https://discord.gg/botpress logo
Join Discord
Powered by
Hi bro, I'm stuck in the third part of Gordy with ...
# 🌎general
h
Hi bro, I'm stuck in the third part of Gordy with the video of the recipe bot, I've already checked my records three times and it's as it is, also when I put the message on the label it takes time to answer gpt. I can't continue I'm stuck
s
This is for the contextual questions part i.e. using the AI task?
Please can you send me a screenshot of the AI task and the result. Thanks.
h
this is the problem that happens
That was the first thing that happened, I send the message and it does not call gpt, I have the spoonacular api installed correctly with the name miapykey, then I put the code, for the reason that it did not call the message and then I got the error as in the first screenshot
s
Thanks that's great! I think there was a similar issue yesterday. Please check if this is the solution: https://discordapp.com/channels/1108396290624213082/1137431150076375261/1137431150076375261
h
Bro, do you mean that I replaced the code but not the message?
s
Yes, try replacing just the code. Maybe @Ruchi can help.
h
same problem, of course jb also told me, download the template, I already have it, but I'm trying to do it myself, because the clients will be different, because I have checked my variables and my code is fine,
the apikey, I changed it because mine is configured with the name miapikey
s
ok, hopefully one of our devs or someone from the community can help you asap. Out of interest, what was different about JB's replacement code versus the code in the template?
h
It is practically the same code, everything is already in the template until the end that one is downloaded.
justin a query when one contracts a botpress plan and wants to make these bots for different companies in different areas, botpress puts developers to solve this type of code as in my previous case as an example.
a
So a 401 error happens from bad authentication. Have you saved your spoonacular API key as a config variable? https://botpress.com/docs/cloud/studio/variables/#how-to-make-configuration-variables
h
I did it the same as in your video in which it would be failing
a
you need to click "show API Key" before copying it, otherwise you'll get a bunch of `****`'s
h
Thank you very much I am going to continue with this bot, any inconvenience I will send my doubt,
Hello, I'm already in part five, but when I ask the bot to answer for the recipe, it gives me a lot of numbers, id, a lot of text and then it gives me the images. Is there a way to remove it? other time when i ask gpt it doesn't give me much context or content like in video this is because of gpt version or notice.
a
Hey @handsome-jordan-74538, double check the text content cards before the carousel and see if your bot is saying any variables (like workflow.recipeInfo) to the user
h
If inside a card, I create a variable that says RecipeInfo, do you mean that I delete it?
bro, i just removed it from this variable, and it works
hello @acceptable-kangaroo-64719 , I was able to create the chain of thought in another flow, it helped me to be reviewing other videos, but since I create the parentheses [... Recipes, chathistory, etc. The variable does not allow entering symbols and that because you put brackets has some meaning.
a
I don't understand. You are trying to name a variable with a name that includes special characters like 
(
or 
[
?
h
hello @acceptable-kangaroo-64719 I was able to solve it, it was with the array variable, now when I am in the chain of thoughts I see that the output is not saved in the objects of the step, due to the questions I asked the bot, and it also does not answer me or recommend me when asked ask about calories, could you help me with this part.
a
Is the output being saved to any variables?
h
the output does not save, I have put the same parameters of the main template, and it does not respond as it should at the end the bot does not give much information as in step 3, also code appears in the conversation
a
You can remove the tags and replace them with a blank newline. It looks like it is the last AI task that is having issues, have you tried debugging the AI task in "make-and-tell-answer"?
h
How to remove tags in which node I replace it, AI debugging task means there are multiple AI tasks in the chain of thought, one is divided into steps, and the other is at the end which of it I debug
a
How to remove the tags
This is the AI task to debug. When I say "debug," I mean do these steps:
1. Check if the prompt, inputs, and output variables are correctly set 2. Check if the variables used as inputs are properly set 3. Check if the output variables are filled 4. Check if the output variables are filled with the correct data 5. Ensure that the prompt + input variables does not exceed 4000 characters 6. Check if there is a task example, and if there is, make sure it is a correct example 7. Ensure the output of the AI task is being used correctly in the flow
There's more after this, but that's a good start
h
hello @acceptable-kangaroo-64719 , I just reviewed it several times, everything you mentioned up to and including, all the steps come up like these, the questions I ask the bot, in the event debugger it does not appear as it appears in the original video, and I was putting all the your parameters with the instructions of the AI with the original template, if you could review this capture and I will send you my file please.
