WIP

chat-ai/Dockerfile

@@ -64,8 +64,7 @@ RUN mkdir -p /home/llama/models
 EXPOSE 8080
 
 CMD ["llama-server", \
-     "--model",   "/home/llama/models/mistral-7b-instruct.gguf", \
+     "--model",   "/home/llama/models/Codestral-22B-v0.1-Q4_K_M.gguf", \
      "--host",    "0.0.0.0", \
      "--port",    "8080", \
-     "-c",        "4096", \
-     "-ngl",      "33"]
+     "-c",        "4096"]

chat-ai/compose.yml

@@ -30,7 +30,7 @@ services:
     environment:
       - GGML_VK_VISIBLE_DEVICES=0
     volumes:
-      - ../ai-models/mistral-7b-instruct.gguf:/home/llama/models/mistral-7b-instruct.gguf:ro
+      - ../ai-models:/home/llama/models:ro
       - ../workspace:/home/llama/workspace
 
     # Health check — vérifie que le serveur répond