[BUG] [VSCode Extension] Package not installed

[tvanvalkenburg]

I am getting an error saying "Can't find tool promptflow_vectordb.tool.common_index_lookup.search". The image below shows that the package is installed, along with promptflow and promptflow-tools. The text printed to the terminal is the output of pip list in the same virtual environment in which the flow is being executed.

Running on Mac OS Sonoma 14.3.1.

[Kevinlee49]

It seems to be a duplicate of #788

And if you use the Vector Index Lookup tool, it has been deprecated. Therefore, I recommend you to upgrade to the new index lookup tool.

It is well written in this link.
https://learn.microsoft.com/en-us/azure/ai-studio/how-to/prompt-flow-tools/index-lookup-tool

[D-W-]

Hi @dans-msft @Adarsh-Ramanathan , could you help confirm @Kevinlee49's comment? Thanks!

[tvanvalkenburg]

It seems to be a duplicate of #788

And if you use the Vector Index Lookup tool, it has been deprecated. Therefore, I recommend you to upgrade to the new index lookup tool.

It is well written in this link. https://learn.microsoft.com/en-us/azure/ai-studio/how-to/prompt-flow-tools/index-lookup-tool

I believe that I am using the index lookup tool. My code is below:

- name: lookup_question_from_indexed_docs
  type: python
  source:
    type: package
    tool: promptflow_vectordb.tool.common_index_lookup.search

This is the code that is auto-generated in the Studio when adding an IndexLookup:

[Kevinlee49]

My index lookup looks like this

[Adarsh-Ramanathan]

@tvanvalkenburg, can you share the full output of running pip list? I suspect you're missing one or more dependencies - maybe from forgetting to install promptflow-vectordb's azure exta? Running pf tool list and looking for errors in the output could also be useful.

[tvanvalkenburg]

@tvanvalkenburg, can you share the full output of running pip list? I suspect you're missing one or more dependencies - maybe from forgetting to install promptflow-vectordb's azure exta? Running pf tool list and looking for errors in the output could also be useful.

The output of pip list is attached.
piplist.txt

[Adarsh-Ramanathan]

@tvanvalkenburg - can you also run pf tool list and share the output?

[tvanvalkenburg]

I've attached the output
pf_tool_list.txt

[Adarsh-Ramanathan]

@tvanvalkenburg , can you run pip install promptflow-vectordb[azure] and see if that resolves your issue?

[tvanvalkenburg]

I tried this previously and got an error. I've tried it again just now to confirm:

[Adarsh-Ramanathan]

This is the likely source of the issue that you're seeing. What python version are you using?

[tvanvalkenburg]

In this environment I am runnning 3.11.6.

[Adarsh-Ramanathan]

That's strange - I ran pip install promptflow-vectordb[azure] in a python 3.11.6 container to verify, and it does work for me. Do you have a custom index source configuration in your environment, by any chance?

In any case, based on your pip list output, the only packages you're missing from the azure extra are azure-keyvault and azureml-rag. Can you try manually installing them?

azure.keyvault>=4.2.0
azureml.rag[azure,azure_cosmos_mongo_vcore,cognitive_search,faiss,milvus,pinecone]>=0.2.23

[tvanvalkenburg]

So I think we may be onto something.

Firstly, I have been using Python installed through pyenv. I think that this may be the cause of problem related to not being able to find promptflow-vectordb[azure].

I've tried installing the packages in a container build on the official Python 3.11.6 image, and pip is able to find the packages. However, when trying to run the install I get the following error:

[Adarsh-Ramanathan]

It's strange that pip is attempting to install azureml-rag 0.2.23 instead of 0.2.25, which is the latest version that's compliant with that version constraint.

I set up container with pyenv, ran pyenv install 3.11.6 followed by pyenv global 3.11.6, pip install promptflow promptflow-tools, and pip install promptflow-vectordb[azure]; everything ran without errors, including installing azureml-rag 0.2.25.

Is there anything else different about your environment that you think would be helpful in getting a repro going?

[diogoazevedo15]

Try to pip install like this:
pip install "promptflow-vectordb[azure]"

[tvanvalkenburg]

Unfortunately it doesn't work. I tried in a fresh Python 3.11.6 container

[tvanvalkenburg]

An update on this issue. I ran into a similar problem trying to install the az ml CLI. The problem with the CLI was that there is no arm64 wheel for azureml-dataprep-rslex. More details can be found at this issue: Azure/azure-cli#28600.

I got around the problem with the CLI by installing an older version, as is documented in the issue linked above. However, I am still not able to make it work when trying to install promptflow-vectordb.

[tvanvalkenburg]

I was able to solve the issue by creating a conda environment with x_86 architecture. Within that environment I am able to install the required packages without issue. It appears that the problem is that the Azure ML Python SDK has gaps on Apple's ARM chips.

[orasik]

I'm having the same issue.

I tried with multiple python versions [3.8, 3.9, 3.10, 3.11, and 3.12].

I have also tried in Apple M1 and Ubunute 22, same issue.

Tried with venv and conda, same issue.

pip install 'promptflow-vectordb[azure]'

Error

ERROR: Cannot install azureml-dataprep because these package versions have conflicting dependencies.

The conflict is caused by:
    azureml-dataprep 5.2.1 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.2.0 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.1.6 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.1.5 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.1.4 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.1.3 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.1.2 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.1.1 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.1.0 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 5.0.0 depends on azureml-dataprep-native<42.0.0 and >=41.0.0
    azureml-dataprep 4.14.1 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.14.0 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.10 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.9 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.8 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.7 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.6 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.5 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.4 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.2 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.1 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.12.0 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.11.7 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.11.6 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.11.5 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.11.4 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.11.3 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.11.2 depends on dotnetcore2<4.0.0 and >=3.0.0
    azureml-dataprep 4.11.1 depends on dotnetcore2<4.0.0 and >=3.0.0

To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict

ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts