Add basic browser tool for agentic env gen interactive editing#803
Draft
qianl-nv wants to merge 17 commits into
Draft
Add basic browser tool for agentic env gen interactive editing#803qianl-nv wants to merge 17 commits into
qianl-nv wants to merge 17 commits into
Conversation
Adds a Streamlit-based live editor for reviewing and editing UnresolvedArenaEnvGraphSpec YAMLs. The review tool renders: - A mermaid.js spatial graph of initial-state constraints with anchor highlighting, task-constraint dashed edges, and objectReference dotted edges - A task table (index, kind, description, params) - A node card grid with type badges and per-node YAML stanzas The YAML editor auto-renders when the spec is valid and changed. Node cards display a two-letter placeholder thumbnail by default. Also adds YAML round-trip serialization to ArenaEnvGraphSpec and moves assert_unique_ids/assert_references_exist to __post_init__ so validation fires on every load-from-yaml/dict path. Signed-off-by: Qian Lin <qianl@nvidia.com>
Replace monolithic review_graph.py and review_app.py with server, streamlit_ui, and a render subpackage; node thumbnails remain placeholders until SimApp lands. Signed-off-by: Qian Lin <qianl@nvidia.com>
Replace html_document.render_html_for_spec with dashboard.render_dashboard_html. Signed-off-by: Qian Lin <qianl@nvidia.com>
Stack visualizer sections vertically with three node thumbnails per row and unary constraints beside the Mermaid spatial graph instead of below it. Signed-off-by: Qian Lin <qianl@nvidia.com>
… import. Wire LLM fetch plus in-process catalogue build and IntentCompiler into the live editor with optional --yaml, and lazy-load USD helpers so library registration avoids early pxr imports. Signed-off-by: Qian Lin <qianl@nvidia.com>
qianl-nv
commented
Jun 16, 2026
| /isaac-sim/python.sh -m isaaclab_arena_examples.agentic_environment_generation.review_gui.server \\ | ||
| --yaml isaaclab_arena/tests/test_data/pick_and_place_maple_table_init_env_graph.yaml | ||
|
|
||
| # Prompt-only (empty editor until you generate or paste YAML): |
Collaborator
Author
There was a problem hiding this comment.
move this to the first usage line. this should be the default mode
write_yaml already validates and serializes; to_yaml was unused in the repo. Signed-off-by: Qian Lin <qianl@nvidia.com>
build_asset_catalogue() after SimApp init: ~10.0s without deferral vs ~9.9s with deferral; pxr loads in both cases once ensure_assets_registered imports the object libraries. Signed-off-by: Qian Lin <qianl@nvidia.com>
The review GUI still validated EnvironmentIntentSpec after fetch_intent_from_prompt, so the split added indirection without skipping work; generate_spec inlines the LLM call and validation again. Signed-off-by: Qian Lin <qianl@nvidia.com>
Relocate the Streamlit CLI from review_gui/server.py to the parent package so the launcher sits beside review_gui/ rather than inside it. Signed-off-by: Qian Lin <qianl@nvidia.com>
Document render_tasks_table, render_node_cards, and render_node_card in panels.py to match the existing render_unary_constraints style. Signed-off-by: Qian Lin <qianl@nvidia.com>
Signed-off-by: Qian Lin <qianl@nvidia.com>
Remove the duplicate usage line and list the no-args launch first in gui_runner.py and streamlit_ui.py. Signed-off-by: Qian Lin <qianl@nvidia.com>
Document the remaining helpers and panel renderers so every public entry point in the Streamlit app has a summary docstring. Signed-off-by: Qian Lin <qianl@nvidia.com>
Cache validation per editor text, drop the duplicate validate on file load, and pre-render generated specs from the compiler output instead of re-parsing on the next Streamlit rerun. Signed-off-by: Qian Lin <qianl@nvidia.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Add a browser-based agentic env gen/review script prompt-driven spec generation, initial spec editing and graph visualization.
Detailed description
User needs a tool to interactive run pieces of the agentic env gen pipeline, make changes and visualizer the impact.
A new script is added to start the browser-based tool. A new module is also added for rendering/server logic used by the toll.
The review GUI supports starting from input prompt, or load an initial graph spec (using --yaml).
An additional user interface to run the agentic env gen pipeline with better support for user intervention. It supplements the e2e script.
Support for persisent SimApp will be in a follow up MR.