I always start with the classic questions hello, number of people and no restrictions. and hence the questions you ask.
a
looking at the screenshot, it looks like the bot is working fine. There have been a lot of UI changes since the original video was published, so I'm not surprised that the debugger is different. But as long as the bot works, that's fine, and it looks like the bot is working
h
hello and regarding the content, he tells me he doesn't know when I ask him, he doesn't give me much context as in yours it refers to the AI, the version or I have to modify something.
a
There has also been a model update from OpenAI since the recipe bot videos were released, and that likely has an impact, too. have you tried modifying the AI task prompts to get the results you want?
Some things to try: * Play around with the words used in your prompt. Tell the bot explicitly to answer the user's question as best as it can, explain inputs, etc. * Add delimiters and identifiers to inputs. In my prompt I put quotes around "chat history" and then identified the input as "chat history" * Change the temperature * Make sure your summary agent is on and there's enough room to handle all the context * Add descriptions to your output variable
h
I have seen that you put a node that disables no-api-key that must be there for some purpose because when I activate it it no longer responds to me even though it already has the key recorded
ok I will take into account what you told me, and where do I go to change the language
a
the no-api-key is there to ensure that there's a spoonacular API key saved to 
env.apiKey
best way to change language is with the translator agent, then use code to set 
{{user.TranslatorAgent.language}}
to a two-char language code like "en" or "es"
f
@handsome-jordan-74538 I would recommend you to ask Gordy since he sent you the advice. Here is my interpretation on it. Play around with the words used in your prompt. Tell the bot explicitly to answer the user's question as best as it can, explain inputs, etc. I believe this would be in the AI task. Try telling it to explicitly to answer the user's question as best as it can, explain inputs and you could try telling it to do more. Add delimiters and identifiers to inputs. In my prompt I put quotes around "chat history" and then identified the input as "chat history" Got no idea about this one except that he want you to put quotes about your inputs. I dont now which inputs. Change the temperature I think he is refering to the tempature of the response from the AI language model in the AI task. I would refer you to here: https://botpress.com/docs/cloud/toolbox/ai-task/#temperature Make sure your summary agent is on and there's enough room to handle all the context When in the studio of the bot --> Go to agents in the top left side --> Select the Summary agent --> Enable the agent --> Make sure you have enough tokens allocated to handle all of the context the bot would need. Same goes for lines. Add descriptions to your output variable I think this one is pretty clear 🙂
Best of luck
h
Hello @acceptable-kangaroo-64719 , I have raised a question to 200 tokens and raised the temperature and when I ask are the other vegan recipes, the context to which this is due still does not give me a good idea, please, I have also changed the AI instructions in the GTP prompt-chat node
I have also noticed that the emulator is now red in color
a
When the emulator is red, that usually means there's an error in the code. You should double-check your code and make sure there's no errors. Also, you can lok in the event debugger bottom panel and click on individual AI tasks to see their inputs and outputs. This can help you verify that the AI tasks are getting the information that they need, as well as outputting the correct info.
h
@acceptable-kangaroo-64719 in the debugger event, you can check which node you're failing on or the code, because I only incremented the tokens, didn't modify the codes, and only changed the language, also recently I've noticed that when I start Studio Cloud I get a little window that says: There are 2 documents in the database that are not used in your knowledge base and may affect the responses of your bots.
I just checked the code in the prompt-gpt node, I changed the instructions of the AI as I show in the screenshot and the emulator does not change color, I also put a single card in the language change, I don't know if it will be fine or I will have than create another flow.
a
did this fix whatever issue you were running into?
h
There are two points: -The emulator is still red, I put a video, you will see that the code changes - the other point is increase the tokens to 200 and the summary agent is enabled and One of them was without a chain of thought.
a
For the red emulator, if refreshing the page doesn't fix it, try exporting the bot and then re-importing it.
Additionally, we've implemented a caching layer to speed chats up, so if you keep prompting with the same message, a new answer won't be generated
So you can try asking a new question or a new recipe
5 Views
PreviousNext
Powered by