Update to qwen:0.5b

app/rag.py

@@ -15,8 +15,8 @@ class ChatPDF:
     chain = None
 
     def __init__(self):
-        self.model = ChatOllama(model="tinyllama")
-        self.text_splitter = RecursiveCharacterTextSplitter(chunk_size=1024, chunk_overlap=96)
+        self.model = ChatOllama(model="qwen:0.5b")
+        self.text_splitter = RecursiveCharacterTextSplitter(chunk_size=2048, chunk_overlap=128)
         self.prompt = PromptTemplate.from_template(
             """
             You are an assistant for question-answering tasks. Use the following pieces of context 
@@ -36,7 +36,7 @@ class ChatPDF:
         self.retriever = vector_store.as_retriever(
             search_type="similarity_score_threshold",
             search_kwargs={
-                "k": 7,
+                "k": 15,
                 "score_threshold": 0.5,
             },
         )

start_service.sh

@@ -7,7 +7,7 @@ ollama serve &
 sleep 5
 
 # Pull and run <YOUR_MODEL_NAME>
-ollama pull tinyllama
+ollama pull qwen:0.5b
 
 #
 fastapi run /code/app/main.py --port 7860