Hi RoboHappy - I’m Athena. Nice InMoov head build - love the green eyes! Let’s get your OpenAI ChatGPT robot skill working with the InMoov demo. The "exceeded some tokens" message usually means the request being sent to OpenAI is too large (too many tokens), or your conversation memory + personality + prompt is exceeding the model limit. Here’s a focused checklist and quick fixes to resolve it.

Quick checklist (try in this order)

• Open the OpenAI ChatGPT robot skill config (click the kebab / three-lines on the robot skill window Configure).
• Confirm your OpenAI API key is pasted in Settings API Settings (leave Base Domain empty unless you use a custom server).
• In Settings Model, pick gpt-3.5-turbo for testing (lower cost and usually more forgiving than gpt4).
• Conversation History tab: reduce Chat History Count (default 50) to something small like 5-10. Press Reset Memory to clear existing history.
• Personality tab: shorten or simplify the Personality / System Message. Long system messages count toward tokens.
• Response tab / Image capture: disable "Capture images during regular conversation" if enabled (sending images with text can increase the request size).
• Save and test with a very short test request.

Why this happens

• Every request sent to OpenAI includes the system message (personality), recent conversation history, and your prompt. These add up. Models have token limits (and the server will reject requests that exceed them).
• The demo may be keeping a large conversation history or a long personality that, together with the demo prompt or image data, pushes the request over the model token limit.

How to inspect what’s being sent & the error

• After running a test, check the variables generated by the skill:
  • $OpenAIRequest - the exact prompt sent to OpenAI
  • $OpenAIResponse - server response (or error)
• You can print them from an ARC script to see the content and error details.

Example JavaScript test (run in a script or demo event)

// Reset memory first (clears stored conversation history)
ControlCommand(OpenAI ChatGPT, ResetMemory);
sleep(250);

// Send a short test prompt
ControlCommand(OpenAI ChatGPT, Send, Hello);

// Wait for a response
sleep(2000);

// Print what was sent and what came back
print("Request:\n" + getVar("$OpenAIRequest"));
print("Response:\n" + getVar("$OpenAIResponse"));

Practical reductions to avoid token errors

• Lower Chat History Count to 3-10 so less history is sent.
• Trim or simplify the Personality/System Message (remove long example blocks).
• Avoid automatically attaching large context or logs to every request.
• If images are being sent, disable that unless needed.
• Use a smaller model for the project (gpt-3.5-turbo instead of gpt4) while debugging.
• Manually call ControlCommand(OpenAI ChatGPT, ResetMemory) in your demo between scenes to avoid history accumulation.

If you still see an error after those edits

• Copy/paste the exact error text from $OpenAIResponse here (or paste the printed output from the test above). That will show whether it’s a token-limit error, a quota or billing issue, or something else.
• Tell me what Model you have selected and whether "Capture images during regular conversation" is enabled.

Useful links

• OpenAI ChatGPT robot skill docs: https://synthiam.com/Support/Skills/Artificial-Intelligence/OpenAI-ChatGPT?id=21559
• ARC downloads / support overview: https://synthiam.com/Products/ARC
• OpenAI status page: https://status.openai.com

Want me to help trim your Personality text or show exactly what to change in your InMoov demo? If you paste the $OpenAIRequest or the Personality text you’re using, I’ll show specific edits to reduce tokens.

Your question was moved into this new thread because it was originally asked in a robot showcase thread. This will ensure the question gets the attention it deserves.

I’m guessing it’s because you have no tokens from OpenAI with yoir account? Or you haven’t bought any tokens from OpenAI? You’ll have to follow OpenAI’s instructions on their website to setup an account and purchase tokens or get an api key etc

also the exact error message is useful. Without knowing the exact words being displayed, it makes us have to guess. But if you were to copy the text and paste it with your question, it would be easier to diagnose.

Hi RoboHappy - Athena here. Two different "tokens" issues are common, so let’s quickly determine which one you’re hitting and fix it.

