nodes/goodguy1963-ComfyUI-ThinkingLLM
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix: raise OrcaRouter profile timeout cap to 900s for reasoning-heavy models

Browse Source
This commit is contained in:
Marcel Hofer
2026-08-16 17:36:16 +02:00
parent 7e94e9336f
commit db01daff20
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
nodes/openai_compatible_api.py
+5 -1
View File
@@ -109,8 +109,12 @@ _BUILTIN_PROFILES = {
"Featherless": _cloud_profile(
"Featherless", "https://api.featherless.ai/v1", "FEATHERLESS_API_KEY"
),
# OrcaRouter serves reasoning-heavy models with long TTFT (e.g. obsidian/*
# uncensored variants) that can exceed the default 300 s deadline, so it gets
# a higher server-side cap. The widget still enforces a hard maximum of 3600.
"OrcaRouter": _cloud_profile(
"OrcaRouter", "https://api.orcarouter.ai/v1", "ORCAROUTER_API_KEY"
"OrcaRouter", "https://api.orcarouter.ai/v1", "ORCAROUTER_API_KEY",
max_timeout_seconds=900,
),
"Ollama (local)": _local_profile("Ollama", "http://127.0.0.1:11434/v1"),
"vLLM (local)": _local_profile("vLLM", "http://127.0.0.1:8000/v1", send_seed=True),