Exciting news for tech enthusiasts! Microsoft has just unveiled a game-changing Mobile App that's completely free - it's your gateway to the incredible capabilities of GPT-4, thanks to the innovative Copilot app. Don't miss out on this cutting-edge technology! Dive into the details and witness the future unfold by watching our latest video. It's packed with info and showcases the amazing potential of this new tool. Your journey into the next level of tech starts here - watch now and be amazed! ua-cam.com/video/Crpek9r3KFY/v-deo.html
do you have any direcitons on how to incorporate sharepoint into the mix? e.g. I want ChatGPT to be able to fetch information from a sharepoint library/database. Thanks!
Yes, there are multiple Sharepoint connectors that should let you achieve that. Just extract the information with one of the connectors and pass it in the prompt to chat GPT.
First off, thanks so much for posting this. Easy and clear instructions. I noticed your response took about 8 minutes. I tested out some simple queries and it's slow to respond. I'm using the gpt-3.5-turbo-instruct model and it has a 150,000 TPM limit, so I expect a bit quicker response. Have you found a way to speed this up?
thank you so much for your video, you showed along with explaining what I had questions on and was stuck at. Showing is always better than just telling.
I copied the text from the video for the body it did not return a answer instead returned the text how do I fix this: Finesse Jones via Power Automate {body('HTTP')['choices'][0]?['text']}
Hey @finessejones3109! Thanks for the kind words about the video. 😊 Regarding the issue you're facing, it seems like there's a small hiccup with the expression you're using to extract the response from ChatGPT. Expression Check: The expression {body('HTTP')['choices'][0]?['text']} is meant to extract the text from the response of the HTTP request to ChatGPT. If you're seeing this exact text returned, it means the expression isn't being evaluated correctly. Power Automate Expression: When you're adding this expression in Power Automate, make sure you're adding it in the expression editor and not directly as plain text. Here's how: In the action where you want to use the response (e.g., "Reply to a message in Teams"), click on the "Expression" tab. Enter the expression body('HTTP')['choices'][0]?['text']. Once entered, it should show up in the "Dynamic content" pane. Click on it to add it to the action. Flow Check: Ensure that the HTTP request to ChatGPT is correctly set up and returning a valid response. If the HTTP request isn't working as expected, the expression won't be able to extract the desired text. Give these steps a shot, and hopefully, you'll get the desired response from ChatGPT in your Teams channel.🔧🚀
Thanks for the walk through, I got it working following your steps. Q), how can I prompt myGPT I created to be used in teams for the same purpose, rather than the generic OpenAI models?
Glad it works. Can you please ellaborate on your use cases. Do you mean you created a GPT in the Chat GPT interface, and now you want to use the same instructions when prompting from team?
@boulangerAI Hey, thanks for the feedback. I created a Human Resources GPT in the interface that is trained on some pdf's for general HR questions, which I intend to use as my "model" parameter in the HTTP request. If I use my ChatGPT keyword, followed by asking it to use my GPT specifically in the Teams prompt I can get it to work, but I do not want the user to ask for the specific GPT in the prompt if that makes sense.
Hey @dyeLucky! Thanks for the kind words about the video. 😊 Hmm, if your flow is running indefinitely or timing out, there could be a few reasons. Let's dive in: HTTP Request Timeout: By default, the HTTP request action in Power Automate has a timeout of 100 seconds. If ChatGPT is taking longer than that (which is rare, but can happen during peak times), it might cause the flow to time out. You can try increasing the timeout in the HTTP action settings. Looping Issue: Ensure there's no accidental loop in your flow. For instance, if the flow is set to trigger on a Teams message and then sends a message back to Teams, it could re-trigger itself. Flow Concurrency: Power Automate has a limit on how many flows can run concurrently. If you have other flows running at the same time, it might cause a delay. Check if there are any other heavy flows running on your tenant. API Rate Limits: OpenAI has rate limits for API requests. If you're making a lot of requests in a short time, you might hit these limits. You can check OpenAI's documentation for the specific rate limits based on your subscription. Error Messages: Even though you mentioned there's no specific reason, it's worth checking the flow's run history in Power Automate. Sometimes, there might be a subtle error message or warning that can give a clue. Flow Structure: Double-check the flow structure. Ensure that after the HTTP request to ChatGPT, there's a proper action to send the response back to Teams. If you've gone through these and still can't pinpoint the issue, it might be worth recreating the flow from scratch. Sometimes, starting fresh can help identify any misconfigurations. Hope this helps
Hey @@samuelboulangerAI! Thanks for the reply! Everything was set up perfect, but I wasn't 100% sure if this would work on a free OpenAI account (it doesn't...); my error was 'You exceeded your current quota, please check your plan and billing details.'
Thanks for a great video! Please make video on how to keep context. And also include how to avoid Power Automate to loop if trigger-keyword is mentioned in reply.
Funny as I tried with two tutorials, and if I put the dynamic contents like this, it will reply again and again until I turn the flow off. Then, when I put "Reply to message ID" and "Message ID" vice versa, it works fine. Wonder why is that..?
Will others in the teams system be able to see all the questions and answers from said CHATGPT that were done inside CHATGPT online or just for that particular channel?
Sure, actually if you're using Copilot Studio which use chat GPT as the model behind you have a lot of supported channels. Other than then, considering we are making an API call, any platform that support the integration with power automate will works.
This is great - I wonder if you could go over how I would set this up so ChatGPT is providing answers based on content I've created. Ideally from a OneNote Notebook or even word doc?
Thanks for the video Samuel, it worked like magic. I have a question on how can we add local files in a workflow such a that the gpt model gives the information through the files as well.
Glad you liked it :) It depends on the type of the document, the lenght and structure. For a small document when only text you can use the OCR capabilities of Ai builder. For a form, form a recognizer as well in AI builder. For more complex scenarios you will have to explore something like azure search paired with azure OpenAI or another vector database to store the document.
Great video but your explanation on Authorization and Bearer were very unclear. Did you type this or pull from the list? I cant get it to work. Also, what model name? Im using model "GPT-4" but that snot working as a model in the BODY. Any ideas?
Is there a way that the Teams message takes the text formatting of the ChatGPT response? It just posts messages in Teams with escape characters and Markdown tags for me. E.g. for code formatted with ```
Great video! Worked as expected. How would one go about integrating ChatGPT as a stand alone bot? My ideal situation would be to add ChatGPT as a user so if I have a question I can just message that user rather than everyone seeing what I am asking the bot in a channel. Any guidance is much appreciated :)
Thanks for the video, Step-by-step instructions. I am not getting any errors but when I am testing, it's not giving me any output. it just shows in the running status
is it possible to give it a prompt and when you type something it will respond based on that prompt you previously gave (not when asking the question but before that, like "give answers in xyz format" and we wont need to type that every single time)?
great session, i have followed all steps but face the error "We could not parse the JSON body of your request. (HINT: This likely means you aren't using your HTTP library correctly. The OpenAI API expects a JSON payload, but what was sent was not valid JSON" not sure where went wrong? thankssss
Awesome integration, but incase we have custom GPT which read data from internal folder using LLM and going with data to chatgpt for NLP processing how we can automate the same ?
Hi I have an doubt. I have chat GPT account API with my personal email ID and i tried to setup this flow on my organization official email ID does this combination will work? please clarify.
@@samuelboulangerAI Thanks am getting the below error message on flow HTTP process { "error": { "message": "You exceeded your current quota, please check your plan and billing details.", "type": "insufficient_quota", "param": null, "code": null } can you let me know what should I do now ?
Great video, thanks Samuel, however Im receiving the following error in my flow - { "error": { "message": "'messages' is a required property", "type": "invalid_request_error", "param": null, "code": null } } If think the model you provided is now deprecated, is there are a substitute. Thank you
I don't see add dynamic content in my flow. Instead I get something says " Select a parameter then provide examples and we'll suggest an expression" so I am not able to add the expression in the last part for converting the body
@@pudgy_buns I do have Premium plan. I sorted it out. I had a different view because I was creating the flow from PowerAutomate and not from powerApps.
Was able to set this up following your instructions. Super cool! Thank you! Is there any way to get the Chat GPT reply to be named as "Chat GPT" and not "[name] via Power Automate"?
Happy you've been able to make this work :) It's possible but you'll have a to create a user called "Chat GPT" and use this user to send the message instead of the bot. Here's a more detailed explanation on the community forum : powerusers.microsoft.com/t5/Using-Flows/Change-the-name-in-the-Via-Power-Automate-Message/td-p/1620106
Thank you for the video, i am facing an error in the http stage "ActionFailed. An action failed. No dependent actions succeeded" "Bad request" any clue?
would it be possible to add a query with power automate, for example, I'm on the phone with a customer, I need to know the status of a shipment, ChatGPT is linked to my audio call, hears the words "status and shipment", and automatically begins a chat with me and my s/r department head?
Is there a way to do this such that a person can post and then then instead of triggering a response from chatgpt, ai searches through a given list of websites instead?
With Power Automate, it would be really complicated. Basically to pass information from a website you first need to extract the data of the website and then pass it to Chat GPT in the prompt. An example would be to add in the prompt : {Your question} Answer only with information from bellow source : #Sources: {Data extracted from the website}
Hi there, I am getting an error when I try this but I am wondering if this is my work VPN causing the issue. Can you give me a way to contact you to figure out what is happening ?
Sorry to hear you're having trouble. Let's try to troubleshoot this: API Key: First, double-check if you've correctly added your OpenAI API key in the HTTP action. A common mistake is missing the "Bearer " part before the key. Triggers: Ensure that you're using the correct keyword "chat GPT" in Teams to trigger the flow. Remember, it's case-sensitive. Power Automate Flow Errors: Head over to Power Automate and check the flow's run history. If there's an error, it should give you more details on what went wrong. Token Limit: Remember the token limit mentioned in the video? If your question and the answer combined exceed the token limit, the flow might fail. Try a shorter query to test. Permissions: Ensure you have the necessary permissions in Teams to send messages. Sometimes, admin settings can restrict bots or automated flows from sending messages. Flow Test: When testing the flow in Power Automate, make sure you're using the "Test" button and then selecting "I'll perform the trigger action" before heading to Teams to type your message. If you've checked all these and it's still not working, let me know more details or any error messages you're seeing. We'll get this sorted out! 👍🔧
Great video! I'm hitting a failure in my testing though. I get an Action HTTP failed - 'messages' is a required property. { "error": { "message": "'messages' is a required property", "type": "invalid_request_error", "param": null, "code": null } }
Thank you for your feedback, I'm glad you liked it. You want to drop a csv file in MS team, than retrieve the data and ask questions about the data, I'm I correct? It would be feasible but too complicated to explain in a message. This will require to use the chat API instead of the completion API. You will need to start with a system message telling GPT to answer only based on the data you provide and then use the user message to ask the question.
@@samuelboulangerAIYou could probably do like a part 2 of this video with features in consideration of all the comments 🤓 Great video btw, really helpful and professional 🙌🏻
Totally yes. Please refer to the documentation for more details : learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=command-line%2Cpython-new&pivots=rest-api But since I released this video, we released Copilot Studio which make the integration so much easier
thank you for the video. I have followed the all of the required steps but am receiving an HTTP error 'messages' is a required property. Is there any help here with this?
ChatGPT starts answering this on loop after it answered my question: Q: What is the difference between a Data Scientist and a Data Analyst? A: Data Scientists are responsible for extracting insights from large datasets and using advanced analytics techniques to identify patterns and trends. They are also responsible for developing predictive models and algorithms to help organizations make better decisions. Data Analysts, on the other hand, are responsible for collecting, organizing, and analyzing data to identify trends and patterns. They use statistical methods to interpret data and create reports and visualizations to help organizations make informed decisions. Data Analysts are also responsible for creating data models and developing data-driven solutions.
Awesome video! I tried it and it works. But there is only 1 problem, the flow cant be stopped lol , the chatgpt keeps spam messaging ' Q: What is the difference between a Data Scientist and a Data Analyst? A: Data Scientists are responsible for extracting insights from large datasets and using them to inform decisions. They use a variety of techniques, such as machine learning, statistical analysis, and data mining, to uncover patterns and trends in data. Data Analysts, on the other hand, focus on collecting, organizing, and analyzing data to answer specific questions. ' under the comment sections. I have to revoke the API keys every time to stop the flow. Any suggestion to resolve this spam responses?
If you are using the "when a new chat" trigger you will need to add an exception for the flow user. If not the chat response from your GPT will trigger a new flow (causing an infinite loop). Use conditionals to except flow user from making a HTTP request. For example IF "from chat details get user of msg" contains flow user, don't send request.
Not using the method in the video but generally speaking yes. Or you can use Microsoft Copilot as well. Simply drop the document and ask it to answer solely based on the information in the document.
Thank you for this video but I need help with this link you promised will go with this video. I can't seems to find it body('HTTP')['choices'][0]}('text'0)
You are right sorry for that. I was leaving for vacation and posted in a hurry. Just updated the description and here's the Fx : {body('HTTP')['choices'][0]?['text']}
You should select manually instead of automatically and then go in MS Team and test it directly. Then the flow will run and you will be able to select automatically next time.
Hi Same issue from my side. When testing manually I can trigger the flow but the flow looks Frozen on the HTTPS connector and then fails few minutes later 😢
I have an error on the HTTP part. When I run the test, it shows an error “Invalid URL (POST/v1/chat/completions”. All the input seems to be okay. Any ideas? Thank you
Here's ChatGPT answer ;) : This error message indicates that there is an issue with the URL you are using to call the OpenAI GPT-3 API from Power Automate. Double-check that you have correctly specified the URL, including the correct API version and endpoint (in this case, "/v1/chat/completions"). Make sure that you have also provided the correct authorization credentials (i.e., API key or token) in your Power Automate action.
@@samuelboulangerAI I have the same issue as well with the Invalid URL . Is this because I am behind a proxy (cisco umbrella). Thanks for a very detailed video
Great session and thank you. I followed everything but received an error message: Action 'HTTP' Failed. Error Details: The provided input content is not valid: 'The provided content-type header value 'application-json' is not well formed.'. What am I missing and thank you again!
FYI, Under HTTP, additional error message below: BadRequest. The provided input content is not valid: 'The provided content-type header value 'application-json' is not well formed.'.
I got error - Flow save failed with code 'OpenApiOperationParameterValidationFailed' and message 'Input parameter 'requestBody' validation failed in workflow operation 'When_keywords_are_mentioned': The API operation 'WebhookKeywordTrigger' requires the property 'requestBody/channels' to have more than '1' items but has '0' items.'.
I'll say that the easiest way would be to change the keyword that trigger the flow to something you know for sure won't be used in the answer. Like " HeyChatGPT" instead of "ChatGPT".
Awesome tutorial only a thing... this expression {body('HTTP')['choices'][0]?['text']} doesn't work for me? anyone who got a solution? anyone who encountered this error too?TY
This is great, set it up, worked fine for about 5 messages, now it's stopped reacting. The trigger word no longer to seems to kick off the flow, plenty of free credit on my account left too, this happened to anybody else?
@@samuelboulangerAI Indeed, it’s a hassle though. Once you’ve had a chat with ChatGPT on a particular topic, it’s incredible how the conversation flows. It even references it’s past responses for clarity. It’s very natural and conversational without this context - it’s hard to put in the effort otherwise. Pretty good that we can have some functionality in steams though.
Very good video but I cannot make this work for the life of me. The flow fails with a "ActionFailed. An action failed. No dependent actions succeeded." and Bad Request in the Get Message Details action. No clue what is wrong.
I have a Power Automate Premium license and using Teams version: 1.6.00.24078. I noticed that I can no longer start a conversation in Teams but rather a "Post" and I must put a headline,etc. Any thoughts?
Hey @jasoncardinal7502! Thanks for the feedback on the video. Let's tackle this step by step: "ActionFailed" & "Bad Request" in Get Message Details: This error typically indicates that the flow is having trouble retrieving the details of the message from Teams. Here are a few things to check: Message ID: Ensure that the "Message ID" you're passing to the "Get Message Details" action is correct. It should be dynamically populated from the trigger. Permissions: Double-check if the account linked to Power Automate has the necessary permissions to read messages in Teams. Teams Version & "Post" Issue: The version you're using seems relatively recent. Microsoft Teams has been evolving, and some features or UI elements might change. If you're seeing a "Post" option instead of starting a conversation, it's likely a UI update. Headline Requirement: If Teams is now requiring a headline for a post, it might be affecting how the flow is triggered. The flow might be expecting a simple message, but if Teams is now structuring it as a post with a headline, body, etc., it could be causing the flow to fail. Workaround: For now, try using the "Reply" option in an existing conversation in Teams to send a message without a headline. This might allow you to bypass the new "Post" structure and trigger the flow as intended. Hang in there! Tech can be a bit finicky sometimes, but we'll get this sorted. Let me know how it goes! 👨💻🛠️🚀
Thank you for your comment. Unfortunately, UA-cam doesn't let me edit a video after it's been uploaded. I can cut a section, but that's it. I would probably make an updated version.
Hi Nice Video But i need help that is why every time we need to type ChatGpt for asking Queries , Why Can't we ask direct like not using any keyword , It is Possible , if it is can you explain please.
