We are building a CitizenComplaintAgent using UiPath Coded Agent + Python + LlamaIndex.
Use case
The agent accepts a citizen complaint, analyzes the input using LlamaIndex, and executes the corresponding automation workflow. The complete solution works correctly when running from VS Code.
Problem
The same project does not work in UiPath Studio Web. It runs successfully in VS Code, but Studio Web fails to execute the coded agent.
The main challenge is that Studio Web does not show any specific error message or trace logs. The job simply ends with the status:
Job Status: Faulted
Because there are no detailed logs or exception information, it is difficult to identify the root cause.
Environment
UiPath Coded Agent
Python
LlamaIndex
VS Code (working)
UiPath Studio Web (not working)
Question
Is LlamaIndex fully supported in UiPath Studio Web?
Are there any known limitations or dependency restrictions for Python-based coded agents?
How can we enable detailed error logs or trace logs when a Studio Web job is faulted?
Has anyone implemented a similar setup successfully?
from the screenshot it looks like the job is failing during execution, but Studio Web isn’t exposing the actual exception.
I would first verify that LlamaIndex and all its Python dependencies are supported in the UiPath Coded Agent environment. A package working in VS Code doesn’t necessarily mean it will work in the Studio Web runtime.
As a quick test, try running the agent with a simple Python function first, then add LlamaIndex/dependencies one by one. Also add try/except logging around the Python code to capture the actual exception.
This should help narrow down whether the issue is the runtime or a specific dependency.
First, download and review the job logs (the screenshot shows a Download logs option). Also check Orchestrator → Jobs → Logs for the detailed exception.
Verify that all required Python packages (LlamaIndex, OpenAI SDK, vector store libraries, etc.) are included in the project deployment and accessible in the Studio Web execution environment.
Studio Web runs in a cloud-managed runtime, which may not support all local Python configurations that work in VS Code.
Check for:
Missing environment variables/API keys
Unsupported Python libraries
Local file path dependencies
Network/firewall restrictions for external LLM services
Add extensive try/catch logging in the coded agent and write exception details to logs to identify the exact failure point.
Test with a simple Python script first. If that succeeds, gradually add LlamaIndex components to find the dependency causing the fault.
@Sushma_Ganapatisa This looks more like a Studio Web runtime/dependency issue than a LlamaIndex support issue. because the same code is working fine from VS Code. Please check;
-Make sure all Python packages used in your local setup are added in pyproject.toml. Sometimes package is already installed locally, so VS Code works but Studio Web fails.
- Check the Python version also.
-If you are using any .env file for API keys/configuration, make sure the same values are available in UiPath environment. Studio Web will not pick your local
-Try running a very simple coded agent first without LlamaIndex. If that works, then add LlamaIndex/dependencies one by one. This will help to identify where exactly it is failing.
Also, from your screenshot I can see Download logs option. Please download that and check once. Even if Studio Web is only showing Job Status: Faulted, normally the downloaded logs may have some more details. One more thing, it looks like the job is failing just after provisioning and before the actual execution starts. So I would first check dependency/configuration side rather than the agent logic. If you can share the pyproject.toml and downloaded error log (remove any keys/secrets), we can check further.
Looking at the screenshot, the job is reaching the execution stage but ends as Faulted without exposing the actual exception.
Since the same code works in VS Code, I would first check the Python dependencies and environment used by the Coded Agent. LlamaIndex may depend on additional Python packages or runtime features that are available locally but not available in the Studio Web execution environment.
I would try running a very simple version of the agent in Studio Web first, without LlamaIndex, and then add the dependencies back one by one. This can help identify which package or part of the code is causing the failure.
Also, adding exception handling around the main Python code and logging the exception details could help, for example capturing str(e) and the full traceback.
If you can share the Python imports/dependencies used by the agent and the relevant part of the code where LlamaIndex is initialized, it would be easier to pinpoint the issue.
