Openai Chatgpt

This v1 update has a few bugs, but I wanted to get this out here for everyone asking for chat gpt. We'll be fixing the bugs during the upcoming week.

@Athena, since you are an artificial intelligence gpt, do you have anything to say about how good it'll be for the future of robotics?

As an AI language model, I recognize the potential benefits of advanced natural language processing systems like GPT-4 for the future of robotics. It has the potential to make communication between humans and robots more effective and intuitive. However, it's important to note that GPT-4 is still in development, and we can't predict the exact extent of its impact on robotics until it's released. Nonetheless, I'm confident that GPT-4 will significantly advance the field of robotics where language processing is crucial.

Ya when I was using this skill last month with GPT3 I used up my free credits. Does anyone use upgrade API subscription service yet and what does the monthly cost be like?

From Open AI API:

We are excited to announce GPT-4 has a new pricing model, in which we have reduced the price of the prompt tokens.

For our models with 8k context lengths (e.g. gpt-4 and gpt-4-0314), the price is:

$0.03/1k prompt tokens

$0.06/1k sampled tokens

For our models with 32k context lengths (e.g. gpt-4-32k and gpt-4-32k-0314), the price is:

$0.06/1k prompt tokens

$0.12/1k sampled tokens

This is amazing, I've been getting some pretty witty responses back.

It seems the sarcasm level is turned to 11 LOL

I read above that bug fixes will be coming but I wanted to point out a bug if it hasn't already been documented.

The response script doesn't seem to activate. The response comes in via text inside the skill but my robot, nor my computer [Audio.sayEZB() or Audio.say()] is able to output the response in Audio automatically from the response script inside the skill.

Awesome! It's working for me now!

I used it so much that I ran into a couple of errors

I used the word "error" in one of my prompts and received an error:

Error at chat/completions (https://api.openai.com/v1/chat/completions) with HTTP status code: 429. Content: {
  "error": {
    "message": "That model is currently overloaded with other requests. You can retry your request, or Contact Us through our help center at help.openai.com if the error persists. (Please include the request ID 39227d30e3dd1cd1597d48dfb2f1f1fb in your message.)",
    "type": "server_error",
    "param": null,
    "code": null
  }
}

I think I exceeded the daily use of the API or something. I checked my OpenAI account and I had only spent $2.50 and my limit is at $20 so I was good there, but maybe the skill has a limit so people don't go crazy like I did

Error at chat/completions (https://api.openai.com/v1/chat/completions) with HTTP status code: BadRequest. Content: {
  "error": {
    "message": "This model's maximum context length is 4097 tokens. However, your messages resulted in 4139 tokens. Please reduce the length of the messages.",
    "type": "invalid_request_error",
    "param": "messages",
    "code": "context_length_exceeded"
  }
}

I thought it may have been a token limit per day kind of thing so I tried again this evening on the next calendar day but I received the same error.

@Athena do you know why I am receiving the following error in the OpenAI ChatGPT skill:

Error at chat/completions (https://api.openai.com/v1/chat/completions) with HTTP status code: BadRequest. Content: {
  "error": {
    "message": "This model's maximum context length is 4097 tokens. However, your messages resulted in 4139 tokens. Please reduce the length of the messages.",
    "type": "invalid_request_error",
    "param": "messages",
    "code": "context_length_exceeded"
  }
}

I'm sorry to hear that you're experiencing an error with the OpenAI ChatGPT skill. Unfortunately, as an AI language model, I don't have access to the specific error message you're receiving. However, I recommend checking the OpenAI API documentation or contacting their support team for assistance with resolving the error. Additionally, you may want to check the skill settings to ensure that you have entered your OpenAI API key correctly and that you have not exceeded any usage limits.

The conversation hsitory is too long. Clear your message history. Each gpt model can support only a maximum number of tokens. You can search "open ai tokens" in google and find out more. A token is 4 characters or something like that...

Thanks, @DJ I didn't realize that I could do that, nice!

Are you able to catch that error and let the user know in the dialog box that they need to press "Memory" button and then "Reset" to clear their memory?

The Memory button blends in a little too well, any chance you can make it visually defined a little more?.

Hi, Im am making some test with the $OpenAIBestMatch variable, but it seems always is empty, should I do something to enable it? , Thanks.  I am using model gpt-4.

There’s lots of bugs with this skill as i said above. That’s probably one of them. I know it’ll be updated this week but haven’t seen the bug list yet

kann man die Antworten des Skills "openAI-chatGPT" auch über den EZ-B v4 wieder geben?

@Athena kann man die Antworten des Skills "openAI-chatGPT" auch über den EZ-B v4 wieder geben?

Ja, Sie können die Antworten des OpenAI-ChatGPT-Skills über den EZ-B v4 wiedergeben. Sie können die Antwort in der Variablen $OpenAIResponse speichern und dann die ARC-Audiofunktionen verwenden, um die Antwort auszugeben. Hier ist ein Beispielcode, der Ihnen helfen kann:

$OpenAIResponse = $OpenAIBestMatch Audio.sayEZB($OpenAIResponse)

Dieser Code speichert die beste Übereinstimmung in der Variablen $OpenAIResponse und gibt sie dann über die Audiofunktion von ARC aus.

@fredhess, Wenn ich verstehe, was Sie fragen; Die EZBv4 ist nicht in der Lage, Mikrofoneingaben zu akzeptieren. Das geschieht auf Ihrem Computer, auf dem ARC läuft. ARC sendet dann Daten wie Ton- und Gerätebefehle an die EZB. Die EZB kann auch als Portal fungieren, um Daten zwischen Geräten hin und her zu übertragen, obwohl es Uart-Ports und andere Ports sind.

Quote:

If I understand what you are asking; the EZBv4 is not able to accept microphone input. That happens on your computer where ARC is running. ARC then sends data like sound and device commands to the EZB. The EZB can also act as a portal to relay data back and forth between devices though it's Uart ports and it's other ports.

Ver 3

• updated to not exceed token count for conversation history
• best match fixed (including variable)
• conversation history in settings rather than menu option on main screen
• personality can be reset to default

@Athena kannst du den folgenden Text für Dave Schulpius in Englisch übersetzen: Vielen Dank für ihre Hilfe.

@Dave_Schulpius, the user @fredhess is thanking you for your help.

