update

Dockerfile

@@ -0,0 +1,16 @@
+FROM python:3.9
+
+RUN useradd -m -u 1000 user
+USER user
+ENV PATH="/home/user/.local/bin:$PATH"
+
+WORKDIR /app
+
+COPY --chown=user ./requirements.txt requirements.txt
+RUN pip install --no-cache-dir --upgrade -r requirements.txt
+
+RUN curl -fsSL https://ollama.com/install.sh | sh
+RUN ollama serve & ollama pull MarcoAland/llama3.1-rag-indo
+
+COPY --chown=user . /app
+CMD python app.py

RAGModule.py

@@ -0,0 +1,63 @@
+# Embedding model builder
+from llama_index.embeddings.huggingface import HuggingFaceEmbedding
+from llama_index.core import Settings, SimpleDirectoryReader, VectorStoreIndex
+from llama_index.core.retrievers import VectorIndexRetriever
+from llama_index.core.query_engine import RetrieverQueryEngine
+from llama_index.core.postprocessor import SimilarityPostprocessor
+
+def set_embed_model(model_name: str,
+                    chunk_size: int = 256,
+                    chunk_overlap: int = 25) -> None:
+    Settings.llm = None
+    Settings.embed_model = HuggingFaceEmbedding(model_name=model_name)
+    Settings.chunk_size = chunk_size
+    Settings.chunk_overlap = chunk_overlap
+
+
+class RAGModule:
+    def __init__(self,
+                 llm_model: str = "MarcoAland/llama3.1-rag-indo",
+                 embedding_model: str = "MarcoAland/Indo-bge-m3",
+                 docs_path: str = "data",
+                 top_k: int = 3,
+                 similarity_cutoff: float = 0.3):
+
+        # Define embedding model
+        set_embed_model(model_name=embedding_model)
+
+        # Set vector DB
+        documents = SimpleDirectoryReader(docs_path).load_data()
+        index = VectorStoreIndex.from_documents(documents)
+        retriever = VectorIndexRetriever(
+            index=index,
+            similarity_top_k=top_k,
+        )
+
+        self.top_k = top_k
+        self.query_engine = RetrieverQueryEngine(
+            retriever=retriever,
+            node_postprocessors=[SimilarityPostprocessor(similarity_cutoff=similarity_cutoff)]
+        )
+
+    def format_context(self, response):
+        context = "Context:\n"
+        for i in range(self.top_k):
+            context += response.source_nodes[i].text + "\n\n"
+        return context
+
+    def query(self, query: str):
+        try:
+            response = self.query_engine.query(query)
+            context = self.format_context(response)
+            return context
+        except:
+            return ""
+
+    def prompt(self, context: str, instruction: str):
+        return f"{context}\n ### Instruksi:\n {instruction}"
+
+    def main(self, instruction: str):
+        context = self.query(query=instruction)
+        prompt = self.prompt(context=context, instruction=instruction)
+        # print(prompt)
+        return prompt

app.py

@@ -0,0 +1,76 @@
+import gradio as gr
+import ollama
+# import asyncio
+# from openai import AsyncOpenAI
+from RAGModule import RAGModule
+
+# Instantiate the RAG module
+ollama.pull("MarcoAland/llama3.1-rag.indo")
+RAG_Trwira = RAGModule()
+
+# # Configure the async OpenAI client
+# client = AsyncOpenAI(api_key="34.69.9.203", base_url="http://34.69.9.203:11434/v1")
+
+# settings = {
+#     "model": "MarcoAland/llama3.1-rag-indo",
+#     "temperature": 0.3,
+#     "max_tokens": 2048,
+# }
+
+# async def generate_response(user_input: str) -> str:
+#     message = "Namamu adalah Mitrakara.\n\n" + user_input
+#     # Call documents options or not
+#     if "dokumen" in message.lower() or "document" in message.lower() or "documents" in message.lower():
+#         prompt = RAG_Trwira.main(message[10:])
+#     else:
+#         prompt = message
+        
+#     # Format the messages as a list of message dictionaries
+#     message_formated = [
+#         {"role": "user", "content": prompt}
+#     ]
+
+#     # Use streaming to handle partial responses
+#     stream = await client.chat.completions.create(messages=message_formated, stream=True, **settings)
+    
+#     response = ""
+#     async for part in stream:
+#         if token := part.choices[0].delta.content or "":
+#             response += token
+
+#     return response
+
+# def chat(user_input: str):
+#     # Call the asynchronous response generation function
+#     response = asyncio.run(generate_response(user_input))
+#     return response
+
+def chat(message: str, chat_history: str):
+    if "dokumen" in message.lower() or "document" in message.lower() or "documents" in message.lower():
+        prompt = RAG_Trwira.main(message[10:])
+    else:
+        prompt = message
+
+    stream = ollama.chat(
+                            model='MarcoAland/llama3.1-rag-indo',
+                            messages=[{'role': 'user', 'content': prompt}],
+                            stream=True,
+                        )
+
+    response_text = ''
+    for chunk in stream:
+        response_text += chunk['message']['content']
+        yield response_text
+
+# Define the Gradio interface
+iface = gr.Interface(
+    fn=chat,
+    inputs=gr.Textbox(label="Masukkan pertanyaan anda", placeholder="Tanyakan saja padaku🌟"),
+    outputs=gr.Textbox(label="Respons Mitrakara"),
+    title="Hai, namaku Mitrakara. Selamat datang!👋",
+    description="Berikut adalah beberapa tips untuk bertanya denganku✨✨✨\n1. Gunakan kata 'document:' jika ingin bertanya mengenai dokumen/administrasi perusahaan.\n2. Gunakan kalimat tanya yang baik.\n3. Enjoy the conversation.😊"
+)
+
+# Launch the Gradio interface
+if __name__ == "__main__":
+    iface.launch(share=False)

requirements.txt

@@ -0,0 +1,13 @@
+openai
+gradio
+ollama
+llama-cloud==0.0.13
+llama-index==0.10.64
+llama-index-embeddings-huggingface==0.2.3
+llama-index-cli==0.1.13
+llama-index-core==0.10.64
+llama-index-legacy==0.9.48
+llama-index-llms-openai==0.1.29
+llama-index-readers-file==0.1.33
+llama-index-readers-llama-parse==0.1.6
+llama-parse==0.4.9