feat: Add GUI for DPK Planning Agent#1369
Conversation
Signed-off-by: dobromir <[email protected]>
dobromiriiliev
force-pushed
the
DPK_Agent_GUI
branch
from
b3db4cc to
d26ebd3
Compare
|
@dobromiriiliev could you please also add |
Thanks, @santoshborse, I also asked @dobromiriiliev in our call yesterday to add the README. |
| @@ -0,0 +1,113 @@ | |||
| # DPK Planning Agent GUI | |||
|
|
|||
| This is a Gradio-based web interface to generate, edit, and run data pipelines using the Data Prep Kit (DPK) transform launcher via LangGraph `StateGraph`. It falls back to a demo mode when the DPK library is unavailable. | |||
There was a problem hiding this comment.
Please add a hyperlink to Gradio when it is mentioned.
| ## Prerequisites | ||
|
|
||
| * **Operating System:** macOS, Linux, or Windows | ||
| * **Python:** 3.8 or newer |
There was a problem hiding this comment.
DPK requires Python 3.10 or newer, so if DPK is to be used, we need to say 3.10 or newer.
|
|
||
| * **Pipeline Generation:** Generate YAML pipelines from natural language prompts. | ||
| * **Pipeline Editing:** Apply edit instructions to existing pipeline YAML. | ||
| * **Pipeline Execution:** Run pipelines via the DPK launcher and displays console output. |
There was a problem hiding this comment.
displays => display
| **Note:** You can install the Data Prep Kit via PyPI (when available), by running: | ||
|
|
||
| ```bash | ||
| pip install data-prep-kit |
There was a problem hiding this comment.
pip install data-prep-toolkit-transforms[language]
dobromiriiliev
force-pushed
the
DPK_Agent_GUI
branch
2 times, most recently
from
128a74a to
eed0f94
Compare
…K install options
touma-I
left a comment
touma-I
left a comment
There was a problem hiding this comment.
Please see various comments. I think more work is needed before this can be merged and used by the broader community.
| print("--- DPK Discovery (Subprocess Execution) ---") | ||
| script_dir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe()))) | ||
|
|
||
| launcher_path = os.path.abspath( |
There was a problem hiding this comment.
Can this be re-written to allow users to pip install the transforms rather than by cloning ?
| error: Optional[str] | ||
|
|
||
|
|
||
| def _run_dpk_as_subprocess(args: List[str]) -> subprocess.CompletedProcess: |
There was a problem hiding this comment.
Not clear on why this runs as a subprocess ? Can you elaborate on this design decision ?
There was a problem hiding this comment.
Some of the design decisions in the code are not obvious. I suggest you enhance the readme.md to include details on the architecture and design of this application.
|
|
||
| # 2. Clone the Data Prep Kit (DPK) as a sibling directory | ||
| cd .. | ||
| git clone https://github.com/data-prep-kit/data-prep-kit.git |
There was a problem hiding this comment.
Why can't we just do pip install ? Why do we need to do git clone ?
refactor: Align DPK integration with library API to resolve errors
Signed-off-by: dobromir <[email protected]>
|
Thank you, @dobromiriiliev. Two small problems that I see:
|
4 participants
Why are these changes needed?
The GUI wrapper streamlines the creation of data pipelines with the DPK Planning Agent by allowing users to use NLP (Natural Language Prompts) instead of creating Jypter notebooks or manually writting YAML configurations. The graphical user interface allows for the iterative design process by allowing users to make recommendations or changes to the YAML configurations; the GUI also allows the final configuration to be run as a subprocess.
Related issue number (if any).
[Feature] GUI Wrapper for the DPK Planning Agent/Front-End Orchestration #1318