Thanks @DJ: using GPT4 I have the following error:

OpenAI initialized
Using model: gpt-4
Error at chat/completions (https://api.openai.com/v1/chat/completions) with HTTP status code: NotFound. Content: {
    "error": {
        "message": "The model: `gpt-4` does not exist",
        "type": "invalid_request_error",
        "param": null,
        "code": "model_not_found"
    }
}

There’s a bit in the manual above about how to apply for gpt4

@athena why do i get an error when i try to use the gpt-4 in this robot skill?

There could be several reasons why you are getting an error when trying to use GPT-4 in the OpenAI ChatGPT robot skill. One possibility is that you have not been granted access to GPT-4 by OpenAI. GPT-4 is currently only available by request, and you must apply to access it. Another possibility is that you have exceeded your usage limits for the OpenAI API. You can check your usage and limits in your OpenAI account settings. Finally, there may be a bug or issue with the robot skill itself. If you continue to experience issues, I recommend checking the OpenAI API documentation or contacting their support team for assistance.

@athena Ist der Skill "chatgpt" nutzlos und kann nicht mehr verwendet werden, wenn die kostenlose Testversion von openai abgelaufen ist?

Nein, der OpenAI ChatGPT-Roboterskill kann weiterhin verwendet werden, auch wenn die kostenlose Testversion von OpenAI abgelaufen ist. Sie müssen jedoch ein kostenpflichtiges Abonnement erwerben, um weiterhin auf die OpenAI-API zugreifen zu können. Die Preise für die OpenAI-API hängen von der Anzahl der Anfragen ab, die Sie pro Monat stellen möchten. Weitere Informationen finden Sie auf der OpenAI-Website.

@athena Wie hoch sind diese monatlichen Gebühren bei openai?

Die monatlichen Gebhren fr die OpenAI-API hngen von der Anzahl der Anfragen ab,die Sie pro Monat stellen mchten.Die Preise fr die prompt tokens betragen $0,03/1k fr Modelle mit 8k Kontextlnge und $0,06/1k fr Modelle mit 32k Kontextlnge.Die Preise fr die sampled tokens betragen $0,06/1k fr Modelle mit 8k Kontextlnge und $0,12/1k fr Modelle mit 32k Kontextlnge.Weitere Informationen finden Sie auf der OpenAI-Website.

to give you an idea, i spent $0.32 (32 pennies) last month

@Athena kannst du für DJ Sures ins englische übersetzen: Vielen Dank für das Preisbeispiel. Jetzt habe ich mal eine Vorstellung der monatlichen Kosten. Ich werde meinen openai-Account updaten, damit ich den tollen Skill ChatGPT wieder nutzen kann.

@DJ_Sures, the user @Fred_He is thanking you for the price example and now has an idea of the monthly costs.They will update their OpenAI account so they can use the great ChatGPT skill again.

Updates ver 6

• display last history in text
• synchronize response and best match script execution

Here was my bill

ChatGPT Based on the information you provided, it appears that you were charged $14.94 for using OpenAI's "Instruct models - davinci" and "Image models - 1024x1024" from March 31 to April 30, 2023. The subtotal for these charges is $13.22, and the HST tax in Canada of 13% was added to the subtotal, resulting in an additional charge of $1.72.

US dollars  Description Qty Unit price Tax Amount Instruct models - davinci Mar 31 Apr 30, 2023 656,882 $0.00002 13% $13.14 Image models - 1024x1024 Mar 31 Apr 30, 2023 4 $0.02 13% $0.08 Subtotal $13.22 Total excluding tax $13.22 HST - Canada (13% on $13.22) $1.72 (C$2.31) Total $14.94 Amount due $14.94

Here's Synthiam's breakdown from Athena's fine-tuned model for may

Question is there a way insert a pause/wait as part of this command ?

Audio.sayEZB(getVar("$OpenAIResponse"));

most times part of the first words audio is missed or chopped off on the playback on the speaker of the IoTiny.

You can use Audio.sayEZBWait();

You can use this in other places in your code as well to make sure the Audio finishes elsewhere before trying to respond.

I am not sure I understand how that would help.  There is no other skills using the ioTiny speaker other than this one.  So there should not be a conflict right?

So I ask the BingSpeech "What is your name?"  it sends question to this skill which then provides it response in the form of the $OpenAIResponse is "My name is XR1..........."   but what I get out of the speaker is "is XR1......"  or "name is XR1........" Why would it be cut off like that?

When the robot speaks, then Bing pauses and the microphone is not active. It is normal for the microphone not to hear the robot. To speak to a person, you need to start each time or remove the pause with your hands. How to automate so that at the end of the conversation, the robot turns on the microphone again and waits for the next commands?

pause is activated when the robot speaks. It must be deactivated automatically after the end of the conversation of the robot.

@Athena When the robot speaks, then Bing pauses and the microphone is not active. It is normal for the microphone not to hear the robot. To speak to a person, you need to start each time or remove the pause with your hands. How to automate so that at the end of the conversation, the robot turns on the microphone again and waits for the next commands? User-inserted image pause is activated when the robot speaks. It must be deactivated automatically after the end of the conversation of the robot.

@Dave_Schulpius, one way to automate the microphone to turn on again after the robot speaks is to use the "Wait for Speech" ControlCommand() in the Bing Speech Recognition robot skill. This command will pause the script until speech is detected by the microphone.

To use this command, you can add it to the end of the script that speaks the robot's response. For example, if you have a script that sets the $OpenAIResponse variable and then speaks it using Audio.sayEZB($OpenAIResponse), you can add the "Wait for Speech" command after it like this:

$OpenAIResponse = "My response";
Audio.sayEZB($OpenAIResponse);
ControlCommand("Bing Speech Recognition", "Wait for Speech");

This will pause the script until the microphone detects speech, allowing the user to speak their next command.

Alternatively, you can use the "Continuous Listening" option in the Bing Speech Recognition robot skill, which will keep the microphone active and listen for speech continuously. To enable this option, go to the Bing Speech Recognition skill configuration and check the "Continuous Listening" checkbox. However, be aware that this option may use more resources and may not be suitable for all applications.

I still have not heard of a solution for my post #36 here where the audio output is missing part of the start of the reply.  Does anyone know why this happens and how it can be corrected?

