`	Source repo for pattern loading.
`PATTERNS_LOADER_GIT_REPO_PATTERNS_FOLDER`	`data/patterns`	Pattern folder inside the source repo.
`PROMPT_STRATEGIES_GIT_REPO_URL`	`https://github.com/danielmiessler/fabric.git`	Source repo for prompt strategies.
`PROMPT_STRATEGIES_GIT_REPO_STRATEGIES_FOLDER`	`data/strategies`	Strategy folder inside the source repo.
`CUSTOM_PATTERNS_DIRECTORY`	empty	Extra directory for custom Fabric patterns.
`LANGUAGE_OUTPUT`	empty	Optional default language output hint.
`JINA_AI_API_KEY`	empty	Optional Jina API key for web extraction helpers.
`YOUTUBE_API_KEY`	empty	Optional YouTube API key for metadata and comments.
`SPOTIFY_CLIENT_ID`	empty	Optional Spotify integration client ID.
`SPOTIFY_CLIENT_SECRET`	empty	Optional Spotify integration client secret.

Deployment Dependencies

Original Fabric repository: https://github.com/danielmiessler/fabric
Railway template source repository: https://github.com/RockinPaul/fabric_railway_template
Railway template deployment docs: https://docs.railway.com/templates/deploy
OpenRouter API keys: https://openrouter.ai/docs/api-keys
OpenAI API keys: https://platform.openai.com/docs/api-reference/authentication/api-keys
Anthropic API getting started: https://docs.anthropic.com/en/api/getting-started
Gemini API keys: https://ai.google.dev/gemini-api/docs/api-key

Implementation Details

The template repository is not a fork of Fabric. It holds only the Railway glue: a Dockerfile, an entrypoint and one patch file. The Dockerfile clones a pinned upstream tag, asserts the tag's commit so a moved tag fails the build, applies the patch and compiles the binary. Bumping Fabric is a two-line change.

The patch carries three small changes upstream has not taken: an unauthenticated /health route for the platform health check, the correct text/event-stream media type on the streaming chat endpoint, and a CORS origin read from FABRIC_ALLOWED_ORIGIN instead of being pinned to a local development server.

Patterns and strategies are baked into the image and copied into the volume on first boot, so a fresh deployment works without reaching out to GitHub. The copy is guarded by a marker file, so your own edits survive later boots.

Template-provided secret:

FABRIC_API_KEY=${{ secret(32) }}

Recommended first-deploy variables:

DEFAULT_VENDOR=OpenRouter
DEFAULT_MODEL=openai/gpt-4.1-mini
OPENROUTER_API_KEY=

OpenAI example:

DEFAULT_VENDOR=OpenAI
DEFAULT_MODEL=gpt-4.1-mini
OPENAI_API_KEY=

After deploy:

curl https://YOUR-FABRIC-DOMAIN/health
curl -H "X-API-Key: $FABRIC_API_KEY" https://YOUR-FABRIC-DOMAIN/models/names
curl -H "X-API-Key: $FABRIC_API_KEY" https://YOUR-FABRIC-DOMAIN/patterns/names

Set FABRIC_ALLOWED_ORIGIN if a browser application will read the streaming chat endpoint; it defaults to *, which is safe here because the API authenticates with a header rather than cookies.

Swagger is available at:

https://YOUR-FABRIC-DOMAIN/swagger/index.html

Why Deploy Fabric on Railway?

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 Fabric 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.