michael/AI-Profile-Router
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Define four-profile production matrix with Medium default

Browse Source
This commit is contained in:
Mikei386
2026-08-22 11:30:38 +02:00
parent 977f8f5c76
commit 41b9c17f0f
33 changed files with 237 additions and 160 deletions
Download Patch File Download Diff File Expand all files Collapse all files
config/install.env.example
+14 -12
View File
@@ -28,15 +28,15 @@ WG_DNS=192.168.1.1
WG_ROUTE_AI_INTERNET=true
# Exact model artifacts. Never put access tokens in these URLs.
FAST_MODEL_FILE=qwen/Qwen3.8-27B-UD-IQ4_XS.gguf
FAST_MODEL_URL=https://huggingface.co/unsloth/Qwen3.8-27B-GGUF/resolve/main/Qwen3.8-27B-UD-IQ4_XS.gguf
FAST_MODEL_SHA256=40fac4050e940397dbf13087afd50f4734a11805bf9d65ef8ddd7483470e6199
MEDIUM_MODEL_FILE=qwen/Qwen3.8-27B-UD-IQ4_XS.gguf
MEDIUM_MODEL_URL=https://huggingface.co/unsloth/Qwen3.8-27B-GGUF/resolve/main/Qwen3.8-27B-UD-IQ4_XS.gguf
MEDIUM_MODEL_SHA256=40fac4050e940397dbf13087afd50f4734a11805bf9d65ef8ddd7483470e6199
LONG_MODEL_FILE=qwen/Qwen3.8-27B-UD-IQ4_XS.gguf
LONG_MODEL_URL=https://huggingface.co/unsloth/Qwen3.8-27B-GGUF/resolve/main/Qwen3.8-27B-UD-IQ4_XS.gguf
LONG_MODEL_SHA256=40fac4050e940397dbf13087afd50f4734a11805bf9d65ef8ddd7483470e6199
FAST_MODEL_FILE=qwen-mix/Qwen3.8-27B-IQ4-MIX.gguf
FAST_MODEL_URL=https://huggingface.co/vmarcelo/Qwen3.8-27B-MIX_GGUF/resolve/main/Qwen3.8-27B-IQ4-MIX.gguf
FAST_MODEL_SHA256=54879ae8738d5938f46cb3b8cbf16bf42b8c85b7d68d7c73f062b612ec183e36
MEDIUM_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
MEDIUM_MODEL_URL=https://huggingface.co/jpetrina/Qwen3.8-27B-IQ4_XS-pure-GGUF/resolve/main/qwen3.8-27b-IQ4_XS-pure.gguf
MEDIUM_MODEL_SHA256=ea5a3c45d407f9b9e5d2c0d647f0ea600f486f6b86b92b56d0823ba073dae675
LARGE_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
LARGE_MODEL_URL=https://huggingface.co/jpetrina/Qwen3.8-27B-IQ4_XS-pure-GGUF/resolve/main/qwen3.8-27b-IQ4_XS-pure.gguf
LARGE_MODEL_SHA256=ea5a3c45d407f9b9e5d2c0d647f0ea600f486f6b86b92b56d0823ba073dae675
ULTRA_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
ULTRA_MODEL_URL=https://huggingface.co/jpetrina/Qwen3.8-27B-IQ4_XS-pure-GGUF/resolve/main/qwen3.8-27b-IQ4_XS-pure.gguf
ULTRA_MODEL_SHA256=ea5a3c45d407f9b9e5d2c0d647f0ea600f486f6b86b92b56d0823ba073dae675
@@ -46,12 +46,14 @@ EXPERIMENTAL_MODEL_SHA256=40fac4050e940397dbf13087afd50f4734a11805bf9d65ef8ddd74
VISION_PROJECTOR_FILE=qwen/mmproj-BF16.gguf
VISION_PROJECTOR_URL=https://huggingface.co/unsloth/Qwen3.8-27B-GGUF/resolve/main/mmproj-BF16.gguf
VISION_PROJECTOR_SHA256=83ee4f4f205fa514161778c41df1ea14144faa0f713510893b63c2395f5c2d53
# FAST/LONG use the MTP tensor embedded in the IQ4-MIX GGUF. A separate
# All standard profiles use the MTP tensor embedded in their GGUF. A separate
# draft-model artifact is neither downloaded nor passed to llama-server.
FAST_CONTEXT=76800
MEDIUM_CONTEXT=94208
LONG_CONTEXT=131072
MEDIUM_CONTEXT=160000
MEDIUM_TENSOR_SPLIT=90,10
LARGE_CONTEXT=192000
LARGE_TENSOR_SPLIT=86,14
ULTRA_CONTEXT=262144
ULTRA_TENSOR_SPLIT=80,20
EXPERIMENTAL_CONTEXT=76800