Improve models configuration

README.md

@@ -31,14 +31,13 @@ Note that it is probably possible to run the project on other GPUs or modern Mac
 
 ## How to launch the server
 
-Choose the models you wish to use in the `docker-compose.yaml` file.
+Choose the models you wish to use in the `.env` file. Put the embedding model first and the default model second.
 
 ```bash
-environment:
-      - MODELS=...
+MODELS=gemma3n:e4b,nomic-embed-text,gpt-oss:20b,llama3.1:8b,llama3.2:3b,gemma3:12b-it-qat
 ```
 
-Add an API key to secure server access by adding a `.env` file like this:
+Add an API key to secure server access by adding a `.env` file like this (you can generate one with  `openssl rand -hex 16`):
 
 ```bash
 LLM_API_KEY=1234567890

docker-compose.yml

@@ -24,7 +24,7 @@ services:
     build:
       dockerfile: src/ollama_provision/Dockerfile
     environment:
-      - MODELS=nomic-embed-text,llama3.1:8b,qwen2.5:7b,gemma3:12b-it-qat,gemma3:27b-it-qat,gpt-oss:20b
+      - MODELS=${MODELS}
     volumes:
       - ollama:/root/.ollama
     restart: no

setup_desktop.sh

@@ -21,6 +21,31 @@ if [[ ! -z $OLLAMA_HOST ]]; then
   cat ~/.config/aichat/config.yaml | sed "s/http/https/" >~/.config/aichat/config.yaml.tmp
   cat ~/.config/aichat/config.yaml.tmp | sed "s/localhost:11434/${OLLAMA_HOST}/" >~/.config/aichat/config.yaml
 fi
+DEFAULT_MODEL=$(cat .env | grep '^MODELS=' | cut -d'=' -f2 | cut -d',' -f2)
+EMBEDDING_MODEL=$(cat .env | grep '^MODELS=' | cut -d'=' -f2 | cut -d',' -f1)
+if [[ -z ${EMBEDDING_MODEL} ]] || [[ -z ${DEFAULT_MODEL} ]]; then
+  echo "MODELS in .env are missing."
+  exit 1
+fi
+
+cat ~/.config/aichat/config.yaml | sed "s/__EMBEDDING_MODEL__/ollama:${EMBEDDING_MODEL}/" >~/.config/aichat/config.yaml.tmp
+cat ~/.config/aichat/config.yaml.tmp | sed "s/__DEFAULT_MODEL__/ollama:${DEFAULT_MODEL}/" >~/.config/aichat/config.yaml
+
+MODELS=$(cat .env | grep '^MODELS=' | cut -d'=' -f2 | cut -d',' -f2- | tr ',' ' ')
+for i in ${MODELS}; do
+  echo "    - name: ${i}" >>~/.config/aichat/config.yaml
+  echo
+  read -p "Does the model ${i} support function calling? [y/N] " -n 1 -r
+  if [[ $REPLY =~ ^[Yy]$ ]]; then
+    echo "      supports_function_calling: true" >>~/.config/aichat/config.yaml
+  else
+    echo "      supports_function_calling: false" >>~/.config/aichat/config.yaml
+  fi
+  echo
+  read -rp "What is the maximum number of input tokens ${i} supports? [128000] " max_input_tokens
+  max_input_tokens=${max_input_tokens:-128000}
+  echo "      max_input_tokens: ${max_input_tokens}" >>~/.config/aichat/config.yaml
+done
 
 mkdir -p ~/.config/aichat/roles
 cp src/aichat/roles/* ~/.config/aichat/roles/

src/aichat/config.yaml

@@ -1,9 +1,9 @@
 # see https://github.com/sigoden/aichat/blob/main/config.example.yaml
 
-rag_embedding_model: ollama:nomic-embed-text
+rag_embedding_model: __EMBEDDING_MODEL__
 rag_chunk_size: 8192
 rag_chunk_overlap: 409
-model: ollama:gemma3:12b-it-qat
+model: __DEFAULT_MODEL__
 temperature: 0
 clients:
 - type: openai-compatible
@@ -11,21 +11,3 @@ clients:
   api_base: http://localhost:11434/v1
   api_key: __LLM_API_KEY__
   models:
-    - name: gemma3:4b
-      supports_function_calling: true
-      max_input_tokens: 128000
-    - name: gemma3:12b-it-qat
-      supports_function_calling: true
-      max_input_tokens: 128000
-    - name: gemma3:27b-it-qat
-      supports_function_calling: true
-      max_input_tokens: 128000
-    - name: gpt-oss:20b
-      supports_function_calling: true
-      max_input_tokens: 128000
-    - name: llama3.1:8b
-      supports_function_calling: true
-      max_input_tokens: 128000
-    - name: qwen2.5:7b
-      supports_function_calling: true
-      max_input_tokens: 128000