---
title: "Deploy Pawwake"
description: "Give your AI long-term memory, across any OpenAI-compatible client"
category: "AI/ML"
url: https://railway.com/deploy/pawwake
---

# Deploy Pawwake

Give your AI long-term memory, across any OpenAI-compatible client

**[Deploy Pawwake on Railway](https://railway.com/template/pawwake)**

Machine-readable deploy manifest (JSON, validated by TemplateCI): https://railway.com/deploy/pawwake/manifest.json

- **Creator:** Pawwake
- **Category:** AI/ML
- **Total deploys:** 1

## Template content

### pawwake

- **Source:** garan0613/pawwake
- **Health check:** /
- **Public domain:** Yes

### postgresql

- **Image:** pgvector/pgvector:0.8.7-pg16

## Documentation

# Deploy and Host Pawwake with Railway

Pawwake is an open-source memory gateway for AI chat. It sits between an OpenAI-compatible chat client and an OpenAI-compatible LLM API, extracts durable facts from conversations, and retrieves relevant memories in later chats across sessions and apps.

## About Hosting Pawwake

This template deploys two services: the Pawwake gateway, built from the public GitHub repository with its Dockerfile, and PostgreSQL with the pgvector extension on a persistent volume. DATABASE_URL is wired automatically, while DASHBOARD_PASSWORD, GATEWAY_SECRET, and SESSION_SECRET are generated at deploy time.

Once the services are live, copy DASHBOARD_PASSWORD and GATEWAY_SECRET from the pawwake service's Variables tab. Sign in at /dashboard, then enter your LLM API base URL, API key, and default model under Settings.

Railway bills by usage. A two-day idle test used about $0.23 in Railway resources, equivalent to roughly $3.4 per month. Actual usage varies, and LLM API charges are separate.

## Common Use Cases

- Give an AI companion or roleplay character memory that lasts across chats
- Share one memory store across multiple chat apps and devices
- Control context size with cache-friendly context rotation and summaries
- Browse, edit, pin, export, restore, and import memories and chat history in a web dashboard
- Let MCP-capable clients add and search memories through the built-in /mcp endpoint

## Dependencies for Pawwake Hosting

- Python with FastAPI, built from the repository's Dockerfile
- PostgreSQL with pgvector
- An OpenAI-compatible LLM API, such as OpenRouter, OpenAI, DeepSeek, or a compatible local endpoint
- Optional: an OpenAI-compatible embedding API for semantic memory search

### Deployment Dependencies

- Source code and documentation: https://github.com/garan0613/pawwake
- The main README is written in Chinese

### Implementation Details

For chat clients that support custom headers:

- API base URL: https://<your-pawwake-domain>/v1
- API key field: any non-empty placeholder
- Custom header: X-Gateway-Key: <GATEWAY_SECRET>

Pawwake is licensed under AGPL-3.0.

### Why Deploy Pawwake on Railway?

Railway runs the Pawwake gateway and pgvector database together, wires their private connection, attaches persistent storage, exposes the gateway over HTTPS, and keeps secrets in service variables. After deployment, only the model provider settings need to be filled in through Pawwake's dashboard.

## 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 v0.1 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/pawwake