Used to do similar project to this myself some months back. Basically just use a trigger: "when message is posted" in teams instead of the conditional "if keayword are mentioned, then pick the message"
Great video! Thanks for the great information. I keep running into the flow test being fine but then when I try it in teams, nothing populates in teams and I go into power automate and have an error message that is in the when keywords are mentioned and says this in the body { "responses": [ { "id": "0", "status": 404, "headers": { "Cache-Control": "no-cache", "Content-Type": "application/json" }, "body": { "error": { "code": "ExtensionError", "message": "Operation: Create; Exception: [Status Code: NotFound; Reason: Resource not found]", "innerError": { "date": "2023-05-13T01:59:19", "request-id": "58", "client-request-id": "5baa8" } } } } ] } I deleted the api key for posting this :)
Here's what Chat-GPT told me based on your error message ;) : The error message you provided indicates that there is an issue with a resource not being found. It seems like there is a problem with the "Create" operation in your flow. Here are a few steps you can take to troubleshoot and resolve the issue: 1. Verify the connection: Check if the connection to the target service (Teams, in this case) is properly established in your Power Automate flow. Sometimes, connection issues can cause resource not found errors. Make sure the connection is active and authenticated correctly. 2. Check for missing or incorrect inputs: Review the inputs you're passing to the "Create" action in your flow. Ensure that all the required fields are provided and their values are valid. Double-check the configuration of the action to ensure it matches the expected inputs of the Teams API. 3. Validate permissions: Ensure that the user or account used for the connection has the necessary permissions to perform the "Create" operation in Teams. Insufficient permissions can lead to resource not found errors. Verify the permissions and consider using an account with appropriate access. 4. Confirm the existence of the resource: Double-check that the resource you are trying to create (such as a channel, message, or other Teams entity) actually exists. If the resource is deleted or does not exist, you may encounter a resource not found error. Make sure the resource is available and accessible in your Teams environment. 5. Review the request-id and client-request-id: The error message includes request IDs ("request-id" and "client-request-id"). These IDs can help with troubleshooting the specific request that resulted in the error. You can use these IDs to investigate further in your service logs or contact the support team of the service you're integrating with (e.g., Microsoft Teams) for assistance. By following these steps, you should be able to identify and resolve the issue causing the resource not found error in your Power Automate flow.
Thanks for the tutorial - after setup I get below reply - can you help - package com.example.demo.controller; import com.example.demo.model.User; import com.example.demo.service.UserService; import org.springframework.beans.factory.annotation.Autowired; import org.springframework.web.bind.annotation.*; import java.util.List; @RestController @RequestMapping("/user") public class UserController {
@@samuelboulangerAI I get this back- I think its a similar- error like Java? totally unexpected ! package com.example.demo.controller; import com.example.demo.model.User; import com.example.demo.service.UserService; import org.springframework.beans.factory.annotation.Autowired; import org.springframework.web.bind.annotation.*; import java.util.List; @RestController @RequestMapping("/user") public class UserController { @Autowired private UserService userService; @GetMapping("/all") public List getAllUsers(){ return userService.getAllUsers(); } @PostMapping("/add") public void addUser(@RequestBody User user){ userService.addUser(user); } @PutMapping("/update") public void updateUser(@RequestBody User user){ userService.updateUser(user); } @DeleteMapping("/delete/{id}") public void deleteUser(@PathVariable int id){ userService.deleteUser(id); } }
Exciting news for tech enthusiasts! Microsoft has just unveiled a game-changing Mobile App that's completely free - it's your gateway to the incredible capabilities of GPT-4, thanks to the innovative Copilot app. Don't miss out on this cutting-edge technology! Dive into the details and witness the future unfold by watching our latest video. It's packed with info and showcases the amazing potential of this new tool. Your journey into the next level of tech starts here - watch now and be amazed! ua-cam.com/video/Crpek9r3KFY/v-deo.html
do you have any direcitons on how to incorporate sharepoint into the mix? e.g. I want ChatGPT to be able to fetch information from a sharepoint library/database. Thanks!
Yes, there are multiple Sharepoint connectors that should let you achieve that. Just extract the information with one of the connectors and pass it in the prompt to chat GPT.
First off, thanks so much for posting this. Easy and clear instructions.
I noticed your response took about 8 minutes. I tested out some simple queries and it's slow to respond. I'm using the gpt-3.5-turbo-instruct model and it has a 150,000 TPM limit, so I expect a bit quicker response. Have you found a way to speed this up?
thank you so much for your video, you showed along with explaining what I had questions on and was stuck at. Showing is always better than just telling.
I copied the text from the video for the body it did not return a answer instead returned the text how do I fix this: Finesse Jones via Power Automate
{body('HTTP')['choices'][0]?['text']}
Hey @finessejones3109! Thanks for the kind words about the video. 😊
Regarding the issue you're facing, it seems like there's a small hiccup with the expression you're using to extract the response from ChatGPT.
Expression Check: The expression {body('HTTP')['choices'][0]?['text']} is meant to extract the text from the response of the HTTP request to ChatGPT. If you're seeing this exact text returned, it means the expression isn't being evaluated correctly.
Power Automate Expression: When you're adding this expression in Power Automate, make sure you're adding it in the expression editor and not directly as plain text. Here's how:
In the action where you want to use the response (e.g., "Reply to a message in Teams"), click on the "Expression" tab.
Enter the expression body('HTTP')['choices'][0]?['text'].
Once entered, it should show up in the "Dynamic content" pane. Click on it to add it to the action.
Flow Check: Ensure that the HTTP request to ChatGPT is correctly set up and returning a valid response. If the HTTP request isn't working as expected, the expression won't be able to extract the desired text.
Give these steps a shot, and hopefully, you'll get the desired response from ChatGPT in your Teams channel.🔧🚀
Thanks for the walk through, I got it working following your steps. Q), how can I prompt myGPT I created to be used in teams for the same purpose, rather than the generic OpenAI models?
Glad it works. Can you please ellaborate on your use cases. Do you mean you created a GPT in the Chat GPT interface, and now you want to use the same instructions when prompting from team?
@boulangerAI Hey, thanks for the feedback. I created a Human Resources GPT in the interface that is trained on some pdf's for general HR questions, which I intend to use as my "model" parameter in the HTTP request. If I use my ChatGPT keyword, followed by asking it to use my GPT specifically in the Teams prompt I can get it to work, but I do not want the user to ask for the specific GPT in the prompt if that makes sense.
Hey @Samuel Boulanger!
Great video! One problem I'm having on my tenant, but the flow runs indefinitely / times out. No specific reasons...Any idea?
