From 969669b6471d5e9baa97a25cb2e6df3748ae71ea Mon Sep 17 00:00:00 2001
From: Classic298 <27028174+Classic298@users.noreply.github.com>
Date: Fri, 24 Jul 2026 09:27:24 +0200
Subject: [PATCH] refac: fetch the terminal system prompt per request without a
 cache

Drop the module-level TTL cache and fetch the system prompt directly in
get_terminal_tools, gathered with the existing uncached per-request cwd
fetch that already follows this pattern. The fetch uses the user's own
credentials and falls back to the set_terminal_servers snapshot, so a
cold or unreachable instance degrades to the previous behaviour instead
of needing an error cache. Also restore the 3s timeout: on the request
path a 30s wait would stall chat completions, and a cold instance is
covered by the snapshot fallback until it warms up.
---
 backend/open_webui/utils/tools.py | 42 +++----------------------------
 1 file changed, 3 insertions(+), 39 deletions(-)

diff --git a/backend/open_webui/utils/tools.py b/backend/open_webui/utils/tools.py
index de9b97ca60..0c55805c9e 100644
--- a/backend/open_webui/utils/tools.py
+++ b/backend/open_webui/utils/tools.py
@@ -8,7 +8,6 @@ import json
 import logging
 import os
 import re
-import time
 from functools import partial, update_wrapper
 from typing import (
     Any,
@@ -1094,8 +1093,7 @@ async def get_terminal_system_prompt(
     base = base_url.rstrip('/')
     try:
         async with aiohttp.ClientSession(
-            # 30s: orchestrator instances may cold-start before answering
-            timeout=aiohttp.ClientTimeout(total=30),
+            timeout=aiohttp.ClientTimeout(total=3),
             trust_env=True,
         ) as session:
             # 1. Check feature flag
@@ -1120,39 +1118,6 @@ async def get_terminal_system_prompt(
     return None
 
 
-_TERMINAL_SYSTEM_PROMPT_TTL = 300  # seconds
-_TERMINAL_SYSTEM_PROMPT_ERROR_TTL = 60  # seconds
-_terminal_system_prompt_cache: dict[tuple[str, str], tuple[float, str | None]] = {}
-
-
-async def get_cached_terminal_system_prompt(
-    base_url: str,
-    headers: dict,
-    cookies: dict | None = None,
-) -> str | None:
-    """Return a terminal server's system prompt, cached per (URL, user).
-
-    Each user may resolve to a different orchestrator instance. Failures
-    are cached briefly so an unreachable instance doesn't stall every
-    request.
-    """
-    key = (base_url, headers.get('X-User-Id', ''))
-    now = time.monotonic()
-
-    cached = _terminal_system_prompt_cache.get(key)
-    if cached and cached[0] > now:
-        return cached[1]
-
-    prompt = await get_terminal_system_prompt(base_url, headers, cookies)
-
-    if len(_terminal_system_prompt_cache) > 4096:
-        for stale_key in [k for k, (expiry, _) in _terminal_system_prompt_cache.items() if expiry <= now]:
-            _terminal_system_prompt_cache.pop(stale_key, None)
-    ttl = _TERMINAL_SYSTEM_PROMPT_TTL if prompt else _TERMINAL_SYSTEM_PROMPT_ERROR_TTL
-    _terminal_system_prompt_cache[key] = (now + ttl, prompt)
-    return prompt
-
-
 async def set_terminal_servers(request: Request):
     """Load and cache OpenAPI specs from all TERMINAL_SERVER_CONNECTIONS."""
     connections = await Config.get('terminal_server.connections', []) or []
@@ -1286,11 +1251,10 @@ async def get_terminal_tools(
     if session_id:
         headers['X-Session-Id'] = session_id
 
-    # Live fetch with the user's credentials; the set_terminal_servers
-    # snapshot is the fallback.
+    # Fetch live with the user's credentials so prompt changes apply without a restart
     terminal_cwd, system_prompt = await asyncio.gather(
         get_terminal_cwd(server_data['url'], headers, cookies),
-        get_cached_terminal_system_prompt(server_data['url'], headers, cookies),
+        get_terminal_system_prompt(server_data['url'], headers, cookies),
     )
     if not system_prompt:
         system_prompt = server_data.get('system_prompt')