Hard-coded timeout causes failures with OpenAI Deep Research models

[ByungsunAhn]

🐛 Bug Report: Hard-coded timeout causes failures with OpenAI Deep Research models

Problem

OpenAI Deep Research models (o3, o4-mini-deep-research, etc.) require extended thinking time (often 10-30 minutes), but LibreChat has a hard-coded 3-minute timeout in api/server/services/Runs/handle.js:

timeout = 60000 * 3,  // 180 seconds (3 minutes)

Error Logs

error in server stream.
SyntaxError: JSON.parse: unexpected end of data at line 1 column 1 of the JSON data

Current Behavior

• Deep Research requests timeout after 3 minutes
• User gets JSON parsing errors
• Connection is lost during thinking phase

Expected Behavior

• Configurable timeout via environment variable
• Support for longer Deep Research operations

Proposed Solution

Add environment variable support:

# .env
OPENAI_TIMEOUT_MS=1800000  # 30 minutes for Deep Research

// api/server/services/Runs/handle.js
timeout = process.env.OPENAI_TIMEOUT_MS || (60000 * 3),

Impact

• Affects all OpenAI Deep Research model users
• Critical for o3/o4-mini-deep-research functionality

[ByungsunAhn]

nvm it was custom/nginx.conf timeout

location / {
   ....
    proxy_read_timeout 1800s;  # 30 min (OpenAI Deep Research)
}

location /api/ {
   ....
    proxy_read_timeout 1800s;  # 30 min (OpenAI Deep Research)
}

[ByungsunAhn]

nvm it was cloudflare default timeout