OpenAI API Error context_length_exceeded: Maximum context length exceeded

Identify: does your log match?

openai.BadRequestError: Error code: 400 - {'error': {'message': "This model's maximum context length is 128000 tokens. However, your messages resulted in 131072 tokens. Please reduce the length of the messages.", 'type': 'invalid_request_error', 'param': 'messages', 'code': 'context_length_exceeded'}}

1. High

   Chat history grows without a cap

   Every turn is resent, so a long conversation eventually exceeds the window.

2. Medium

   Retrieved documents stuffed in whole

   RAG pipelines paste entire files instead of the top-k relevant chunks.

3. Low

   Large tool schemas

   Dozens of function definitions count toward input tokens on every request.

Fix

resp = client.chat.completions.create(model="gpt-4o", messages=history)

import tiktoken
enc = tiktoken.encoding_for_model("gpt-4o")

def fit(history, budget=120_000):
    system, rest = history[0], history[1:]
    keep, used = [], 0
    for m in reversed(rest):              # newest first
        n = len(enc.encode(m["content"])) + 4
        if used + n > budget: break
        keep.append(m); used += n
    return [system] + keep[::-1]          # always keep the system prompt

resp = client.chat.completions.create(model="gpt-4o", messages=fit(history))

Verify: reproduce it on purpose

$ python3 -c "from openai import OpenAI; OpenAI().chat.completions.create(model='gpt-4o-mini', messages=[{'role':'user','content':'word '*200000}])"

Prevent

• Log token counts per request so growth shows up before it fails.
• Send top-k retrieved chunks, not whole documents.
• Only include the tools the current step can use.