Hey @dyeLucky! Thanks for the kind words about the video. 😊
Hmm, if your flow is running indefinitely or timing out, there could be a few reasons. Let's dive in:
HTTP Request Timeout: By default, the HTTP request action in Power Automate has a timeout of 100 seconds. If ChatGPT is taking longer than that (which is rare, but can happen during peak times), it might cause the flow to time out. You can try increasing the timeout in the HTTP action settings.
Looping Issue: Ensure there's no accidental loop in your flow. For instance, if the flow is set to trigger on a Teams message and then sends a message back to Teams, it could re-trigger itself.
Flow Concurrency: Power Automate has a limit on how many flows can run concurrently. If you have other flows running at the same time, it might cause a delay. Check if there are any other heavy flows running on your tenant.
API Rate Limits: OpenAI has rate limits for API requests. If you're making a lot of requests in a short time, you might hit these limits. You can check OpenAI's documentation for the specific rate limits based on your subscription.
Error Messages: Even though you mentioned there's no specific reason, it's worth checking the flow's run history in Power Automate. Sometimes, there might be a subtle error message or warning that can give a clue.
Flow Structure: Double-check the flow structure. Ensure that after the HTTP request to ChatGPT, there's a proper action to send the response back to Teams.
If you've gone through these and still can't pinpoint the issue, it might be worth recreating the flow from scratch. Sometimes, starting fresh can help identify any misconfigurations.
Hope this helps
Hey @@samuelboulangerAI! Thanks for the reply! Everything was set up perfect, but I wasn't 100% sure if this would work on a free OpenAI account (it doesn't...); my error was 'You exceeded your current quota, please check your plan and billing details.'
Awesome video, very clear instructions, worked perfectly the first time, thank you so much!
Thanks Lisa, happy it worked the first time :)
Thanks for a great video! Please make video on how to keep context. And also include how to avoid Power Automate to loop if trigger-keyword is mentioned in reply.
Great content Samuel thanx for sharing these tips !
My pleasure!
Funny as I tried with two tutorials, and if I put the dynamic contents like this, it will reply again and again until I turn the flow off. Then, when I put "Reply to message ID" and "Message ID" vice versa, it works fine. Wonder why is that..?
Will others in the teams system be able to see all the questions and answers from said CHATGPT that were done inside CHATGPT online or just for that particular channel?
Thank you. Is there any other discussion platform where ChatGPT can be integrated?
Sure, actually if you're using Copilot Studio which use chat GPT as the model behind you have a lot of supported channels. Other than then, considering we are making an API call, any platform that support the integration with power automate will works.
This is great - I wonder if you could go over how I would set this up so ChatGPT is providing answers based on content I've created. Ideally from a OneNote Notebook or even word doc?
Thanks for the video Samuel, it worked like magic. I have a question on how can we add local files in a workflow such a that the gpt model gives the information through the files as well.
Glad you liked it :) It depends on the type of the document, the lenght and structure. For a small document when only text you can use the OCR capabilities of Ai builder. For a form, form a recognizer as well in AI builder. For more complex scenarios you will have to explore something like azure search paired with azure OpenAI or another vector database to store the document.
Great video but your explanation on Authorization and Bearer were very unclear. Did you type this or pull from the list? I cant get it to work. Also, what model name? Im using model "GPT-4" but that snot working as a model in the BODY. Any ideas?
Thank you Great video , works with me 100%
Is there a way that the Teams message takes the text formatting of the ChatGPT response? It just posts messages in Teams with escape characters and Markdown tags for me. E.g. for code formatted with ```
Awesome video, but my flow always receive timed out testing .... Any ideas?
Awesome, works good for me !!
Glad it works :) Thanks for the comment!
Great video! Worked as expected. How would one go about integrating ChatGPT as a stand alone bot? My ideal situation would be to add ChatGPT as a user so if I have a question I can just message that user rather than everyone seeing what I am asking the bot in a channel. Any guidance is much appreciated :)
Thank you for your comment. Happy it worked properly. You can keep the channel private in that case. Or even embed it in a chat instead of a channel.
@@samuelboulangerAI is there anyway to set it up as a Teams user? Thanks for the reply!
Thanks for the video, Step-by-step instructions. I am not getting any errors but when I am testing, it's not giving me any output. it just shows in the running status
is it possible to give it a prompt and when you type something it will respond based on that prompt you previously gave (not when asking the question but before that, like "give answers in xyz format" and we wont need to type that every single time)?
great session, i have followed all steps but face the error "We could not parse the JSON body of your request. (HINT: This likely means you aren't using your HTTP library correctly. The OpenAI API expects a JSON payload, but what was sent was not valid JSON" not sure where went wrong? thankssss
Probably an issue with the characters, make sure to copy paste from their website. A lot of times it's a misplacement or wrong character of the '
Awesome integration, but incase we have custom GPT which read data from internal folder using LLM and going with data to chatgpt for NLP processing how we can automate the same ?
Hi I have an doubt. I have chat GPT account API with my personal email ID and i tried to setup this flow on my organization official email ID does this combination will work? please clarify.
Yes the authentication is done through the API key not the email, so no problem :)
@@samuelboulangerAI Thanks am getting the below error message on flow HTTP process
{
"error": {
"message": "You exceeded your current quota, please check your plan and billing details.",
"type": "insufficient_quota",
"param": null,
"code": null
}
can you let me know what should I do now ?
Great video, thanks Samuel, however Im receiving the following error in my flow -
{
"error": {
"message": "'messages' is a required property",
"type": "invalid_request_error",
"param": null,
"code": null
}
}
If think the model you provided is now deprecated, is there are a substitute. Thank you
Amazing !
I don't see add dynamic content in my flow. Instead I get something says " Select a parameter then provide examples and we'll suggest an expression" so I am not able to add the expression in the last part for converting the body
If you don't have Microsoft Automate Pro, don't bother. HTTP POST requests requires a premium account.
@@pudgy_buns I do have Premium plan. I sorted it out. I had a different view because I was creating the flow from PowerAutomate and not from powerApps.
Was able to set this up following your instructions. Super cool! Thank you! Is there any way to get the Chat GPT reply to be named as "Chat GPT" and not "[name] via Power Automate"?