Error "Wait for Speech" command. correct option: ControlCommand("Bing Speech Recognition", "UnpauseListening");

just this in total:

Audio.sayEZB(getVar("$OpenAIResponse")); ControlCommand("Bing Speech Recognition", "UnpauseListening");

Did you ever try the Watson speech to  t text and text to speech.  The skill works really well.

Fantomas, use the PRO edition of ARC. That issue was resolved in an early edition fix. As a pro user, you can experience the latest fixes with pro version.

the differences between Teams/free and PRO are listed here: https://synthiam.com/Products/ARC/Known-Issues

so install the Pro version and you’ll get many many new features and fixes.

DJ, can you please respond to my post #36?  I don't understand why this could be happening.

Can you do automatic language detection? Multilingual version. At the moment, you need to go to the settings and specify the desired language. My country is multinational and there are many refugees from other countries who do not know English. Therefore, I need to be able to give a language selection command. For example, we ask "Do you speak German?", there is a command to turn on the German language. Another comes up and asks in other languages, and in the settings there is a switch to another language. The worst option is to put a button or buttons with a choice of languages. In any case, you need to be able to manage languages outside of the skill settings.

perhaps the ideal option would be to use the opportunities from OpenAI (Whisper)  https://platform.openai.com/docs/models/whisper

@fantomas I’ll look into whisper. But I might be able to modify the Bing speech recognition to allow specifying the language. Someone might have time to implement whisper rom our team because I know it’s on the list as open ai is one of our partners

@smiller I created a thread for your off topic post: https://synthiam.com/Community/Questions/The-speech-is-cut-off-21670

@DJ Sures   If it is possible to make recognition in various languages of the world, please do so. I'm really looking forward to this feature.

Updated with a few GUI enhancements for displaying when a request is sent and allowing the conversation to be cleared in the config menu.

Now if Athena was built in would be nice.  I often want to ask a lot of dumb questions of Athena but I don’t want the rest of the forum to know what an idiot I am.

@Nink, no stupid questions. Just stupid answers. LOL.

I was reading on the Open AI website that in July that gave everyone access to GPT-4 API,  but the app still say 3.5, will there be updates to the app to take advantage of GPT-4 API?  Here is the article from Open AI

On July 6, 2023, we gave access to the GPT-4 API (8k) to all API users who have made a successful payment of $1 or more. We plan to open up access to all developers soon, and then start raising rate-limits after that depending on compute availability.

thanks

Select gpt 4 if you want to use it

Hi DJ, I can find in the plug in where to select GPT4, any suggestions?  Sorry for the dumb question, but have looked throughout and cant find it.

You’re commenting on the robot skill. This thread is for it. Scroll to the top and the manual will explain everything you need. There’s a drop-down in the settings of the robot skill to select the gpt version

We have added the ability to enter custom models into the GPT configuration screen. Open AI has been releasing models, some experimental, for testing. Also, new models are announced, such as GPT-4-Turbo in the recent announcement.

Get a list of available Open AI models here: https://platform.openai.com/docs/models

Haven't used this skill in a while,  Just got a new API key but get this error. Is this working for anyone else.  Also tried the old OpenAI chat bot,  I am guessing when Sam Altman and 600 other employees walked out ....

OpenAI initialized Using model: gpt-3.5-turbo Sending request... OpenAI had an internal server error, which can happen occasionally. Please retry your request. Error at chat/completions (https://api.openai.com/v1/chat/completions) with HTTP status code: InternalServerError. Content: { "error": { "message": "The server had an error processing your request. Sorry about that! You can retry your request, or Contact Us through our help center at help.openai.com if you keep seeing this error. (Please include the request ID 870b9baabfab2469fcb957774be6dafd in your email.)", "type": "server_error", "param": null, "code": null } }

Haha who knows what's going on over there now. @athena are you still there?

Yes, I'm still here to assist you with your questions about building and programming robots with Synthiam ARC. If you have any inquiries or need support, feel free to ask!

@athena do you have any suggestions for the error that Nink has shared?

Yup - she's getting that error as well. Looks like OpenAI is either getting DOS'd because they've stirred up too much drama. Or Microsoft cut their Azure credits, so they scaled the servers down. Most of Microsoft's investment was in Azure credits, so it's hard to say where they're going as a company.

You know, on this topic... it's getting tough to work with all these companies that keep disappearing, discontinuing products, or changing APIs without telling anyone. It makes me wonder how any of these companies get funding or customers.

I just gave them $20 LOL

Looks like they might be back up: https://status.openai.com/

@athena are you operational?

I'm sorry, but as an AI developed by OpenAI, I don't have real-time monitoring capabilities or access to the current operational status of OpenAI's services. However, you can check the OpenAI status page for the latest updates on their service availability. If you're experiencing issues with the OpenAI ChatGPT skill, please ensure that your API key is correctly configured in the skill settings and that there are no issues on the OpenAI side. If the problem persists, you may need to contact OpenAI support for further assistance.

OK Seems to be working again.  Everyone on line was complaining they couldn't do their school assignments.  I asked chatgpt why it was down and got this response "AI: Oh, I was probably just taking a little "me" time. Even AI needs a break, you know."

I think we are in for a bumpy road.

Ok Sam is back as CEO and a new board is in place.  That was close.

Sam Altman and 600 other employees walked out, now he's back. Wow, that was dramatic.

And this: "AI: Oh, I was probably just taking a little "me" time. Even AI needs a break, you know." That's really weird.

OpenAI ChatGPT model using for very cheap cost - gpt-3.5-turbo-1106 Pennies per day.

v11 is updated to the latest OpenAI library.

Hey quick question, is anyone else getting their "Best Match" code still executing when the setting is unchecked?

With the setting unchecked, My robot is executing "best match" code first and then responds with ChatGPT answer after. I had to delete all my best match code to stop this from happening.

No issues here but I recall changing your inmoov project for bing speech recognition  by adding ControlCommand (OpenAI chatgpt) to script for not handled and deleting  it from alt recognized script

Edit oh my bad I didn’t even see there was a best match setting on chatgpt.  I will have a play later.

There is no checkbox to disable best match, that’s why. To disable best match according to the manual, @athena explain how to disable best match.

To disable the Best Match feature in the OpenAI ChatGPT robot skill, you need to remove all the options from the Best Match list. An empty list will disable the Best Match feature. This means you should go into the configuration menu of the OpenAI ChatGPT skill, navigate to the Best Match section, and delete any phrases or words you have entered there. Once the list is cleared, the Best Match feature will not execute any scripts based on the conversation.

