[Bug] Core setup failures on fresh install: API rename, missing class, wrong docs path, undocumented deps

[31groot]

Environment

• OS: Ubuntu (Linux)
• Python: 3.11
• Ianvs: latest main branch
• Sedna: installed from kubeedge/sedna github main branch

Summary

During a fresh setup using the official quick start guide, I encountered multiple blocking issues that are not currently documented. As it stands, a new contributor is unlikely to run any example successfully without running into these problems.

Bug 1: Incorrect docs path for PCB-AoI example

The path mentioned in the documentation is incorrect.

Documented path:
examples/pcb-aoi/singletask_learning_bench/testalgorithms/fpn/benchmarkingjob.yaml

Actual path in the repository:
examples/pcb-aoi/singletask_learning_bench/fault_detection/benchmarkingjob.yaml

The fpn/ directory does not exist at the documented location, so the docs should be updated to point to the correct path

Bug 2: JsonlDataParse was renamed in Sedna, but Ianvs still uses the old name

core/testenvmanager/dataset/dataset.py imports JsonlDataParse from
sedna.datasources, but in Sedna this class has been renamed to JSONDataParse.

Current error:
ImportError: cannot import name 'JsonlDataParse' from 'sedna.datasources'

Required fix:
Replace JsonlDataParse with JSONDataParse in dataset.py lines 25 and 587.

Bug 3: JSONMetaDataParse is imported in Ianvs, but does not exist in Sedna

dataset.py imports JSONMetaDataParse from sedna.datasources but
this class does not exist anywhere in Sedna.

Running dir(sedna.datasources) shows that JSONDataParse is available, but JSONMetaDataParse is not.

This breaks the JSONFORLLM dataset format at import time, which means all LLM-based examples depending on it currently fail before execution starts.

Error:
ImportError: cannot import name JSONMetaDataParse from sedna.datasources

Bug 4: PCB-AoI example depends on undocumented packages

The PCB-AoI example requires additional packages that are not documented anywhere in the repository.

At minimum, the example depends on:

• opencv-python
• tensorflow

These dependencies are not listed in requirements.txt and are not mentioned in the example README or setup instructions.

As a result, a fresh setup fails with ModuleNotFoundError,, and new contributors are given no indication of which packages need to be installed.

Reproduction steps

1. Clone kubeedge/ianvs
2. Run pip install -e .
3. Install sedna from GitHub source
4. python benchmarking.py -f examples/pcb-aoi/singletask_learning_bench/fault_detection/benchmarkingjob.yaml
5. Hit the errors above in sequence

Suggested fixes

• Update docs path from fpn/ to fault_detection/
• Replace JsonlDataParse with JSONDataParse in dataset.py
• Implement JSONMetaDataParse in Sedna or create a fallback in Ianvs
• Add requirements.txt to each example folder listing its dependencies