ucode-agent 1.61.1 → 1.62.0

package/README.md

@@ -96,6 +96,9 @@ Served by Google (aistudio.google.com), free with a key.
 
 `/model` shows them and switches. `ucode -m gemini-3.5-flash` starts on one.
 
+When Google is overloaded and Flash-Lite stops answering, ucode carries on with
+3.5 Flash by itself and goes back to Flash-Lite a few minutes later.
+
 ## What it does
 
 **Twenty-one tools.** `create_app`, `read_file`, `read_files`, `write_file`,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ucode-agent",
-  "version": "1.61.1",
+  "version": "1.62.0",
   "description": "ucode - a terminal coding agent that reads, edits and runs your code, on Google Gemini models.",
   "type": "module",
   "main": "ucode.js",

package/src/core/loop.js

@@ -27,7 +27,7 @@ import { spawn } from 'node:child_process';
 
 import {
   ask, model, setModel, modelName, modelList, contextLimit, rateLimits,
-  estimateConversation, MODELS, DEFAULT_MODEL, PROVIDER, fallbackFor,
+  estimateConversation, MODELS, DEFAULT_MODEL, PROVIDER, fallbackFor, backupFor,
 } from './provider.js';
 import {
   tools, runTool, describe, setRoot, setConfirm, setRequest,
@@ -2095,6 +2095,8 @@ export class Agent {
     if (++this.failovers > MAX_FAILOVERS) return false;
     const from = model();
     let next = fallbackFor(from, this.tried);
+    const backup = err.kind !== 'rate_limit' && backupFor(from);
+    if (!next && backup && !this.tried?.has(backup)) next = backup;
 
     const why = err.kind === 'rate_limit' ? 'busy' : err.kind === 'timeout' ? 'too slow to answer' : 'not answering';
 

package/src/core/provider.js

@@ -104,6 +104,15 @@ export function fallbackFor(id, tried = new Set()) {
   return null;
 }
 
+/**
+ * Where a Flash-Lite model goes when Google is overloaded (503) or it freezes.
+ * Unlike fallbackFor this is on by default: both Lite models go down together
+ * at peak hours while Flash keeps answering, and a build that waits on them
+ * can wait for hours. UCODE_BACKUP=0 turns it off.
+ */
+export const backupFor = (id) =>
+  /flash-lite/.test(id) && process.env.UCODE_BACKUP !== '0' ? 'gemini-3.5-flash' : null;
+
 /** Seconds to wait on successive rate limits that come with no retry-after. */
 const RATE_LIMIT_BACKOFF = [5, 10, 20];
 
@@ -711,7 +720,7 @@ export async function ask(messages, tools = [], opts = {}) {
           kind: 'stalled',
           attempted: `asking ${modelName(id)} for a reply`,
           failed: `${modelName(id)} froze ${MAX_STALLS} times in a row — it took the request and then sent nothing.`,
-          fix: 'Its free endpoint is struggling right now. Run /model and pick another one; North Mini Code answers soonest.',
+          fix: 'Google is struggling with it right now. Run /model and pick another one.',
           cause: problem,
         });
       }
@@ -741,6 +750,7 @@ export async function ask(messages, tools = [], opts = {}) {
       if (!worthRetrying || attempt === attempts || opts.signal?.aborted) break;
       if (printed > 0) break; // half an answer is on screen; do not print it twice
       if (problem.detail?.handed) break; // its first tool calls are already running
+      if (problem.kind === 'timeout' && backupFor(id)) break; // one freeze is enough: the loop moves to the backup
 
       // A stalled provider needs longer to come back than a dropped socket
       // does, and the wait is narrated so a slow turn never looks like a hang.