feat: add Railway deployment support and improvements

- Add Railway configuration and deployment guide
- Update entrypoint.sh to support Railway PORT env variable
- Add Bun server idle timeout configuration (300s) to prevent premature timeouts
- Set X-Initiator to "agent" for claude-sonnet-4.5 model requests
- Update .dockerignore to exclude Railway-specific files

Author: sachnun

.dockerignore

@@ -11,3 +11,7 @@ tests/
 *.md
 
 .eslintcache
+
+# Railway specific
+railway.toml
+RAILWAY.md

RAILWAY.md

@@ -0,0 +1,36 @@
+# Railway Deployment Guide
+
+## Environment Variables
+
+Set these in Railway dashboard under your service settings:
+
+### Required
+- `GH_TOKEN` - Your GitHub token (generate using `bun run auth` locally first)
+
+### Optional
+- `PORT` - Railway will set this automatically (default: 4141)
+- `ACCOUNT_TYPE` - "individual", "business", or "enterprise" (default: individual)
+
+## Deployment Steps
+
+1. **Generate GitHub Token locally** (one-time setup):
+   ```bash
+   bun run auth
+   # This will save your token to ~/.local/share/copilot-api/github-token
+   # Copy the token content for Railway
+   ```
+
+2. **Deploy to Railway**:
+   - Connect your GitHub repository to Railway
+   - Railway will automatically detect the Dockerfile
+   - Set environment variable `GH_TOKEN` in Railway dashboard
+
+3. **Access your deployed API**:
+   - Railway will provide a public URL like `https://your-app.railway.app`
+   - Test with: `curl https://your-app.railway.app/v1/models`
+
+## Additional Options
+
+You can pass additional flags via Railway's service settings if needed:
+- Set `RAILWAY_RUN_COMMAND` to override the default command
+- Example: `bun run dist/main.js start -g $GH_TOKEN -a business`

entrypoint.sh

@@ -1,9 +1,17 @@
 #!/bin/sh
+
+# Use PORT from Railway if available, otherwise default to 4141
+PORT="${PORT:-4141}"
+
 if [ "$1" = "--auth" ]; then
   # Run auth command
   exec bun run dist/main.js auth
 else
-  # Default command
-  exec bun run dist/main.js start -g "$GH_TOKEN" "$@"
+  # Default command - use GH_TOKEN if provided
+  if [ -n "$GH_TOKEN" ]; then
+    exec bun run dist/main.js start -g "$GH_TOKEN" -p "$PORT" "$@"
+  else
+    exec bun run dist/main.js start -p "$PORT" "$@"
+  fi
 fi
 

railway.toml

@@ -0,0 +1,10 @@
+[build]
+builder = "dockerfile"
+dockerfilePath = "Dockerfile"
+
+[deploy]
+numReplicas = 1
+restartPolicyType = "on_failure"
+restartPolicyMaxRetries = 10
+
+# Railway will automatically expose the PORT that your app listens on

src/services/copilot/create-chat-completions.ts

@@ -22,10 +22,13 @@ export const createChatCompletions = async (
     ["assistant", "tool"].includes(msg.role),
   )
 
+  // Check if model is claude-sonnet-4.5 (should also be treated as agent)
+  const isClaudeSonnet45 = payload.model === "claude-sonnet-4.5"
+
   // Build headers and add X-Initiator
   const headers: Record<string, string> = {
     ...copilotHeaders(state, enableVision),
-    "X-Initiator": isAgentCall ? "agent" : "user",
+    "X-Initiator": isAgentCall || isClaudeSonnet45 ? "agent" : "user",
   }
 
   const response = await fetch(`${copilotBaseUrl(state)}/chat/completions`, {

src/start.ts

@@ -117,6 +117,12 @@ export async function runServer(options: RunServerOptions): Promise<void> {
   serve({
     fetch: server.fetch as ServerHandler,
     port: options.port,
+    // Bun-specific options for longer timeouts
+    bun: {
+      // Disable request timeout (set to 0 or very high value)
+      // Default is 10 seconds, we set to 5 minutes for long streaming requests
+      idleTimeout: 300, // 5 minutes in seconds
+    },
   })
 }