How to Troubleshoot Port 11434 and CORS Bind Failures on Local Ollama and vLLM AI Deployments (Error 903)
Is your local AI service failing to serve API requests, or returning CORS errors when accessed from web interfaces? Learn to bind network hosts and allow origins correctly.
Resolving Network and CORS Access Blocks on AI APIs
When developing custom frontend web clients (like chat interfaces) that interact directly with local AI backends like Ollama or vLLM, you will frequently run into connection timeouts on port 11434 or Cross-Origin Resource Sharing (CORS) exceptions in your browser's console.
# Step 1: Diagnose Port Collisions
If Ollama fails to run, verify if another instance is already binding the local service address:
On Windows (Cmd):
``
netstat -ano | findstr 11434
`
On Linux/Mac:
`
sudo ss -lptn 'sport = :11434'
`
If another service is listening, terminate the corresponding process ID (PID) or change your API configurations port parameter.
# Step 2: Configure Global Host Bindings
By default, Ollama binds exclusively to 127.0.0.1 (localhost). If you want to access the API from other computers on your LAN or inside virtual containers, you must configure it to listen on all interfaces (0.0.0.0).
Add the environment variables in your active system environment profile:
# Step 3: Solve Browser CORS Blocks via OLLAMA_ORIGINS
If you access the Ollama API from a web browser app hosted on port 3000, the browser blocks responses because they originate from different origins. Establish origin rules:
Environment Variable:
(Allows absolute access) or http://localhost:3000,http://localhost:5173` for development scopes.After defining these, restart your Ollama/vLLM system services to enable the changes.