Bug: Wrong token count for attachments

[e-ndorfin]

When running /tokens the token count is inaccurate. This is also true for the output following normal messages

[e-ndorfin]

Also missing for thinking. I think we can use data returned by openrouter rather than tokenizing with tiktoken (as right now it's still inaccurate)

Example from https://openrouter.ai/docs/api-reference/chat-completion

{
  "id": "gen-12345",
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "The meaning of life is a complex and subjective question...",
        "refusal": ""
      },
      "logprobs": {},
      "finish_reason": "stop",
      "index": 0
    }
  ],
  "provider": "OpenAI",
  "model": "openai/gpt-3.5-turbo",
  "object": "chat.completion",
  "created": 1735317796,
  "system_fingerprint": {},
  "usage": {
    "prompt_tokens": 14,
    "completion_tokens": 163,
    "total_tokens": 177
  }
}

[e-ndorfin]

This is also missing for the system prompt which is not calculated in the tokens

[oop7]

So what do you think, the best approach to fix it

[e-ndorfin]

{
"provider": "OpenAI",
"model": "openai/gpt-3.5-turbo",
"object": "chat.completion",
"created": 1735317796,
"system_fingerprint": {},
"usage": {
"prompt_tokens": 14,
"completion_tokens": 163,
"total_tokens": 177
}
}

OpenRouter gives direct token counts - we can use this instead of calculating ourselves (I referenced this above)