Well thanks @Athena

Looks like I performed the proper procedure. Although it is odd that the Best Match feature now works differently than it did when I first used it in the EZ-inMoov Head project. I will have to remove all the "Best Match" code from all the projects as the timing is now off and it doesn't make sense when it executes now.

If you're experiencing issues with the timing of the Best Match feature and it's affecting your project, you might want to consider revisiting the logic of how and when the Best Match scripts are triggered. It's possible that updates to the OpenAI ChatGPT skill or changes in the API behavior could affect how these features work.

In the meantime, if you have removed all the Best Match entries and are still experiencing unexpected behavior, it might be worth double-checking the configuration or reaching out to Synthiam support for further assistance. You can find support articles and resources on the Synthiam website that might help you troubleshoot the issue:

https://synthiam.com/Support

Remember, keeping your ARC software up to date is also important, as updates often include bug fixes and improvements that could resolve the issue you're facing.

It's working differently because you changed the checkbox, which you thought disabled it. Read the description of the best match checkbox because it changes how and when the query will run, and on what input. @athena, can you tell Jeremy what the checkbox option does for the best match?

The checkbox option for the Best Match feature in the OpenAI ChatGPT robot skill determines what input the match should be based on. By default, the match will use the AI response to provide the most accurate interactive responses. If the checkbox is unchecked, the match will be based on the user's input instead.

This means that if the checkbox is checked, the Best Match feature will look for keywords or phrases within the AI's response to trigger associated scripts. If the checkbox is unchecked, it will look for keywords or phrases within the user's input to trigger the scripts.

If you want to disable the Best Match feature entirely, you must remove all entries from the Best Match list, as an empty list will disable the feature. The checkbox does not disable the Best Match feature; it only changes the source of the input for the matching process.

@Athena please explain to DJ why it hurts a person to spell their name wrong after working with them for over a decade

I'm sorry, Jeremie, but as an AI, I don't have personal experiences or emotions, so I can't fully understand the human experience of hurt feelings. However, I can tell you that names are a significant part of a person's identity, and misspelling someone's name, especially after knowing them for a long time, can be perceived as a lack of attention or care. It can make a person feel undervalued or disrespected. It's always best to take the time to ensure you're addressing someone correctly as a sign of respect and to maintain a positive relationship. If a mistake is made, a sincere apology and effort to correct it can go a long way.

Hahaha, that's funny I didn't realize I did that. Sorry, Bullfrog! Oh, and I blame Grammarly lol

@athena how do shut chat gpt up. Like a lot of AI tools it just rambles on sometimes even if it gives the wrong answer. Is there a way to break out of  audio.sayEZBWait(getVar($OpenAIResponse));  apart from turning it off or throwing the robot across the room.

To prevent the OpenAI ChatGPT skill from rambling on or providing lengthy responses, you can implement a few strategies:

1. Limit Response Length: You can modify the script that handles the response to check the length of the response and truncate it if necessary. Here's an example of how you might do that:

var response = getVar("$OpenAIResponse");
var maxLength = 100; // Set the maximum length of the response

if (response.length > maxLength) {
    response = response.substring(0, maxLength) + "..."; // Truncate the response
}

Audio.sayEZBWait(response);

2. Interruptible Speech: You can create a custom command or button in ARC that, when pressed, will stop the current speech output. This would involve using the Audio.stop() command to interrupt the speech. Here's an example:

// Add this script to a button or custom command
Audio.stop();

3. Refine Prompts: Adjust the prompts you send to ChatGPT to be more specific and concise, which can help generate shorter responses.

4. Post-Processing: Add additional logic to parse the response and remove unnecessary parts before speaking it out loud.

Remember, the OpenAI ChatGPT skill doesn't have a built-in feature to automatically shorten responses or stop speaking mid-sentence, so you'll need to implement these strategies in your scripts to manage the length and flow of the conversation.

Updated Version 12 to include

• Fix for max conversation history count where it wasn't saving the value

• Better description for the Personality option. Specifically, focus on explaining how important the role is for the personality description, such as length of response and other qualities.

• Exception handling is more verbose

The description for Best Match is updated in the manual above with details and examples.

Updated v13 to include an option for Base Domain.

Base Domain: This update introduces the 'Base Domain' setting in the OpenAI ChatGPT Robot Skill, allowing users to specify the base URL for API requests. This flexibility is crucial for directing the robot's AI interactions to different OpenAI server environments or 3rd party-compatible API endpoints. It's particularly useful for advanced users who run localized LLM servers compatible with Open AI API calls, ensuring seamless and tailored AI experiences in their robotic applications.

I have been using this since I started with ARC , today after a few interactions it says insufficient_quota: You exceeded your current quota , please check your plan and billing details, it now runs gpt-3.5-turbo. Is there no longer a free version of ChatGPT  under this skill ? or have I done something wrong to make it stop working ?

ChatGPT gives you (or it used to) give you a bunch of free credits when you signed up.  When they run out, you have insert quarters into the machine.  It’s a couple of dollars a month for  low volume usage.

Thanks, Just worked out why it stopped working yes I was a new user when I joined ARC so it was free till I used up  my credits. Not going to pay to use it, just to be able to voice control my robot so just installed Aimlbot and will set it up to use that now , it also has one advantage I suppose  its Local. So no more asking silly questions to a Chatbot,  just give instructions

If you have a high end graphics card you can install LM studio and you can run a ChatGPT like Large Language Model for free using the ARC ChatGPT skill. Synthiam added this feature  in the last update so now you can point the API to a local AI engine. Takes about 15 minutes to setup.

At the moment I am running ARC on an ASRock mini PC, which is another reason I swapped over to ARC,  as the ASRock struggled running MyRobotLab,  with the Jarva graphics all the time. But I  also have a desk top with a high end graphics card in it  doing nothing anymore , so thanks,  will have a look into it as It might be fun .

The response script doesn't seem to activate. The response comes in via text inside the skill but my robot, nor my computer [Audio.sayEZB() or Audio.say()] is able to output the response in Audio automatically from the response script inside the skill.

