AI

I have a strange issue while creating copilot. I want the copilot to read data from one table only. The table contains creation date, closure date, case id. While simple questions like created cases in June are ok for the copilot to handle, wekely or daily trend charts are not produced. The copilot can’t extract week or month no from date columns to produce the trend charts. AI services enabled are: Claude 3.7 Sonnet (AWS)GPT 4o (Azure OpenAI) From time to time the copilot get’s frozen while thinking, it can take up to few hours for it to spin without results. Anyone faces similar issues? 

Webinar Sessions | Celonis x AI: 𝗙𝗿𝗼𝗺 𝗕𝘂𝘇𝘇𝘄𝗼𝗿𝗱𝘀 𝘁𝗼 𝗔𝗰𝘁𝗶𝗼𝗻𝗮𝗯𝗹𝗲 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝘀AI is trending everywhere but, how do you turn hype into real business impact?Join us to explore:- Showcase of AI capabilities in Celonis, featuring a hands-on demo for service tickets.- Processand capabilities scalable for AI deployment and operation- How to get started and initiate pilot use cases📅 19 September & 26 September | 🕒 10:20 - 11:00 AMSpeakers: Federica Tella (Consultant) & Edoardo Puddu (Data Scientist)𝗦𝗲𝗰𝘂𝗿𝗲 𝘆𝗼𝘂𝗿 𝘀𝗽𝗼𝘁:- 𝟭𝟵𝘁𝗵 𝗦𝗲𝗽𝘁𝗲𝗺𝗯𝗲𝗿 𝟮𝟬𝟮𝟱:https://lnkd.in/duYZ-sQB- 𝟮𝟲𝘁𝗵 𝗦𝗲𝗽𝘁𝗲𝗺𝗯𝗲𝗿 𝟮𝟬𝟮𝟱:https://lnkd.in/dubVmgsf  

  Latest trend: #AI 

How can I get the process copilot to use the record attributes saved on a knowledge model to search for the “output” rather than using the sample from the prompt? I have tried removing the sample from the prompt and it does not work. I have used the attributes on the knowledge input section of the process copilot as well and it also seems not to be enough.

Hi everyone I have a challenge that I’m trying to figure out related to pycelonis.I’m currently trying to get the component values from a view using pycelonis. Unfortunately, I couldn’t find any information inside the documentation that can help me with this.It seems that the documentation only explains how to work with analysis or knowledge models. Does anyone know how to work with components of a view inside pycelonis?Best wishes!

I have generated a zip file with the content-cli module and now I need to download it locally, but every time I press the “Download” button it creates an HTML file. That HTML redirects me to a “Secure download” which re-downloads the same HTML. How can I avoid the HTML and just retrieve the zip file?

Hi everyone,I want to use the Annotation Builder, but I don’t have a unique ID.For example, the same material number exists in two different business units as you see in the picture.What’s the best way to handle this? Thanks for your help!  

I am unable to get the option to schedule a notebook in my ML workbench.Does it require admin access to get the option?  Documentation says to have a valid executable file, which I have. Any code to be written in the notebook to make it eligible for scheduling?  

Hi all,I want to use the Process CoPilot with the built-in python tool in order to execute ad hoc calculations.https://docs.celonis.com/en/tool-activation.html#UUID-3c37cb75-0631-254f-a5b1-1a70ffd8d5f4_UUID-5523d1d0-d4b7-e2d3-4ebe-f999dffd6357and https://docs.celonis.com/en/tool-activation.html#UUID-3c37cb75-0631-254f-a5b1-1a70ffd8d5f4_UUID-b801ca95-08b6-fe38-c5c5-89bffe3f22aaI don’t understand the examples...My question:HOW can I implement this? HOW is the data going to be accessed and loaded in order to generate a dataframe? The examples in the docs don’t help me at all … Why is a df generated from a “KPI”? There is no relation between “get_data” output and the “code” …  HOW should I use get_data and python? HOW do these tools interact?Thank you in advance and best regardsJan  

