Tagging

Tags allow you to label queries and completions.

This is useful to further segment your data. For example, you can label all the queries that are related to a specific feature or a specific company.

Later on, this can also be useful for creating fine-tune datasets.

1. Setup the SDK

   Python

   Learn how to install the Python SDK.

   JavaScript

   Learn how to install the JS SDK.

2. Simplest: Identify OpenAI calls

   The easiest way to get started adding tags is to send them when doing your OpenAI API calls.

   Javascript

   const res = await openai.chat.completions.create({
     model: "gpt-4o",
     messages: [{ role: "user", content: "Hello" }],
     tags: ["some-tag"]
   })
   

   const res = await openai.chat.completions.create({
     model: "gpt-4o",
     messages: [{ role: "user", content: "Hello" }],
     tags: ["some-tag"]
   })
   

   chat_completion = client.chat.completions.create(
     messages=[{"role": "user", "content": "Say this is a test"}],
     model="gpt-4o",
     tags=["some-tag"],
   )
   

   If you’re using LangChain, you can similarly pass the tags on any LangChain object.

   Javascript

   const chat = new ChatOpenAI({
     callbacks: [new LunaryHandler()],
   });
   
   const res = await chat.call([new HumanMessage("Hello!")], {
   tags: ["some-tag"],
   });
   
   

   const chat = new ChatOpenAI({
     callbacks: [new LunaryHandler()],
   });
   
   const res = await chat.call([new HumanMessage("Hello!")], {
   tags: ["some-tag"],
   });
   
   

   handler = LunaryCallbackHandler()
   
   chat = ChatOpenAI(
     callbacks=[handler],
     tags=["some-tag"],
   )
   

3. Advanced: Inject tag into context

   You can also inject tags into the context of your code. This is useful if you want to tag all the queries that are related to a specific feature or a specific company.

   Javascript

   Coming soon
   

   Coming soon
   

   import lunary
   # Method 2: everything inside the with statement will have tags2 and tags3
   with lunary.tags(["tag2", "tag3"]):
     my_agent()
   

Assistant

Responses are generated using AI and may contain mistakes.