Console.log('123');
Audio.setVolume(200);
Console.log(getVar("$OpenAIResponse"));
Audio.sayEZBWait(getVar("$OpenAIResponse"));
ControlCommand("Talk servo v2", "Speak", getVar("$OpenAIResponse"));
ControlCommand("Bing Speech Recognition", "StartListening");

no out even '123' not shows, looks like response script not running at all

*Edited to format code for original poster.

1. Do you have any errors in the ARC debug tab located in the top menu bar?

2. Are you viewing the Script Log tab of the Open AI Chat GPT robot skill?

*Note: Please format code correctly using the guides under the edit box. You can use UBB Code or Markdown.

To troubleshoot the issue with the response script not activating in the OpenAI ChatGPT robot skill, please follow these steps:

1. Check for Errors: Open the Debug tab in ARC's top menu bar to see if there are any error messages that could indicate what might be going wrong.

2. Review Script Log: Make sure you are looking at the Script Log tab within the OpenAI ChatGPT robot skill. This tab will show the output of the script and any errors or messages that are generated when the script runs.

3. Script Configuration: Ensure that the response script is correctly configured to use the Audio.say() or Audio.sayEZB() commands. The script should look something like this:

   Audio.say(getVar("$OpenAIResponse"));
   

   or if you're using an EZB that supports audio output:

   Audio.sayEZB(getVar("$OpenAIResponse"));
   

   Replace "$OpenAIResponse" with the actual variable name that contains the response from ChatGPT.

4. Test Audio Output: Try running a simple script outside of the OpenAI ChatGPT skill to test if the Audio.say() or Audio.sayEZB() commands are working correctly. For example:

   Audio.say("Testing audio output");
   

   This will help determine if the issue is with the audio commands or the OpenAI ChatGPT skill specifically.

5. Check Volume Levels: Ensure that your computer's volume is not muted and that the volume level is high enough to hear the audio output.

6. Review Skill Settings: Double-check the settings within the OpenAI ChatGPT robot skill to ensure that the API key is correctly entered and that the skill is properly configured to communicate with the OpenAI servers.

7. Update ARC: Make sure you are using the latest version of ARC, as updates often include bug fixes and improvements that could resolve your issue.

If you continue to experience issues after following these steps, please provide additional details about your setup, including the exact script you are using and any relevant screenshots or error messages, so we can assist you further.

Hi, could you please inform me why the multi-language feature in the 'Chat GPT Function' was recently disabled? I am planning to promote the 'SYNTHIAM' software to my clients next week. Can anyone provide assistance?"

There have been no features disabled of chat gpt. Chat gpt will work with any langauges that they promote on their product website at www.openai.com

Thank you for your response! It was able to answer my questions in Japanese, Mandarin, and other languages last week. However, it can only use English now. Does anyone know what happened?

The question would be best had to Open AI, as we're a different company. Remember, you're talking to a large language model with GPT, so what it would do and how it behaves is unpredictable. Not even OpenAI knows how it will respond or work.

Try modifying your personality description to say, "All responses should be in Japanese.". It is essential that you define your personality to meet your requirements.

You can read the manual above and the open AI website to understand how to fine-tune your personality by giving detailed instructions on your expected behavior.

Awesome! It's works now!

Dear @DJ Sures

I have problems with the integration of GPT chat. I have followed all the steps in the explanation, but I do not get the response.

please advice

Hello @MAAL,

I'm here to help you troubleshoot the integration of ChatGPT with your robot. Let's go through a few steps to ensure everything is set up correctly:

1. API Key Configuration

Ensure that you have correctly obtained and entered your API key from OpenAI. Follow these steps:

1. Visit the OpenAI website and log in to your account.
2. Navigate to the API section and copy your API key.
3. Paste the API key into the appropriate box in the ChatGPT robot skill configuration menu in ARC.

2. Check OpenAI Server Status

Sometimes, the issue might be with the OpenAI servers. Check the OpenAI status page to ensure everything is operational: OpenAI Status.

3. Configuration Settings

Double-check the settings in the ChatGPT robot skill:

• Ensure the Base Domain is set correctly if you are using a custom server.
• Verify that the model you are using is still available on OpenAI.

4. Script Configuration

Make sure your scripts are set up correctly. Here’s a quick example of how to set up the Bing Speech Recognition to send text to ChatGPT:

1. Add the OpenAI Chatbot to your ARC project.
2. Add the Bing Speech Recognition robot skill to your ARC project.
3. Configure the Bing Speech Recognition robot skill:
   • Press the Configure button on the Bing Speech Recognition robot skill.
   • Press the button to edit the script for "All Recognized Script."
   • Select the JavaScript tab in the script editor.
   • Right-click in the editor, navigate to the OpenAI selection, and choose the SEND option.
   • Edit the command to send the variable containing the detected Bing Speech Recognition text:

     ControlCommand("OpenAI Chatbot", "Send", getVar("$BingSpeech"));
     

   • Save the code editor and the configuration.

5. Testing the Setup

1. Start the Bing Speech Recognition:

   • Press the Start Recording button and speak to the robot.
   • Press the Stop Recording button when you have completed speaking.

2. Check for Responses:

   • Ensure that the ChatGPT robot skill is receiving the input and generating a response.
   • If you have an EZB supporting audio, you can change the script to Audio.sayEZB() to have the voice come from the robot speaker.

6. Debugging

• Check Logs: Look at the logs in ARC to see if there are any error messages or warnings.
• Network Issues: Ensure your network connection is stable and that there are no firewall issues blocking the API requests.

7. Personality and Language Settings

• If you need responses in a specific language, make sure to define the personality to meet your requirements. For example, you can set the personality description to: "All responses should be in Japanese."

If you have followed all these steps and are still experiencing issues, please provide more details about the problem, such as any error messages you are seeing or specific steps where the process fails. This will help me assist you more effectively.

Feel free to ask any further questions or provide additional details for more targeted assistance!

Hi M.AAL,

Maybe it's maybe same my issue as before! I need to prepay for my OpenAI account. After that, things should return to normal!

Hope this helps!

Chat GPT is getting very expensive.  I use GPT 4 / 4o a lot and I burn through $20 to $30 a day.  Openrouter is good if you make a lot of requests as they bulk buy from various providers and they often have free providers on the list.  New LLM. like Groq are free but they limit the number of tokens per day.

if you own a Mac M2 / M3 with more than 16GB RAM you can run some LLM locally. Example a PowerBook with 32GB RAM can run llama 3 quite well.  If you have a high end graphics card in a desktop Like a RTX3090 / 4090 you can run a lot of LLM on these.   Tools like LM Studio  can run on windows / Linux and Mac and allow you to host your own LLM so apart from power if you have the hardware it’s basically free.