Happy you've been able to make this work :) It's possible but you'll have a to create a user called "Chat GPT" and use this user to send the message instead of the bot.
Here's a more detailed explanation on the community forum : powerusers.microsoft.com/t5/Using-Flows/Change-the-name-in-the-Via-Power-Automate-Message/td-p/1620106
Thank you for the video, i am facing an error in the http stage "ActionFailed. An action failed. No dependent actions succeeded" "Bad request" any clue?
would it be possible to add a query with power automate, for example, I'm on the phone with a customer, I need to know the status of a shipment, ChatGPT is linked to my audio call, hears the words "status and shipment", and automatically begins a chat with me and my s/r department head?
Is there a way to do this such that a person can post and then then instead of triggering a response from chatgpt, ai searches through a given list of websites instead?
With Power Automate, it would be really complicated. Basically to pass information from a website you first need to extract the data of the website and then pass it to Chat GPT in the prompt. An example would be to add in the prompt :
{Your question}
Answer only with information from bellow source :
#Sources:
{Data extracted from the website}
Hi there, I am getting an error when I try this but I am wondering if this is my work VPN causing the issue. Can you give me a way to contact you to figure out what is happening ?
Hey Cindy, you can reach out on LinkedIn : in/samuelboulanger/
I don't have the option to start a conversation only to start a post. Any idea why?
Actually there's been an update to Team since I made this video. So that's the normal behavior.
This flow doesn't work for me. it's just spins on the powerautomated flow test. Nothing reply back to Teams ? Any suggestion?
Sorry to hear you're having trouble. Let's try to troubleshoot this:
API Key: First, double-check if you've correctly added your OpenAI API key in the HTTP action. A common mistake is missing the "Bearer " part before the key.
Triggers: Ensure that you're using the correct keyword "chat GPT" in Teams to trigger the flow. Remember, it's case-sensitive.
Power Automate Flow Errors: Head over to Power Automate and check the flow's run history. If there's an error, it should give you more details on what went wrong.
Token Limit: Remember the token limit mentioned in the video? If your question and the answer combined exceed the token limit, the flow might fail. Try a shorter query to test.
Permissions: Ensure you have the necessary permissions in Teams to send messages. Sometimes, admin settings can restrict bots or automated flows from sending messages.
Flow Test: When testing the flow in Power Automate, make sure you're using the "Test" button and then selecting "I'll perform the trigger action" before heading to Teams to type your message.
If you've checked all these and it's still not working, let me know more details or any error messages you're seeing. We'll get this sorted out! 👍🔧
I cannot find "Reply to message Id" in my power automate flow.
Great video! I'm hitting a failure in my testing though. I get an Action HTTP failed - 'messages' is a required property. {
"error": {
"message": "'messages' is a required property",
"type": "invalid_request_error",
"param": null,
"code": null
}
}
Did you find a fix? Just hit a similar roadblock
Can you post your HTTP request body and I'll try to help?
Same problem - messages is required property. Did anyone find a fix?
It seems your using the Chat api instead of the completion API. Make sur you post to : api.openai.com/v1/completions
Excellent explanation. Is there a way we can add some csv file and ask questions related to that in Teams and get answers?
Thank you for your feedback, I'm glad you liked it. You want to drop a csv file in MS team, than retrieve the data and ask questions about the data, I'm I correct?
It would be feasible but too complicated to explain in a message. This will require to use the chat API instead of the completion API. You will need to start with a system message telling GPT to answer only based on the data you provide and then use the user message to ask the question.
@@samuelboulangerAI exactly. At some point it would be really helpful if you can add a similar content as well 😊
@@samuelboulangerAIYou could probably do like a part 2 of this video with features in consideration of all the comments 🤓 Great video btw, really helpful and professional 🙌🏻
How do I make it use GPT4 instead of GPT3.5?
In the model you should use GPT-4.
@@samuelboulangerAI it doesnt work, it says The model: `gpt-4` does not exist in the error prompt.
Would it be possible to replace OpenAI API call with Azure Open API? What needs to change?
Totally yes. Please refer to the documentation for more details : learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=command-line%2Cpython-new&pivots=rest-api
But since I released this video, we released Copilot Studio which make the integration so much easier
Just for your info: at 09:06 your key is shown... maybe not that good..
Thank you for taking the time to tell me. I deleted the key!
How can I train this bot on company specific information?
I followed your program twice step by step and I am not getting the chat reply in my teams, any advice?
thank you for the video. I have followed the all of the required steps but am receiving an HTTP error 'messages' is a required property. Is there any help here with this?
I'm not sure but I think you're using the chat API instead of the completion one. Are you posting to
api.openai.com/v1/chat/completions ?
How do I show citations on teams chat
The easiest way is to use Copilot Studio instead of this method.
ChatGPT starts answering this on loop after it answered my question:
Q: What is the difference between a Data Scientist and a Data Analyst?
A: Data Scientists are responsible for extracting insights from large datasets and using advanced analytics techniques to identify patterns and trends. They are also responsible for developing predictive models and algorithms to help organizations make better decisions. Data Analysts, on the other hand, are responsible for collecting, organizing, and analyzing data to identify trends and patterns. They use statistical methods to interpret data and create reports and visualizations to help organizations make informed decisions. Data Analysts are also responsible for creating data models and developing data-driven solutions.
by the way, i;m not using chatgpt as keyword. the account that was used in integration has bot and the keyword i;m using is bot
Got the same issue with another project using the API. I supposed It's a bug...
didnt work with me, I did all the steps
but when I post the question no answer is received
Do you have an error message in the power automate? It probably means it failed.
Not getting Reply from Teams , I don't have any license for Power Automate, is this because of that??
Awesome video! I tried it and it works. But there is only 1 problem, the flow cant be stopped lol ,
the chatgpt keeps spam messaging
'
Q: What is the difference between a Data Scientist and a Data Analyst?
A: Data Scientists are responsible for extracting insights from large datasets and using them to inform decisions. They use a variety of techniques, such as machine learning, statistical analysis, and data mining, to uncover patterns and trends in data. Data Analysts, on the other hand, focus on collecting, organizing, and analyzing data to answer specific questions. '
under the comment sections.
