Create a Powerful RAG Chatbot in 10 Minutes with IBM watsonx | No Coding | IBM Business Partners

[@saiabhilashbabu6382]

This is fantastic. Good Job

[@shenglan-channel]

Thank you very much!

[@fredericobarreto5921]

Thanks, what a great demo! Very helpful

[@shenglan-channel]

My pleasure!

[@shaiksaadabdultawwab8170]

Thank you very much, it helped me a lot.

[@shenglan-channel]

Glad it helped!

[@ShahilNikhailJagjivan]

do you know of a way to get rid of the privacy feature?

[@shenglan-channel]

Could you elaborate more which privacy feature you refer to? Is it AI guardrails from watsonx.ai?

[@SKini18]

Hi Sheng, can we build RAG to retreieve information from excel files using langchain and watsonx ? (Without discovery as it is not free). For vector databse it should use local memory or chromadb. Is there a way you can create a demo or if you already have any running project..if you can share the git link

[@shenglan-channel]

Hi SK, thanks for your question. Yes, here is a example: dataplatform.cloud.ibm.com/exchange/public/entry/view/d3a5f957-a93b-46cd-82c1-c8d37d4f62c6?context=wx I am thinking to create a video to demo how to use watsonx.ai in the popular library, such as LangChain, LlamaIndex.

[@Rafaelgarbulho]

how users can provision resources w/o the techzone?

[@shenglan-channel]

After you sign up IBM cloud, there will be free tier/trials for the service mentioned in the video.

[@faizziplays]

Hi, I am facing issue, in Watsonx Assistance in Actions "Invoke watsonx generation API" it is not getting set, It says Entension not set. I followed your tutorial. Other action is resolved but this is to.

[@shenglan-channel]

Wondering if you try to config the watson ai extension as the following image, thanks github.com/lansheng919/watsonx-RAG-Demo-action/blob/main/watsonx%20ai%20extension%20parameter%20config.png

[@jyotiranjanmallick5967]

Very Nice👍

[@shenglan-channel]

Thank you! Cheers!

[@ataishnehra6321]

Hi, the watsonx ai openapi file you are using in the demo is different and the one in the description is different. Hence a lot of people are facing errors with “Invoke watsonx generation api”. The link you have provided is the default file available on ibm documentation and it does not go with your custom actions file. Can you share the watsonx ai openapi json file you used in the demo please?

[@shenglan-channel]

Thanks for the info! Just uploaded.

[@ShahilNikhailJagjivan]

@@shenglan-channel is it the screenshot? or is it a file. The link would be highly appreciated🙏

[@shenglan-channel]

Hi @ShahilNikhailJagjivan, please check the video descriptions, it has the links to the file on github repo. Thanks

[@lichun00868]

Hi Sheng, thanks for the great sharing! it seems the JSON of watson assistant RAG actions doesn't work now, may I learn is there any version change during these two months?

[@shenglan-channel]

Hi Li Chun, please try this action json file: github.com/lansheng919/watsonx-RAG-Demo-action which I used in the video. Feel free to reach out if any questions ([e-mail]). Thanks.

[@mubeenarasack3625]

@@shenglan-channel the url is not working

[@cammediname]

​Thanks @shenglan-channel for the video.@@mubeenarasack3625 the link works, remove the final ), unfortunately it also doesn't let me load the .json file in watson assistant like @lichun00868, it generates "Internal Error"

[@shenglan-channel]

Hi @Cammediname, thanks for your comment, I add more details in the video description, which maybe helpful for you to troubleshot. Thanks

[@javiercalderon6480]

Hi, I´m having the same problem. Did you find the way to solve it?

[@LVs-lq4mk]

Thank you very much for this wonderful introduction on how to set up the chat bot! I still have a question about the capabilities. I want to use a chat bot to ask customers specific questions and have the responses analzed and stored in a standardised way in a data base. For Instance: The chat bot asks for the satisfaction and then analyzes the response to give it a numerical value from 1 to 5. Is it generally possible to use watsonx in such a way? How do I develop such a tool? Any help would be much appreciated!

[@shenglan-channel]

Thanks for your kind words! Yes, watsonx assistant is capable for this more structured conversation. Check how to build action in watsonx assistant should help and get you start to build the chatbot, here is the link:cloud.ibm.com/docs/watson-assistant?topic=watson-assistant-build-actions-overview Let me know if any more questions, thanks!

[@LVs-lq4mk]

@@shenglan-channel awesome, thank you very much, I really appreciate it!

[@muni2363]

Hi Sheng, I have followed the same instructions but unfortunately I see missing extension error for "Invoke watsonx generation API" .. is there anything needs to be fixed. However other 2 are fine. please advice

[@shenglan-channel]

Thanks for your questions, maybe try to inspect your extension to get more information to debug. Here is a link to give your more information how to do it, cloud.ibm.com/docs/watson-assistant?topic=watson-assistant-call-extension#extension-debug

[@LIFEOFCL_]

What’s chatbot provider is better botpress voice flow or ibm?

[@praasshant18]

IBM

[@Sai-ns2sk]

Can i hopedully use this on the Free plan?

[@shenglan-channel]

Yes, I think the watsonx assistant has a free tier, and watson discovery has free trial.

[@adrian-maulana]

AI that answers questions sometimes still hallucinates; is there a way to address this issue? And also how you can get the "for more context on this answer please see" info in the assistant? Thank you very much

[@shenglan-channel]

Great questions, one of the main purpose of RAG is to reduce hallucination by providing context. But still, I haven't seen a solution to remove hallucination 100%. Most of approaches I have seen are to "beg" LLM not to answer if the context is not relevant in RAG solution and always check the response. For more context part is via Watson Assistant (WA), you can manipulate the context variables in WA and display it in the response. I can share you the action Json for more details, feel free to email me ([e-mail]), Thanks!

[@jdsvic]

The web crawl for the 2022 pdf takes a very long time, cant we just upload the pdf?

[@shenglan-channel]

Yes, upload pdf is another option!