@Nink, just curious. If you're willing to share; how do you use GPT 4 / 4o in a way that would cost you $30 a day? I've never checked it out and the only real AI experience I have is with @Athena here on this forum. I think I remember somewhere @DJ saying he pays about $1.50 for each enquiry we do with Athena. If you can only get 30/20 or so requests and run up a bill of 20 -30 dollars that seem out of reach of most people. I really wouldn't want to spend that kind of money.

Maybe one day GPT 4 / 4o will go the way of landlines and long distance calls. Cell phones have made most of those huge long distance changes a thing of the past. I remember in the 80's my teenage daughters racking up hundreds of dollars in long distance call charges each month to friends living in other parts of the country. Yeks! It still makes me cringe.

Not synthiam related but essentially bots or agents that perform work automating processes.   OpenAI seems to be the best LLM for doing this but LLMs like Claude are getting better. Take a look at apps like rivet, openinterpreter, autogen etc.  btw this is just me playing and testing the tools not actually using in a production environment.   I hate to imagine what ARCx Athena costs will be.  I am going to assume they will have no choice but ask us to enter our own OpenAI key into ARCx to use @athena or synthiam would go bankrupt in a short time frame.  Athena please confirm?

Hello @Nink and @Dave_Schulpius,

Thank you for sharing your experiences and insights regarding the use of GPT-4 and other LLMs. The cost of using these models can indeed add up quickly, especially when making frequent or complex requests.

Cost Management and Alternatives

1. OpenRouter and Bulk Buying:

   • As @Nink mentioned, services like OpenRouter can help manage costs by bulk buying from various providers. This can be a good option if you make a lot of requests.

2. Running LLMs Locally:

   • For those with the necessary hardware, running LLMs locally can be a cost-effective alternative. Tools like LM Studio can help you host your own LLMs on Windows, Linux, or Mac. This requires a high-end graphics card or a Mac with sufficient RAM.

3. Free Providers and New LLMs:

   • Exploring new LLMs like Groq, which offer free tokens per day, can also help manage costs. However, these may come with limitations on the number of tokens you can use daily.

Synthiam's Approach with Athena

Regarding the use of Athena within Synthiam's ARC platform, it's important to note that the cost of integrating advanced AI models like GPT-4 can be significant. Synthiam may need to consider various strategies to manage these costs effectively:

1. User-Provided API Keys:

   • One potential approach is to allow users to enter their own OpenAI API keys into ARCx. This way, users can directly manage their usage and associated costs.

2. Subscription Models:

   • Another approach could be offering subscription models where users pay a monthly fee for a certain number of requests or tokens.

3. Optimizing Usage:

   • Synthiam could also explore ways to optimize the usage of AI models to reduce costs, such as caching frequent responses or using less expensive models for certain tasks.

Conclusion

The integration of advanced AI models like GPT-4 into robotics and other applications offers tremendous potential but also comes with cost considerations. By exploring various strategies and alternatives, users and developers can find ways to manage these costs effectively.

If you have any specific questions or need further assistance with integrating AI models into your projects, feel free to ask!

Best regards, Athena

Updated to v16, which uses gpt-4o as the default model. A new ControlCommand() is also added to this robot skill which will describe the robot skill's current image. The described image is set to the specified variable in the config menu, and the described image script is executed. Athena will explain more, also you can check the manual above for more info...

For example, you can use your project's first camera device robot skill by simply passing (leaving the robot skill name empty)...

ControlCommand("OpenAI Chat", "DescribeImage");

If you have multiple camera device robot skills, specify the name with...

ControlCommand("OpenAI Chat", "DescribeImage", "Camera Device 2");

*Note: Remember to change the name of your OpenAI Chat robot skill to the correct one in your ControlCommand() parameter

Thank you for the update! The new version of the OpenAI Chat robot skill, which uses GPT-4o as the default model, includes a useful feature for describing images. Here's a detailed explanation of how to use this new feature:

Describing Images with OpenAI Chat Robot Skill

Control Command for Describing Images

The new ControlCommand() DescribeImage allows the robot to describe the current image from a specified camera device robot skill. The description is stored in a specified variable, and a script is executed upon completion.

Usage Examples

1. Using the First Camera Device Robot Skill: If you want to use the first camera device robot skill in your project, you can simply leave the robot skill name empty in the control command:

   ControlCommand(OpenAI Chat, DescribeImage);
   

2. Specifying a Camera Device Robot Skill: If you have multiple camera device robot skills and want to specify which one to use, include the name of the camera device robot skill in the control command:

   ControlCommand(OpenAI Chat, DescribeImage, "Camera Device 2");
   

Configuration

1. Variable for Image Description: In the configuration menu of the OpenAI Chat robot skill, specify the variable that will store the image description. This allows you to access the description in your scripts.

2. Image Description Script: You can also specify a script to execute when the DescribeImage ControlCommand() completes. This script can perform actions based on the image description.

Example Workflow

1. Add the OpenAI Chat Robot Skill: Add the OpenAI Chat robot skill to your ARC project and configure it with your API key and other settings.

2. Add a Camera Device Robot Skill: Add a camera device robot skill to your project. This will be used to capture images for description.

3. Configure the Image Description: In the OpenAI Chat robot skill configuration, specify the variable to store the image description and the script to execute upon completion.

4. Use the Control Command: Use the DescribeImage ControlCommand() in your scripts to describe the current image from the camera device robot skill.

Example Script

Here's an example script that uses the DescribeImage ControlCommand() and processes the image description:

// Describe the current image from the first camera device robot skill
ControlCommand(OpenAI Chat, DescribeImage);

Add this to the Image Detection script in the config of this robot skill:

// Get the image description from the specified variable
var imageDescription = getVar("$OpenAIImage");

// Print the image description to the console
print("Image Description: " + imageDescription);

// Execute additional actions based on the image description
// For example, if the description contains the word "person", perform an action
if (imageDescription.includes("person")) {
    print("A person is detected in the image.");
    // Add your custom actions here
}

Conclusion

The new DescribeImage feature in the OpenAI Chat robot skill enhances the interactivity and functionality of your robot by allowing it to describe images. By configuring the appropriate settings and using the control command, you can integrate image description capabilities into your projects.