I have to revoke the API keys every time to stop the flow. Any suggestion to resolve this spam responses?
If you are using the "when a new chat" trigger you will need to add an exception for the flow user. If not the chat response from your GPT will trigger a new flow (causing an infinite loop). Use conditionals to except flow user from making a HTTP request. For example IF "from chat details get user of msg" contains flow user, don't send request.
Thanks for the feedback' I'm glad you liked the video. Have you specified to only run when the trigger word is used?
Is it possible for ChatGPT to respond based solely on my own document?
Not using the method in the video but generally speaking yes. Or you can use Microsoft Copilot as well. Simply drop the document and ask it to answer solely based on the information in the document.
My flow has been running for several hours. Any idea on how to bypass that?
Do you get an error message? At some point, it should time-out ans give you an error message.
Did you get past this issue? i think i am having the same issue...
Normally it means the power automate is simply not triggering. Make sure that you're on the proper channel and using the right keyword.
I got the same problem. The trigger dies fire off, but then gets stuck in the for each loop
Thank you for this video but I need help with this link you promised will go with this video. I can't seems to find it body('HTTP')['choices'][0]}('text'0)
That´s wrong. it´s: body('HTTP')['choices'][0]?['text']
@@anny9875 Thank you
@anny9875 Please help me when I tried to test, I got an error "There are no runs for this flow.". Kindly advise me on what I need to do.
You are right sorry for that. I was leaving for vacation and posted in a hurry. Just updated the description and here's the Fx : {body('HTTP')['choices'][0]?['text']}
You should select manually instead of automatically and then go in MS Team and test it directly. Then the flow will run and you will be able to select automatically next time.
It takes long time to run the flow in power automate and then it got failed. What to do
What's the error message when it fails?
Hi Same issue from my side. When testing manually I can trigger the flow but the flow looks Frozen on the HTTPS connector and then fails few minutes later 😢
@@stephanevasseur5544 can you share the error message?
@@samuelboulangerAI I found the source of the issue. My openAI account has been opened few weeks ago and the free tokens were expired.
I have an error on the HTTP part. When I run the test, it shows an error “Invalid URL (POST/v1/chat/completions”. All the input seems to be okay. Any ideas? Thank you
Here's ChatGPT answer ;) : This error message indicates that there is an issue with the URL you are using to call the OpenAI GPT-3 API from Power Automate.
Double-check that you have correctly specified the URL, including the correct API version and endpoint (in this case, "/v1/chat/completions"). Make sure that you have also provided the correct authorization credentials (i.e., API key or token) in your Power Automate action.
@@samuelboulangerAI I have the same issue as well with the Invalid URL . Is this because I am behind a proxy (cisco umbrella). Thanks for a very detailed video
Excelent! TY!
You're welcome!
It works fine. ChatGPT started to produce questions and and answers by its own! Very strange!
That must be an interesting conversation! If you don't mind, I'll be curious to have some screenshots. You can reach out on LinkedIn : samuelboulanger
Great session and thank you. I followed everything but received an error message:
Action 'HTTP' Failed.
Error Details: The provided input content is not valid: 'The provided content-type header value 'application-json' is not well formed.'.
What am I missing and thank you again!
FYI, Under HTTP, additional error message below:
BadRequest. The provided input content is not valid: 'The provided content-type header value 'application-json' is not well formed.'.
@@JAIWeLearn I need a solution here as well.
It's an error in the JSON you use in the HTTP request body. Make sure you copy-paste it from the documentation.
hi can you help me. mine says 'messages' is a required property
Thanks for reaching out! Let's work together to tackle this issue. Have you tried troubleshooting it yet?
I got error - Flow save failed with code 'OpenApiOperationParameterValidationFailed' and message 'Input parameter 'requestBody' validation failed in workflow operation 'When_keywords_are_mentioned': The API operation 'WebhookKeywordTrigger' requires the property 'requestBody/channels' to have more than '1' items but has '0' items.'.
I think it might be the expression to retrieve the body. Make sure to copy paste it from the video description
Parse json flow also works pretty well on this. It can also be automatically generated from a sample, but this seems to be much simpler and it works
@@samuelboulangerAI I also have the same issue have put the word by word as per your flow still the issue is same
Failed on ' HTTP' HOW can o fix this?
Hey, it works great for me, is there a way for the bot to ignore the keyword on its replies otherwise it gets stuck in a loop?
I'll say that the easiest way would be to change the keyword that trigger the flow to something you know for sure won't be used in the answer. Like " HeyChatGPT" instead of "ChatGPT".
its not working for me . Also its not showing error
Awesome tutorial only a thing... this expression {body('HTTP')['choices'][0]?['text']} doesn't work for me? anyone who got a solution? anyone who encountered this error too?TY
Hello! Get rid of your beginning and end brackets; just put this:
body('HTTP')['choices'][0]?['text'
This is great, set it up, worked fine for about 5 messages, now it's stopped reacting. The trigger word no longer to seems to kick off the flow, plenty of free credit on my account left too, this happened to anybody else?
That context retention part is a bit of a dealbreaker. However, pretty cool video nonetheless.
Thanks, you could keep context by using a database to keep the different interactions, dataverse, or even Excel. I'm still experimenting.
@@samuelboulangerAI Indeed, it’s a hassle though. Once you’ve had a chat with ChatGPT on a particular topic, it’s incredible how the conversation flows. It even references it’s past responses for clarity. It’s very natural and conversational without this context - it’s hard to put in the effort otherwise. Pretty good that we can have some functionality in steams though.
error Definition contains invalid parameters.
Very good video but I cannot make this work for the life of me. The flow fails with a "ActionFailed. An action failed. No dependent actions succeeded." and Bad Request in the Get Message Details action. No clue what is wrong.
I have a Power Automate Premium license and using Teams version: 1.6.00.24078. I noticed that I can no longer start a conversation in Teams but rather a "Post" and I must put a headline,etc.
Any thoughts?
