---
title: "Deploy Open Fiesta | (Just Updated) Multi-Model AI Chat, Compare ChatGPT, Gemini And Llama Side By Side"
description: "Open Fiesta multi-model AI chat UI for OpenRouter, Gemini and Ollama"
category: "AI/ML"
url: https://railway.com/deploy/open-fiesta-or-just-updated-multi-model-
---

# Deploy Open Fiesta | (Just Updated) Multi-Model AI Chat, Compare ChatGPT, Gemini And Llama Side By Side

Open Fiesta multi-model AI chat UI for OpenRouter, Gemini and Ollama

**[Deploy Open Fiesta | (Just Updated) Multi-Model AI Chat, Compare ChatGPT, Gemini And Llama Side By Side on Railway](https://railway.com/template/open-fiesta-or-just-updated-multi-model-)**

Machine-readable deploy manifest (JSON, validated by TemplateCI): https://railway.com/deploy/open-fiesta-or-just-updated-multi-model-/manifest.json

- **Creator:** SuperSlowSloth
- **Category:** AI/ML

## Template content

### open-fiesta

- **Image:** ghcr.io/jaainil/open-fiesta-ghcr:latest@sha256:404737ed2e4dd8a9bc1967edd35fd5b5b5e9f75042eba4a7b9de93936a1c851b
- **Health check:** /
- **Public domain:** Yes

## Documentation

# Deploy and Host Open Fiesta on Railway

**Built for an LLM agent to deploy and use.** The deploy form has nothing to fill in: the service has no account system, no database and no secret of its own, so there is no step to do by hand after the deploy. Open Fiesta is an open-source (MIT) chat interface that sends one prompt to many AI models at once and shows the answers side by side, with support for OpenRouter, Google Gemini, Mistral, Ollama and a set of free models. The contract below lists the address, the endpoints and one script that was run against a fresh deploy.

## About Hosting Open Fiesta

This template runs one service from the upstream Open Fiesta image (`ghcr.io/jaainil/open-fiesta-ghcr`, the newest build upstream publishes, pinned by digest). It is a Next.js server that listens on Railway's injected `PORT`, so the public domain and the healthcheck reach the same port. The app is stateless: conversations, settings and provider keys are kept in the visitor's browser, so the template has no volume and nothing to back up.

**Keys are entered in the browser.** After the deploy, open the service domain, choose Settings and paste your own OpenRouter, Gemini, Mistral or Ollama details. The server forwards them to the provider for that one request and does not store them. The template deliberately sets no provider key as a variable, so a public domain can never spend yours.

**Know what the upstream app does before you share the domain.** The free-model route (`/api/open-provider`) falls back to a shared Pollinations token written into the upstream source when no key is supplied, so those models work out of the box but share one upstream allowance with everyone using that token. The Ollama routes (`/api/ollama`, `/api/ollama/validate`) fetch whatever `baseUrl` the request names, so anyone who can reach the domain can make the server request an address of their choice. Keep the domain to people you trust, or put it behind an access proxy, if that matters to you.

### Agent contract

- Public URL: `https://RAILWAY_SERVICE_OPEN_FIESTA_URL` (the generated domain of the `open-fiesta` service)
- Private URL: `http://open-fiesta.railway.internal:8080`
- Credentials: none to present. There is no login and the template defines no variables; a provider key (OpenRouter, Gemini, Mistral) is supplied by the caller in the request body or in the browser settings, never read from the server
- Auth: no header or cookie. Endpoints are `POST` with a JSON body: `/api/openrouter/validate` (`slug`, optional `apiKey`), `/api/open-provider` (`model`, `messages`, optional `apiKey`), `/api/ollama/validate` (`slug`, `baseUrl`); a missing field answers `400`
- Health: `GET /` returns 200

```bash
#!/bin/sh
# needs: RAILWAY_SERVICE_OPEN_FIESTA_URL (the generated domain of the open-fiesta service); also curl
R="https://$RAILWAY_SERVICE_OPEN_FIESTA_URL"; J="Content-Type: application/json"
n=0; until [ "$(curl -s -o /dev/null -w '%{http_code}' "$R/")" = 200 ]; do n=$((n+1)); [ "$n" -le 60 ] || break; sleep 5; done
curl -s -o /dev/null -w "health /: %{http_code}\n" "$R/"
curl -s "$R/" | grep -q 'OpenSource Fiesta'; echo "page names the app: $?"
# write-auth: none - the chat API takes no credential of its own; each provider key travels in the request body and is never stored server-side
curl -s -w " write no credential (model catalog lookup): %{http_code}\n" -H "$J" -d '{"slug":"openai/gpt-4o-mini"}' "$R/api/openrouter/validate"
curl -s -w " write bad input: %{http_code}\n" -H "$J" -d '{}' "$R/api/open-provider"
curl -s -w " ollama unreachable: %{http_code}\n" -H "$J" -d '{"slug":"llama3","baseUrl":"http://127.0.0.1:1"}' "$R/api/ollama/validate"
```

A fresh deploy prints `health /: 200`, then `page names the app: 0`, then `{"ok":true,"exists":true} write no credential (model catalog lookup): 200`, `{"error":"Missing model id"} write bad input: 400` and the Ollama answer `Cannot connect to Ollama instance` with `ollama unreachable: 502`. The catalog lookup needs outbound internet from the service; the chat calls themselves need a provider key or a reachable Ollama.

## Why Deploy Open Fiesta 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 Open Fiesta 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.

- **Agent-ready** - no form, no secret to generate and one script that was run against a fresh deploy.
- **Stateless** - no database or volume; restart or scale it without migrating anything.
- **Bring your own models** - OpenRouter, Gemini, Mistral and a self-hosted Ollama on the same private network.
- **Pinned** - the upstream image is fixed by digest, so a deploy does not change under you.

## Common Use Cases

- **Compare ChatGPT, Gemini and Llama answers side by side** - send one prompt to several models and read the results together.
- **Team playground for OpenRouter models** - one shared page where everyone brings their own key.
- **Front end for a private Ollama on Railway** - point the app at `http://ollama.railway.internal:11434`.
- **Self-hosted alternative to a multi-model chat subscription** - own the interface, pay only the providers you use.

## Dependencies for Open Fiesta Hosting

- The one `open-fiesta` service created by the template. Nothing else is required; add your own provider keys in the app, or an Ollama service on the same project for local models.

### Deployment Dependencies

- [Open Fiesta repository](https://github.com/jaainil/open-fiesta)
- [OpenRouter](https://openrouter.ai)


## Similar templates

- [Chat Chat](https://railway.com/deploy/-WWW5r) — Chat Chat, your own unified chat and search to AI platform.
- [Weaviate](https://railway.com/deploy/weaviate-1) — Vector database with hybrid search and an API key-protected API
- [Möbius Personal AI OS](https://railway.com/deploy/mobius-v01-personal-ai-os) — Self-hosted AI agent that builds and runs apps for you, with memory.

Open this page in a browser: https://railway.com/deploy/open-fiesta-or-just-updated-multi-model-
