feat: add Ollama (local AI) integration for AIUI

- Create Ollama provider plugin in AIUI (ollama-provider.ts)
- Register Ollama alongside Claude in plugin system
- Auto-detect installed models via /api/tags endpoint
- Add Ollama proxy in mock backend (forwards to localhost:11434)
- Add nginx proxy rules for /aiui/api/ollama/ (both HTTP and HTTPS)
- Rebuild AIUI dist with Ollama provider included
- Qwen 2.5 Coder 3B installed on dev server via Ollama

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

neode-ui/docker/nginx-demo.conf

@@ -86,6 +86,18 @@ http {
             proxy_set_header X-Real-IP $remote_addr;
         }
 
+        # Proxy Ollama (local AI) requests to backend
+        location /aiui/api/ollama/ {
+            proxy_pass http://neode-backend:5959;
+            proxy_http_version 1.1;
+            proxy_set_header Host $host;
+            proxy_set_header X-Real-IP $remote_addr;
+            proxy_buffering off;
+            proxy_cache off;
+            proxy_read_timeout 300s;
+            proxy_set_header Connection "";
+        }
+
         # Proxy Claude API requests to backend (which handles API key + streaming)
         location /aiui/api/claude/ {
             proxy_pass http://neode-backend:5959;

neode-ui/mock-backend.js

@@ -1087,6 +1087,74 @@ app.post('/aiui/api/claude/*', (req, res) => {
   proxyReq.end()
 })
 
+// Ollama (local AI) proxy — forwards to localhost:11434
+app.all('/aiui/api/ollama/*', (req, res) => {
+  const ollamaPath = '/' + req.params[0]
+  const bodyStr = JSON.stringify(req.body)
+
+  const options = {
+    hostname: '127.0.0.1',
+    port: 11434,
+    path: ollamaPath,
+    method: req.method,
+    headers: {
+      'Content-Type': 'application/json',
+      'Content-Length': Buffer.byteLength(bodyStr),
+    },
+  }
+
+  const proxyReq = http.request(options, (proxyRes) => {
+    res.writeHead(proxyRes.statusCode, proxyRes.headers)
+    proxyRes.pipe(res)
+  })
+
+  proxyReq.on('error', (err) => {
+    const msg = err.message || err.code || 'Ollama not available'
+    console.error('[Ollama Proxy] Error:', msg)
+    if (!res.headersSent) {
+      res.status(502).json({ error: msg })
+    }
+  })
+
+  if (req.method !== 'GET' && req.method !== 'HEAD') {
+    proxyReq.write(bodyStr)
+  }
+  proxyReq.end()
+})
+
+// =============================================================================
+// Ollama Local AI Proxy (forwards to Ollama on localhost:11434)
+// =============================================================================
+app.all('/api/ollama/*', (req, res) => {
+  const ollamaPath = '/' + req.params[0]
+  const isPost = req.method === 'POST'
+  const bodyStr = isPost ? JSON.stringify(req.body) : null
+
+  const options = {
+    hostname: '127.0.0.1',
+    port: 11434,
+    path: ollamaPath,
+    method: req.method,
+    headers: { 'Content-Type': 'application/json' },
+  }
+
+  const proxyReq = http.request(options, (proxyRes) => {
+    res.writeHead(proxyRes.statusCode, proxyRes.headers)
+    proxyRes.pipe(res)
+  })
+
+  proxyReq.on('error', (err) => {
+    const msg = err.message || err.code || 'Ollama not available'
+    console.error('[Ollama Proxy] Error:', msg)
+    if (!res.headersSent) {
+      res.status(502).json({ error: msg })
+    }
+  })
+
+  if (bodyStr) proxyReq.write(bodyStr)
+  proxyReq.end()
+})
+
 // Web search stub (no search engine configured in demo)
 app.get('/api/web-search', (req, res) => {
   res.json({ results: [] })