A bit confused about limits of the context window

[jchudzynski]

I have a txt file with dates and events. It's rather longer, 6,717 tokens to be exact long.

Since it's a bit bigger file I decided to add it to a vector database to better deal with the context limits. In the example below I load it from just a variable, but I was getting the same results reading it from a txt or a json file

data= [{"start":3,"end":5,"name"="event1"},{"start":9,"end":15,"name"="event2"}]
dataString = str(data)
text_splitter = CharacterTextSplitter(chunk_size=300, chunk_overlap=100)
docs = text_splitter.split_text(dataString)

embeddings = OpenAIEmbeddings()
vectordb = FAISS.from_texts(docs, embeddings)

qa_chain = RetrievalQA.from_chain_type(
       llm=ChatOpenAI(model_name="gpt-3.5-turbo-0613", temperature=0),
       chain_type="stuff",
       retriever=vectordb.as_retriever())
   
query = " Given a data with timestamps of events, which ones do you think were the most important ones"
print(qa_chain.run(query))

This request results in a following error:

InvalidRequestError: This model's maximum context length is 4097 tokens. However, your messages resulted in 7004 tokens. Please reduce the length of the messages.

I am bit confused why. I thought the vector databased prevent this issue.
Should I chunk the data differently? Use a different prompt?
Thank you for help :)