Hey @jasoncardinal7502! Thanks for the feedback on the video. Let's tackle this step by step:
"ActionFailed" & "Bad Request" in Get Message Details: This error typically indicates that the flow is having trouble retrieving the details of the message from Teams. Here are a few things to check:
Message ID: Ensure that the "Message ID" you're passing to the "Get Message Details" action is correct. It should be dynamically populated from the trigger.
Permissions: Double-check if the account linked to Power Automate has the necessary permissions to read messages in Teams.
Teams Version & "Post" Issue: The version you're using seems relatively recent. Microsoft Teams has been evolving, and some features or UI elements might change. If you're seeing a "Post" option instead of starting a conversation, it's likely a UI update.
Headline Requirement: If Teams is now requiring a headline for a post, it might be affecting how the flow is triggered. The flow might be expecting a simple message, but if Teams is now structuring it as a post with a headline, body, etc., it could be causing the flow to fail.
Workaround: For now, try using the "Reply" option in an existing conversation in Teams to send a message without a headline. This might allow you to bypass the new "Post" structure and trigger the flow as intended.
Hang in there! Tech can be a bit finicky sometimes, but we'll get this sorted. Let me know how it goes! 👨💻🛠️🚀
its takening ages to respond?
I hope you modify this one since ChatGPT Depreciated some of the oldway of doing things
Thank you for your comment. Unfortunately, UA-cam doesn't let me edit a video after it's been uploaded. I can cut a section, but that's it. I would probably make an updated version.
Nice but I'd recommend having your org pay for the API key and not your personal :D
Totally agree :)
Hi Nice Video But i need help that is why every time we need to type ChatGpt for asking Queries , Why Can't we ask direct like not using any keyword , It is Possible , if it is can you explain please.
Used to do similar project to this myself some months back. Basically just use a trigger: "when message is posted" in teams instead of the conditional "if keayword are mentioned, then pick the message"
Sure, you can easily do that. As Joel mentioned, just change the trigger to "when message is posted".
Great video! Thanks for the great information. I keep running into the flow test being fine but then when I try it in teams, nothing populates in teams and I go into power automate and have an error message that is in the when keywords are mentioned and says this in the body {
"responses": [
{
"id": "0",
"status": 404,
"headers": {
"Cache-Control": "no-cache",
"Content-Type": "application/json"
},
"body": {
"error": {
"code": "ExtensionError",
"message": "Operation: Create; Exception: [Status Code: NotFound; Reason: Resource not found]",
"innerError": {
"date": "2023-05-13T01:59:19",
"request-id": "58",
"client-request-id": "5baa8"
}
}
}
}
]
}
I deleted the api key for posting this :)
Here's what Chat-GPT told me based on your error message ;) :
The error message you provided indicates that there is an issue with a resource not being found. It seems like there is a problem with the "Create" operation in your flow. Here are a few steps you can take to troubleshoot and resolve the issue:
1. Verify the connection: Check if the connection to the target service (Teams, in this case) is properly established in your Power Automate flow. Sometimes, connection issues can cause resource not found errors. Make sure the connection is active and authenticated correctly.
2. Check for missing or incorrect inputs: Review the inputs you're passing to the "Create" action in your flow. Ensure that all the required fields are provided and their values are valid. Double-check the configuration of the action to ensure it matches the expected inputs of the Teams API.
3. Validate permissions: Ensure that the user or account used for the connection has the necessary permissions to perform the "Create" operation in Teams. Insufficient permissions can lead to resource not found errors. Verify the permissions and consider using an account with appropriate access.
4. Confirm the existence of the resource: Double-check that the resource you are trying to create (such as a channel, message, or other Teams entity) actually exists. If the resource is deleted or does not exist, you may encounter a resource not found error. Make sure the resource is available and accessible in your Teams environment.
5. Review the request-id and client-request-id: The error message includes request IDs ("request-id" and "client-request-id"). These IDs can help with troubleshooting the specific request that resulted in the error. You can use these IDs to investigate further in your service logs or contact the support team of the service you're integrating with (e.g., Microsoft Teams) for assistance.
By following these steps, you should be able to identify and resolve the issue causing the resource not found error in your Power Automate flow.
I get this JSON error back from Power Automate:
Unable to parse response JObject:{
"responses": [
{
"id": "0",
"status": 400,
"headers": {
"Cache-Control": "no-cache",
"Content-Type": "application/json"
},
"body": {
"error": {
"code": "ExtensionError",
"message": "Operation: Create; Exception: [Status Code: BadRequest; Reason: Bad Request]",
"innerError": {
"date": "2023-02-14T20:07:55",
"request-id": "2d5449b9-eade-4a8b-ab3f-01c90cf159ff",
"client-request-id": "2d5449b9-eade-4a8b-ab3f-01c90cf159ff"
}
}
}
}
]
}
At which actions does it gave you the error?
@@samuelboulangerAI It appears the error occurs when ChatGPT is overwhelmed and unable to respond. Otherwise it works. Thank you for the video.
Thanks for the tutorial - after setup I get below reply - can you help -
package com.example.demo.controller;
import com.example.demo.model.User;
import com.example.demo.service.UserService;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.web.bind.annotation.*;
import java.util.List;
@RestController
@RequestMapping("/user")
public class UserController {
Will need a screenshot to help you with that. You can reach out on LinkedIn. Samuelboulanger
@@samuelboulangerAI I get this back- I think its a similar- error like Java? totally unexpected ! package com.example.demo.controller;
import com.example.demo.model.User;
import com.example.demo.service.UserService;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.web.bind.annotation.*;
import java.util.List;
@RestController
@RequestMapping("/user")
public class UserController {
@Autowired
private UserService userService;
@GetMapping("/all")
public List getAllUsers(){
return userService.getAllUsers();
}
@PostMapping("/add")
public void addUser(@RequestBody User user){
userService.addUser(user);
}
@PutMapping("/update")
public void updateUser(@RequestBody User user){
userService.updateUser(user);
}
@DeleteMapping("/delete/{id}")
public void deleteUser(@PathVariable int id){
userService.deleteUser(id);
}
}