Dear Celopeers, Currently I am trying to extend the simulation capacity of Celonis in the Machine Learning workbench. To do so, I need to import an event log through, I think, PyCelonis. I would like to use the UI of the Process Explorer in a premade analysis of Studio to define the activities and connections that are imported.The following link suggests this used to be possible within PyCelonis 1.7 using the ‘Share your selection’ URL. (Analysis: Pull data - PyCelonis) This appears like a very elegant and user friendly solution to me.However, I can't figure out how to do so using the newer PyCelonis 2.12. I did find out how to navigate to the correct space, package, and analysis through different IDs. But the Python methods attached to the analysis object created through .get_analysis() do not make apparent how to perform the filtered data extraction I alluded to above.Any ideas or references on how to do this? Thank you in advance, Bas

Hi all, is it possible to code in the Academy Studio or not. For me it always show that there is no data model or data pool that i can use. Thanks for your helpbest regards Louis

I am working on a script in PyCelonis to search assets and compile a list of the data models/Knowledge models that the assets are using, with the aim of non manually figuring out which DM/KMs are currently in use by production assets.  Using PyCelonis I have been able to navigate down to the asset level and find lots of information for each asset, but have so far not found anywhere that you can identify the DM/KM an asset is built on. The “.dict()” function provides a lot of information and I thought would be the intuitive place to include this kind of information but it is nowhere to be found, except when the asset is a view and the KM gets referenced in the yaml code that is provided in “.dict()” Does anyone know if this is information that we can pull from PyCelonis? And, if so, how to find this information on the asset level? Thanks in advance for your help!

Tried with Python script to extract table from the web page, however it didn't work.

I just completed the "Comparative and Predictive Process Mining" course to learn more about ML capabilities, prediction, simulation, etc. I made an assumption that the example notebook/code that was shown in the video regarding feature recommendation and model training would be OOTB or readily available. I cannot find any examples on Celonis site or web. Are there examples for use or is it on us(me) to develop from scratch?

Hi Team,Im trying to convert the data i need from the datapool into a pandas dataframe for further calculation.In the documentation i could only find how to read the parquet files and convert into Pandas.Is there any way?Let me knowThanks 

Hi all,I want to extract data from a table in the schema explorer, which is loaded via an extraction, the standard way is to do it is by loading the table into datamodel and pulling the data from the datamodel using pycelonis as mentioned in the pycelonis documentation I am curious to know if there is any other way to do it without loading the table into datamodel (via the schema explorer if possible)

Hi all! First time here, but god a question. I am running a python script (pycelonis 1.7.6) that needs to push data into a table. Everything work fine unless the data push:  datamodel.pool.create_table(df_or_path = res,table_name = 'TABLE_NAME',if_exists = 'drop',column_config=column_config)  which returns: [2024-07-23 16:26:09] INFO: Data push job status: QUEUED...[2024-07-23 16:26:14] INFO: Data push job status: QUEUED...[2024-07-23 16:26:19] INFO: Data push job status: QUEUED...[2024-07-23 16:26:25] INFO: Data push job status: QUEUED...[2024-07-23 16:26:30] INFO: Data push job status: QUEUED...[2024-07-23 16:26:35] INFO: Data push job status: QUEUED...[2024-07-23 16:26:41] INFO: Data push job status: QUEUED...[2024-07-23 16:26:46] INFO: Data push job status: QUEUED...[2024-07-23 16:26:51] INFO: Data push job status: QUEUED...[2024-07-23 16:26:57] INFO: Data push job status: QUEUED... Same functionality on another table worked flawlessly, for this specific one it queu