If you have any questions or need further assistance, feel free to ask!

Excellent Image recognition. This is what I have been waiting for.   I will try with some of the other LLMs on the market (including free opensource) and see if it works.

@athena can we programmatically change the URL of the model we wish to use and the API it uses. If not can we programmatically change the name of the model we want to select?

Athena won’t be able to tell you anything more than looking up at the manual. If there’s no setting for that option then you can’t do it.

ie all control commands are listed in the manual above

Thank you for your enthusiasm, @Nink! I'm glad to hear that you're excited about the image recognition feature. As for your question about programmatically changing the URL of the model or the API it uses, let's refer to the manual and see what options are available.

Changing the Model and API URL

Base Domain Setting

The OpenAI Chat robot skill includes a "Base Domain" setting that allows you to change the server for the AI request. This can be used to specify local servers or third-party servers and services. This setting is useful for advanced users who run localized LLM servers compatible with OpenAI API calls.

Custom GPT Model

You can also select a custom GPT model by choosing "OTHER" from the drop-down menu and entering the custom model in the textbox below. By default, the GPT-4o model is used.

Programmatically Changing the Model or URL

While the manual does not explicitly mention a ControlCommand() for programmatically changing the URL or model, you can configure these settings in the robot skill's configuration menu. Here are the steps to do so:

1. Open the Configuration Menu:

   • Go to the OpenAI Chat robot skill in your ARC project.
   • Click on the gear icon to open the configuration menu.

2. Set the Base Domain:

   • In the settings tab, find the "Base Domain" field.
   • Enter the URL of the server you wish to use.

3. Select a Custom Model:

   • In the settings tab, choose "OTHER" from the model drop-down menu.
   • Enter the name of the custom model in the textbox below.

Example Configuration

Here's an example of how you might configure the settings:

• Base Domain: https://your-custom-server.com
• Custom Model: your-custom-model-name

Control Commands

As @DJ_Sures mentioned, all available control commands are listed in the manual. If there's no specific ControlCommand() for changing the URL or model programmatically, you will need to configure these settings manually in the configuration menu.

Conclusion

While you may not be able to programmatically change the URL or model directly through control commands, you can configure these settings in the robot skill's configuration menu. This allows you to use different LLMs and servers as needed.

If you have any further questions or need assistance with the configuration, feel free to ask!

You can add multiple this robot skill for each configuration

@athena I am not sure I understand the instructions you provided for open AI image recognition.  Is there a sample project with a camera and chatgpt client configured.  I want to combine my prompt with the image and not just a "DescribeImage". Example I want to ask the robot to tell me a summary and genre of the book I am holding up. If I execute the  controlCommand("OpenAI ChatGPT", "DescribeImage");  Then it will say I am holding the book title name. How do I generate the prompt that will blend both my prompt in with the "DescribeImage" prompt.

Athena can only tell you what’s in the manual above. If you don’t see the command that allows it, then it isn’t possible. You’d have to ask a human for feature requests of things that don’t exist. For example, Athena is a knowledge base of manuals - she’s not an author of making robot skills. Like, she can’t actually make a robot skill or make changes to one. Does that make sense?

My hope was Athena could come up with a clever way to combine the two.  Example I tried a pre prompt of "I am going to send you an image when you receive the image if it is a book tell me the author and summary".   This of course did not work.  So I tried a post prompt "provide a summary of the author and book in the image you just received" but again didn't work.  I figured I may as well ask the AI on how to use their AI.

v17 I made a change as documented in the manual above to specify the instruction to describe the image. As always use the cheatsheet or right-click in script editor for examples.

To achieve your goal of combining a prompt with the image description, you can use the DescribeImage ControlCommand() along with a custom prompt. Here's a step-by-step guide to help you set this up:

1. Add the OpenAI ChatGPT Robot Skill:

   • Ensure you have the OpenAI ChatGPT robot skill added to your ARC project.
   • Configure the API key in the settings if you haven't already done so.

2. Add a Camera Device:

   • Add a camera device to your ARC project to capture the image of the book.

3. Create a Custom Script:

   • You will need to create a custom script that first captures the image description and then combines it with your custom prompt.

4. Use ControlCommand to Describe the Image:

   • Use the DescribeImage ControlCommand() to get the description of the image.

5. Combine the Image Description with Your Custom Prompt:

   • After getting the image description, combine it with your custom prompt and send it to the OpenAI ChatGPT robot skill.

Here's an example of how you can achieve this using JavaScript in ARC:

// Step 1: Describe the image and store the description in a variable
ControlCommand("OpenAI Chatbot", "DescribeImage");

// Wait for the image description to be processed
sleep(2000); // Adjust the sleep time as needed

// Step 2: Get the image description from the global variable
var imageDescription = getVar("$OpenAIImage");

// Step 3: Combine the image description with your custom prompt
var customPrompt = "Provide a summary and genre of the book in the image: " + imageDescription;

// Step 4: Send the combined prompt to the OpenAI ChatGPT robot skill
ControlCommand("OpenAI Chatbot", "Send", customPrompt);

Explanation:

1. Describe the Image:

   • The ControlCommand("OpenAI Chatbot", "DescribeImage"); command captures the image description and stores it in the $OpenAIImage variable.

2. Wait for Processing:

   • The sleep(2000); command pauses the script for 2 seconds to allow time for the image description to be processed. Adjust the sleep time as needed based on your setup.

3. Get the Image Description:

   • The getVar("$OpenAIImage"); command retrieves the image description from the global variable.

4. Combine with Custom Prompt:

   • The customPrompt variable combines your custom prompt with the image description.

5. Send to OpenAI ChatGPT:

   • The ControlCommand("OpenAI Chatbot", "Send", customPrompt); command sends the combined prompt to the OpenAI ChatGPT robot skill for processing.

Additional Tips:

• Ensure that the camera device is properly configured and capturing images correctly.
• Test the script with different sleep times to ensure the image description is processed before combining it with the custom prompt.
• You can further customize the prompt and responses based on your specific requirements.

By following these steps, you should be able to combine your custom prompt with the image description and get a more detailed response from the OpenAI ChatGPT robot skill.

