gandalf/secubox-openwrt
gandalf/secubox-openwrt
1
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(luci-app-localai): Fix chat timeout and port issues

Browse Source 
- Change default API port from 8080 to 8081
- Increase chat API timeout to 120 seconds (LLMs can be slow on ARM)
- Use custom fetch-based chat call with AbortController for timeout control
- Fix wget/curl timeout for RPCD backend

Resolves "XHR request timed out" errors when using chat with TinyLlama.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
CyberMind-FR 2026-01-22 05:09:14 +01:00
parent 612a1be6ea
commit 6b07a613f1
2 changed files with 52 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

56
package/secubox/luci-app-localai/htdocs/luci-static/resources/view/localai/chat.js
View File

@ -9,12 +9,52 @@ var callModels = rpc.declare({
expect: { models: [] }
});
var callChat = rpc.declare({
object: 'luci.localai',
method: 'chat',
params: ['model', 'messages'],
expect: { response: '' }
});
// Custom chat function with longer timeout (LLMs can be slow)
function callChatWithTimeout(model, messages, timeoutMs) {
return new Promise(function(resolve, reject) {
var timeout = timeoutMs || 120000; // 2 minutes default
var controller = new AbortController();
var timeoutId = setTimeout(function() {
controller.abort();
reject(new Error('Request timed out - model may need more time'));
}, timeout);
// Use ubus directly via L.Request or fetch
var payload = JSON.stringify({
jsonrpc: '2.0',
id: 1,
method: 'call',
params: [
L.env.sessionid || '00000000000000000000000000000000',
'luci.localai',
'chat',
{ model: model, messages: JSON.parse(messages) }
]
});
fetch(L.env.requestpath + 'admin/ubus', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: payload,
signal: controller.signal
})
.then(function(response) { return response.json(); })
.then(function(data) {
clearTimeout(timeoutId);
if (data.result && data.result[1]) {
resolve(data.result[1]);
} else if (data.error) {
reject(new Error(data.error.message || 'RPC error'));
} else {
resolve({ response: '', error: 'Unexpected response format' });
}
})
.catch(function(err) {
clearTimeout(timeoutId);
reject(err);
});
});
}
return view.extend({
title: _('LocalAI Chat'),
@ -137,8 +177,8 @@ return view.extend({
// Build messages array
this.messages.push({ role: 'user', content: message });
// Send to API
callChat(this.selectedModel, JSON.stringify(this.messages))
// Send to API (120s timeout for slow models)
callChatWithTimeout(this.selectedModel, JSON.stringify(this.messages), 120000)
.then(function(result) {
var loading = document.getElementById('loading-msg');
if (loading) loading.remove();

7
package/secubox/luci-app-localai/root/usr/libexec/rpcd/luci.localai
View File

@ -10,7 +10,7 @@ LOCALAI_CTL="/usr/sbin/localaictl"
# Load UCI config
load_config() {
config_load "$CONFIG"
config_get API_PORT main api_port "8080"
config_get API_PORT main api_port "8081"
config_get DATA_PATH main data_path "/srv/localai"
config_get MODELS_PATH main models_path "/srv/localai/models"
config_get MEMORY_LIMIT main memory_limit "2G"
@ -388,13 +388,14 @@ do_chat() {
local tmpfile="/tmp/localai_chat_$$"
local tmpfile_err="/tmp/localai_chat_err_$$"
# Use longer timeout for LLM responses (120 seconds)
if command -v curl >/dev/null 2>&1; then
curl -s -X POST "http://127.0.0.1:$API_PORT/v1/chat/completions" \
curl -s -m 120 -X POST "http://127.0.0.1:$API_PORT/v1/chat/completions" \
-H "Content-Type: application/json" \
-d "$request_body" \
-o "$tmpfile" 2>"$tmpfile_err"
else
wget -q -O "$tmpfile" --post-data "$request_body" \
wget -q -T 120 -O "$tmpfile" --post-data "$request_body" \
--header="Content-Type: application/json" \
"http://127.0.0.1:$API_PORT/v1/chat/completions" 2>"$tmpfile_err"
fi