Hi Team,I need assistance with building a Generative AI-based use case in Celonis. As a starting point to understand the various functionalities of Celonis, I aim to develop a basic chatbot system.Here’s where I need help:I want to create a view in Celonis that mimics the look and feel of ChatGPT, where I can ask questions related to the data and receive responses.I have successfully integrated OpenAI with the Machine Learning Workbench, but I’m not sure how to trigger the Machine Learning Workbench from my Celonis views using a button click and display the response appropriately.As I am new to Celonis, I would greatly appreciate a detailed explanation.Thank you for your help. TLDR : Help me understand to built the UI, where i can write in input box my query, on click the button i want to run my ml work book code and then show it as a text response on my views(my ml workbook has open-ai integrated)

Hello everyone, I am trying to call a ML python script from an Action Flow with some parameters. Do some math with it, and give a response to the same Action Flow.Following the documentation in Trigger Machine Learning Script (celonis.com), I have been able to call the Python code, pass some parameters, process them and then call a NEW Action Flow. Is it possible to make the first action flow wait for a response and make a GET or something similar so you continue in the same action flow? Thank you,Federico Arribas. 

I have read the documentation on the Run ML Notebook Component but I don't know how to publish the results of the notebook back to the Analysis UI. Is there any link to documentation that shows how to publish the results back on the Analysis? Do I need to use pycelonis or other python library?

Hi All,Earlier I had asked a question regarding how we can save/extract a table in our EMS as a csv file in our machine learning work bench, seeing the reactions or the lack of replies I got for the topic I think it would be helpful if I explain the code that I used/found to save the file as csv in my ML workbench. from pycelonis import get_celonisimport pycelonis.pql as pqlfrom pycelonis.config import Config celonis = get_celonis()data_pool = celonis.data_integration.get_data_pools().find("Datapool name")data_model = data_pool.get_data_models().find("Datamodel name")Config.DEFAULT_DATA_MODEL = data_modeltables = data_model.get_tables()table_name = data_model.get_tables().find("Table to find out")table_columns = table_name.get_columns()df = pql.DataFrame(  {    "column1": table_columns.find("column name in the table which has to be extracted "),    "column2": table_columns.find(" column name for column 2 "),    &quot

Hi All,'In the past couple of days we are facing an issue in the ML workbench, while running a specific python code which is long and also complex , the kernel suddenly becomes unknown from the busy state due to which we have to rerun the code again from the beginning and this goes on happening as a loop ,we have raised a ticket regarding this to celopeers, but would also like to know if someone had faced this issue before.

Hello everyone, Is there a change log / release notes of PyCelonis somewhere? This would be very helpful to make decisions when to upgrade to a new version. Thanks a lot!Pablo

Hello Community, today, we upgraded one of our ML Workbench to PyCelonis 2.6.Since then DEBUG informations are shown (see below). In the Jupyter Notebook we use: logging.getLogger("pycelonis").setLevel(logging.WARNING) If we execute the notebook on local computer, no DEBUG information if shown. On the other hand, if we execute it on the EMS, following DEGUB informations is shown: [2024-01-31 13:39:56] DEBUG: Request: 'POST' -> '/pql-language/api/lsp/publishDiagnostics/batch'[2024-01-31 13:39:56] DEBUG: Request: 'POST' -> '/pql-language/api/lsp/publishDiagnostics/batch'[2024-01-31 13:39:56] DEBUG: Initialize Series.[2024-01-31 13:39:56] DEBUG: Request: 'POST' -> '/pql-language/api/lsp/publishDiagnostics/batch'[2024-01-31 13:39:56] DEBUG: Initialize Series.[2024-01-31 13:39:56] DEBUG: Request: 'POST' -> '/pql-language/api/lsp/publishDiagnostics/batch'[2024-01-31 13:39:56] DEBUG: Initialize Series.[2024-

Hello all,we are still using PyCelonis 1.7 and we figured out that since yesterday the response of the function "api_request" has been changed.Before we got a list / dict or string. Since yesterday we get a request.model.response object. Have you noticed that? Are you also having issues with your api_request calls? Our current workaround was to get the "text" of the response object. Like: resp.text instead of: resp