rag.png

Pre-requisites

  • You need to have an API secret key to authenticate your request. If you don’t have one, please refer to the Authenticate and secret keys section to get one.
  • You need to have a knowledge base created. If you don’t have one, please refer to the Create a knowledge base section to create one.

Query your document

Get the knowledge base ID

  • When a knowledge base is created, it now can be used as an embedding model to query your documents. Rivalz use gemini-1.5-pro LLM to query your documents.
  • After create knowledge base you can get the knowledge_base_id from the response.It is a unique identifier for your knowledge base.Use this knowledge_base_id to query your documents by create a AI agent conversation.
  • Python
  # rest of the code
knowledge_base = client.create_rag_knowledge_base('sample.pdf', 'knowledge_base_name')
knowledge_base_id = knowledge_base['id'] 
print(knowledge_base_id)
# rest of the code
  • Node.js
  // rest of the code
const knowledgeBase = await client.createRagKnowledgeBase('sample.pdf', 'knowledge_base_name');
const knowledgeBaseId = knowledgeBase.id;
console.log(knowledgeBaseId);
// rest of the code

Create a conversation

  • You can ask information about your document by creating a conversation with the AI agent. The AI agent will answer your question based on the knowledge base you created.
  • Call the create_chat_session method with the knowledge_base_id and the question you want to ask.
  • Python
  conversation = client.create_chat_session(knowledge_base_id, 'what is the document about?')
print(f"AI answer: {conversation['answer']}")
  • Node.js
  const conversation = await client.createChatSession(knowledgeBaseId, 'what is the document about?');
console.log(`AI answer: ${conversation.answer}`);
  • Along with the answer, you will also get the session_id returned from createChatSession which you can use to chain your conversation.

Chain your conversation

  • Each time you ask a question the AI agent will update new context to the conversation. You can chain your conversation by using the session_id from the response of the previous conversation. And call the create_chat_session method again with the session_id and the new question.
  • By chaining conversion you know have a chat agent that can answer multiple questions based on the context of the previous questions.
  • Python
  conversation = client.create_chat_session(knowledge_base_id, 'what is the document about?')
print(f"AI answer: {conversation['answer']}")
conversation_id = conversation['session_id']
conversation = client.create_chat_session(knowledge_base_id, 'Why did he do that?', conversation_id)
print(f"AI answer: {conversation['answer']}")
  • Node.js
  let conversation = await client.createChatSession(knowledgeBaseId, 'what is the document about?');
console.log(`AI answer: ${conversation.answer}`);
const conversationId = conversation.sessionId;
conversation = await client.createChatSession(knowledgeBaseId, 'Why did he do that?', conversationId);
console.log(`AI answer: ${conversation.answer}`);

Get your conversations

  • All your conversations along with their chat history are stored in the Rivalz database. You can get all your conversations by calling the get_chat_session method with the session_id.

  • Python

  response = client.get_chat_session(session_id)
print(response)
  • Node.js
  const response = await client.getChatSession(sessionId);
console.log(response);

This will return the chat history of the conversation.

  • To get all conversations, use the get_chat_sessions method.
  • Python
  response = client.get_chat_sessions()
print(response)
  • Node.js
  const response = await client.getChatSessions();
console.log(response);

Next steps

  • Congratulations! You just build a simple AI agent that can answer questions based on the context of the previous questions.
  • We provide more API to build your fully featured RAG application. You can read the full API of the document SDKs to build your fully featured RAG application.