From 121575754c45feee7735ed39f093863542b1e6fa Mon Sep 17 00:00:00 2001 From: boy Date: Fri, 24 Jul 2026 21:29:40 +0200 Subject: [PATCH] feat(litellm): add local-qwen-3.6-hass model route and update request_timeout to 600s --- litellm/config.yaml | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) diff --git a/litellm/config.yaml b/litellm/config.yaml index c96724f..f13f481 100644 --- a/litellm/config.yaml +++ b/litellm/config.yaml @@ -91,13 +91,24 @@ model_list: max_input_tokens: 524288 is_public_model_group: true -# Re-enabled 2026-07-08 — llama.cpp, canonical endpoint set via .env - litellm_params: api_base: LLAMA_CLASSIFIER_URL_PLACEHOLDER api_key: local-token model: openai/local-qwen-3.6 - request_timeout: 120 + extra_body: + chat_template_kwargs: + preserve_thinking: true + request_timeout: 600 model_name: local-qwen-3.6 +- litellm_params: + api_base: LLAMA_CLASSIFIER_URL_PLACEHOLDER + api_key: local-token + model: openai/local-qwen-3.6-hass + extra_body: + chat_template_kwargs: + enable_thinking: false + request_timeout: 600 + model_name: local-qwen-3.6-hass - litellm_params: api_base: LLAMA_CLASSIFIER_URL_PLACEHOLDER api_key: local-token