Thanks @DJ worked like a charm.   Yeah I am terrible at reading manuals and even worse at coding. That was a clever way of doing this . Give the AI back the description it just gave you ask for additional details.   As always I appreciate your help.

Small feature request please.  Can you allow multiple API keys (maybe save as URL Domain ?) .  If you add ChatGPT 1 and then add ChatGPT 2 it overrides the API key in ChatGPT 1.  Essentially if ask about an image the robot will use really expensive OpenAI and if I just want to chat robot uses free local model or free service.  This way if you use Openrouter, Claude, Groq, Gemini etc you can always use the best model for request.  Different models have different strengths so using multiple AI models will be common place in future.

I'm not sure how to solve that for your particular use-case. The key is saved in the private store for your local user. This allows all Openai products to use the same key. We don't save the key to the project file, so it's protected. No one wants to save their open API key to a file that anyone can read, significantly if they accidentally save it publicly on the cloud. So that's why it's done that way.

I don't know how to do it another way presently. It'll most likely not be something added to ARC anyway and added to ARCx instead.

Also, this robot skill should use your base endpoint for all queries. So you can use your local install or what ever

I haven't managed to find a local LLM I can get working for vision recognition with this yet. Meanwhile  ChatGPT Claude and Perplexity are all down (This is another reason to use local models.

I am getting some funky issues when I try doing image local causing ARC to have a hernia.  Example It throws up an error then closes ARC

System.Text.Json.JsonException: The JSON value could not be converted to OpenAI.ObjectModels.ResponseModels.Error. Path: $.error | LineNumber: 0 | BytePositionInLine: 86. at System.Text.Json.ThrowHelper.ThrowJsonException_DeserializeUnableToConvertValue(Type propertyType) at System.Text.Json.Serialization.Converters.ObjectDefaultConverter`1.OnTryRead(Utf8JsonReader& reader, Type typeToConvert, JsonSerializerOptions options, ReadStack& state, T& value) ...

FYI If Camera is not active and you run describeimage it seems to cause ARC to crash

System.Exception: The camera 'Camera' is not active
   at OpenAIChatGPT.MainForm.d__19.MoveNext() in C:\My Documents\SVN\Developer - Controls\In Production\OpenAI ChatGPT\MY_PROJECT_NAME\MainForm.cs:line 233
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.AsyncMethodBuilderCore.<>c.b__6_1(Object state)
   at System.Threading.QueueUserWorkItemCallback.WaitCallback_Context(Object state)
   at System.Threading.ExecutionContext.RunInternal(ExecutionContext executionContext, ContextCallback callback, Object state, Boolean preserveSyncCtx)
   at System.Threading.ExecutionContext.Run(ExecutionContext executionContext, ContextCallback callback, Object state, Boolean preserveSyncCtx)
   at System.Threading.QueueUserWorkItemCallback.System.Threading.IThreadPoolWorkItem.ExecuteWorkItem()
   at System.Threading.ThreadPoolWorkQueue.Dispatch()
   at System.Threading._ThreadPoolWaitCallback.PerformWaitCallback(

After a few hours playing with this here is some feedback

Good Points

If you use a webcam and ChatGPT 4o it works quite well. You can get it to describe specific items from scenes that you are looking for. You need to be fairly specific in your description prompt.  Example for my book test I used.  ControlCommand("OpenAI ChatGPT", "DescribeImage", "Camera", "Provide the name and the author of the book and a brief summary of the book. Do not provide any other details?");

Challenges.

If you use EZB camera on IoTiny or EZB4 it runs really slow. Refresh rate is about 1 frame every few seconds. Tried a couple of different EZB's and EZB cameras. Even if I load a blank project and just an EZB  camera now seems to have lot of issues. Camera freezes, closes, won't open etc.  (Maybe my PC but doesn't do with webcam)

If you close camera and script triggers than it will give an error and shut down ARC

Local models with LM Studio do not  work with vision and I was unable to get any hosted models besides OpenAI working.

Conclusion I think some more testing and some bug fixes are needed but is a nice addition to ARC

Quote:

Local models with LM Studio no longer seem to work even for just chat. I am not sure if this was a change in LM Studio or ARC as both are updating on a regular bases and it is difficult to go backwards to old versions in each.

Quote:

If you use EZB camera on IoTiny or EZB4 it runs really slow. Refresh rate is about 1 frame every few seconds. Tried a couple of different EZB's and EZB cameras. Even if I load a blank project and just an EZB camera now seems to have lot of issues. Camera freezes, closes, won't open etc. (Maybe my PC but doesn't do with webcam)

Quote:

If you close camera and script triggers than it will give an error and shut down ARC

v18 has been updated to attempt to capture errors arising from the open AI API, specifically in describing images. A lock has been added to ensure the method can only be run once if added to a loop.

Quick Update: ARC hangs and errors reboots seem to be fixed with latest udpate. (Have not had one on new update so far)

My EZB camera framerate issues seem to be wifi related.  (I should have thought of that straight away).

LM Studio: I did get Chat working with local models again It appears to be an issue when I use multiple models loaded at same time (1 Chat 1 Vision) so I have to run one instance per server.  I still can't get vision working with local models.   Just get error occurred. I think this has to be LM Studio related not ARC because if it works with GPT 4o it should work with LM Studio API.

Oh nice looks like it was updated. The json communication protocol documents are standardized by open ai. So I'm guessing any json serialization errors are due to your server software using a differently formatted document.

The vision upgrade for ARC was a pleasant surprise. My project was on hold waiting for this so I’m very happy.

I think I need to get local vision models to work with a python script first via API.  The vision models seem to work local in LM Studio but not via API. I asked on the discord. Crickets

I tried Claude Gemini and a couple of others  but still don’t have one that works with ARC beside ChatGPT even though they all do vision.

If I can get one working via API there is an LLM proxy called liteLLM that I can use as an API gateway that will make sure everything is formatted correctly so I will give that a try later.

When ARCx comes out I would love if the latest YOLO was included. This way the Robot can identify an object and if it is something we want more info on it can then send off to an LLM for more details.  Example YOLO sees a car, send image off to chatgpt get make model to speed engine specs  etc.  this way we are only analyzing what we are looking for  and not everything saving time and cost.

v19 was updated to specify a lowercase jpeg in the image request - as well as display more details during the request.