# Using llama\_index Python modules in Omniverse

**URL:** <https://forums.developer.nvidia.com/t/using-llama-index-python-modules-in-omniverse/333803>\
**Category:** Omniverse Technical Support\
**Tags:** llama\
**Created:** [May 20, 2025, 3:09pm UTC](https://forums.developer.nvidia.com/t/using-llama-index-python-modules-in-omniverse/333803 "2025-05-20T15:09:25Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![dvelazco](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/dvelazco/32/397377_2.png) [@dvelazco](https://forums.developer.nvidia.com/u/dvelazco)\
**Post date:** [May 20, 2025, 3:09pm UTC](https://forums.developer.nvidia.com/t/using-llama-index-python-modules-in-omniverse/333803/1 "2025-05-20T15:09:25Z")

</div>

Hello,

I am trying to integrate some AI scripts into an Omniverse extension, but ran into an issue trying to import the llama\_index Python modules. I am able to install the llama\_index modules by listing them in the `[python.pipapi]` as such:

```auto
[python.pipapi]
requirements = [
    "llama-index",
    "llama-index-llms-ollama",
    "llama-index-embeddings-ollama",
    "llama-index-vector-stores-chroma",
    "chromadb",
    "sqlalchemy"
]
use_online_index = true

```

However, when I try to import llama\_index in Omniverse (either through an extension or the Script Editor). I get the following error:

```auto
ImportError: cannot import name 'async_sessionmaker' from 'sqlalchemy.ext.asyncio' (<redacted>/_build/windows-x86_64/release/extscache/omni.services.pip_archive-0.15.0+107.0.3.wx64.cp311/pip_prebundle/sqlalchemy/ext/asyncio/ __init__.py)

```

Any advice?

---

<div class="post-metadata">

**Author:** ![Richard3D](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/richard3d/32/207062_2.png) [@Richard3D](https://forums.developer.nvidia.com/u/Richard3D)\
**Post date:** [May 21, 2025, 3:22am UTC](https://forums.developer.nvidia.com/t/using-llama-index-python-modules-in-omniverse/333803/2 "2025-05-21T03:22:23Z")

</div>

Well it seems it has an issue with that specific name ‘async\_sessionmaker’ from ‘sqlalchemy.ext.asyncio’, so I would open that up and examine that more. Maybe it has a bad naming convention. I can try to ask the engineers but this is really outside the Omniverse scope.

What are you trying to do with Omniverse kit and AI?

---

<div class="post-metadata">

**Author:** ![dvelazco](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/dvelazco/32/397377_2.png) [@dvelazco](https://forums.developer.nvidia.com/u/dvelazco)\
**Post date:** [May 21, 2025, 3:35pm UTC](https://forums.developer.nvidia.com/t/using-llama-index-python-modules-in-omniverse/333803/3 "2025-05-21T15:35:52Z")

</div>

I am trying to use the `llama_index` Python modules to experiment implementing a custom RAG model. I am no expert in that, so it is a learning experience as well.

I was able to make it work by manually installing `SQLAlchemy` into the Kit’s Python environment. From what I observed, Omniverse Kit already includes `SQLAlchemy`, likely for its own internal use, but it might be a different version than what `llama_index` requires. After installing the latest version with `pip`, everything started working as expected.

It seems that the manually installed version takes precedence over the bundled one, at least in this context. So far, I haven’t noticed any side effects on other Omniverse features.

---

<div class="post-metadata">

**Author:** ![Richard3D](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/richard3d/32/207062_2.png) [@Richard3D](https://forums.developer.nvidia.com/u/Richard3D)\
**Post date:** [May 21, 2025, 4:37pm UTC](https://forums.developer.nvidia.com/t/using-llama-index-python-modules-in-omniverse/333803/4 "2025-05-21T16:37:42Z")

</div>

You know that we have our very own ChatUSD AI model in Omniverse already? Just install that. If you are on kit 106.5 you can look for ChatUSD. It will be coming for kit 107 soon.

---

<div class="post-metadata">

**Author:** ![dvelazco](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/dvelazco/32/397377_2.png) [@dvelazco](https://forums.developer.nvidia.com/u/dvelazco)\
**Post date:** [May 24, 2025, 1:05pm UTC](https://forums.developer.nvidia.com/t/using-llama-index-python-modules-in-omniverse/333803/5 "2025-05-24T13:05:52Z")

</div>

I’ll check it out, thanks!