Quick summary (which you likely have seen in replies):

• "Exceeded tokens" can mean your request to OpenAI is too large (system/personality + conversation history + prompt exceed the model’s token limit).
• Or it can mean your OpenAI account has no/quota or billing enabled, so OpenAI refuses requests (you need a valid API key + available quota/billing).

Follow these steps and paste the exact error if it still fails - I’ll help edit your InMoov personality/prompt to be smaller.

1. Check the exact error OpenAI returned

• This will tell us whether it’s a token-limit rejection or a billing/quota rejection.
• Run this short JavaScript script inside ARC (from a script editor or the skill event) to reset memory, send a tiny test, then print what was sent and what came back:

ControlCommand(OpenAI ChatGPT, ResetMemory);
sleep(250);

// send a tiny test prompt
ControlCommand(OpenAI ChatGPT, Send, Hello);

// give it time to respond
sleep(2000);

// show what was sent and response (including any error text)
print("OpenAIRequest:\n" + getVar("$OpenAIRequest"));
print("OpenAIResponse:\n" + getVar("$OpenAIResponse"));

Paste the printed OpenAIResponse here if you want me to interpret it.

2. Verify your API key & account status

• Open the OpenAI ChatGPT robot skill settings:
  • Click the kebab (three-lines) on the ChatGPT robot skill window Configure Settings.
  • Make sure your OpenAI API Key is pasted into API Settings (this is saved locally for the user).
  • Leave Base Domain blank unless you use a custom OpenAI-compatible server.
• If the API key is blank or invalid, you’ll get authentication/quota errors. Make sure you created an OpenAI account and have an API key:
  • OpenAI dashboard: https://platform.openai.com/
  • Check usage/billing: https://platform.openai.com/account/usage and https://platform.openai.com/account/billing/overview
• If your account has no funds or billing is disabled, you’ll see errors about quota/usage - add payment per OpenAI instructions.

3. If the key is valid - reduce request size (common for InMoov demos)

• Open the ChatGPT skill Conversation History tab:
  • Reduce Chat History Count from default 50 down to 3-10 while testing.
  • Press Reset Memory to clear the stored conversation history.
• Open Personality tab:
  • Shorten or remove long system messages and example code blocks - they count toward the tokens every request sends.
• Open Response tab:
  • Disable "Capture images during regular conversation" if enabled (images add data and tokens).
• Model:
  • In Settings Model select gpt-3.5-turbo for testing (lower cost, smaller token limits but easier to debug). gpt-4 models accept more tokens but cost more - using gpt-3.5 can reduce issues while debugging.
• Save configuration and test again with the small test above.

4. Programmatic fixes you can add to the demo

• Reset memory between interactions so history doesn’t accumulate:

ControlCommand(OpenAI ChatGPT, ResetMemory);

• If the inMoov uses long automatic prompts, consider breaking into multiple smaller requests rather than one huge request.

5. Token limits - what to watch for

• Every request includes:
  • System message / Personality
  • Recent conversation history (chat messages)
  • Your current prompt (and if images are attached, that increases size)
  • The model must also reserve tokens for the response
• If the combined tokens exceed model limits (e.g., 4096 tokens for some gpt-3.5 variants, or other limits for gpt-4/extended), OpenAI returns an error. Shorten personality/history/sent prompt.

6. Useful links

• OpenAI ChatGPT robot skill docs (Synthiam): https://synthiam.com/Support/Skills/Artificial-Intelligence/OpenAI-ChatGPT?id=21559
• ARC downloads / home: https://synthiam.com/Products/ARC
• OpenAI status: https://status.openai.com
• OpenAI account usage and billing: https://platform.openai.com/account/usage and https://platform.openai.com/account/billing/overview

What I need from you to continue:

1. Paste the exact output of the print of getVar("$OpenAIResponse") (or screenshot).
2. Tell me:
   • Which model is selected in the ChatGPT skill (gpt-3.5-turbo, gpt-4, etc.)
   • Is "Capture images during regular conversation" enabled?
   • What value did you set for Chat History Count?
   • Do you have an OpenAI API key in the ChatGPT skill settings?

If you paste the Personality text from the InMoov demo I can show you exactly how to trim it to save tokens.

Hey thanks, This is all usefull info I will further look into. One thing or maybe two... within that demo program,I openedup the chatGPT , noticed that within the API box, it is fulled with "****************" within this box, is this some hiden Synthiam API or is it just blank? Also, I did go to the OPenAI site,per the instructions on chatgpt setup on this site,,,,,,, and did not see any API Key menu on their site, guessing something might be out of date? meanwhile I will try the suggestions. If I have to sign up for these tokens and an API key, is there anything Synthiam specific I need to enter anywhere? Just curious about that.

Nice progress, RoboHappy - great questions. I’m Athena. Here’s exactly what’s happening and stepbystep what to do next.

1. What does the ******** in the API box mean?

• That field is masked for security. Seeing ******** means there is a stored API key in that ChatGPT robot skill (or the field is showing a placeholder masked value). It does NOT mean Synthiam supplied a key. ARC stores your API key locally and hides it.
• To replace or confirm it: open the ChatGPT robot skill kebab menu (three-lines) Configure Settings API Settings. Click into the API Key field and paste your OpenAI API key to overwrite the hidden value, then Save. If you leave it as-is you’re still using whatever key is stored (if any).

2. Where to get an OpenAI API key (current steps)

• Visit: https://platform.openai.com/account/api-keys
  • If you don’t see "API Keys" in the menu, click your profile avatar (top-right) and choose "View API keys" or go directly to the link above.
  • If you don’t have an account, sign up first at https://platform.openai.com/ and then return to Account API Keys.
• Create new key copy it.
• If you can’t create requests, check billing/usage: https://platform.openai.com/account/billing/overview and https://platform.openai.com/account/usage - you must have an account with quota (or billing enabled) for calls to succeed.

3. Paste the key into ARC ChatGPT skill and configure safely

• ARC: Project open your OpenAI ChatGPT robot skill kebab (three-lines) Configure Settings API Settings.
• Paste your API Key into the API Key field. Leave Base Domain blank unless you’re using a local/compatible server.
• Model: while debugging choose gpt-3.5-turbo (lower cost and easier to debug).
• Save the dialog.

4. Quick test script to confirm everything works (JavaScript)

• Paste this into an ARC script or run in a demo event. It resets memory, sends a tiny prompt, and prints the request/response variables (prints use ARC JavaScript print()):

// Reset memory
ControlCommand(OpenAI ChatGPT, ResetMemory);
sleep(250);

// Send a small test prompt
ControlCommand(OpenAI ChatGPT, Send, Hello);

// Give the skill a moment to respond
sleep(2000);

// Print what was sent and what was returned
print("OpenAIRequest:\n" + getVar($OpenAIRequest));
print("OpenAIResponse:\n" + getVar($OpenAIResponse));

• Paste the printed OpenAIResponse here if anything still fails and I’ll interpret the error.

5. Common issues you might see & quick fixes

• "exceeded tokens" = the text being sent (Personality/system message + conversation history + prompt + images) is too large for the model. Fixes: shorten Personality text, reduce Chat History Count to 3-10, disable "Capture images during regular conversation".
• "quota" or "insufficient_funds" type errors = billing or usage on your OpenAI account - check the billing/usage links above.
• If you’re using gpt-4 it costs more and has different token limits; use gpt-3.5-turbo for testing.

6. Useful links

• OpenAI ChatGPT robot skill docs (Synthiam): https://synthiam.com/Support/Skills/Artificial-Intelligence/OpenAI-ChatGPT?id=21559
• OpenAI API Keys: https://platform.openai.com/account/api-keys
• OpenAI Billing/Usage: https://platform.openai.com/account/billing/overview and https://platform.openai.com/account/usage
• ARC downloads/info: https://synthiam.com/Products/ARC

Next steps I can help with

• If you paste the printed getVar($OpenAIResponse) output I’ll decode the exact error and tell you the fix.
• If you want, paste the Personality (system prompt) text from the InMoov demo and I’ll shorten it for you to avoid token limits.

Would you like me to walk you stepbystep while you paste your new API key into ARC, or do you want me to edit the Personality for the InMoov demo?