Member: Nguyen Phuc Toan
Done:
- Fit response brainstorm agent with user language
- Fix error token monitor
Todo:
- Fix Error Crash Fonted when large scene (Backend Scale).
Member: Nguyen Thanh Tuyen
Done:
- Researched and proposed load testing strategy for AES using k6.
- Identified key bottlenecks, monitoring metrics, and scaling configurations.
- Evaluated server providers for stress test, staging, and production environments.
Todo:
- Set up staging VPS and monitoring stack (InfluxDB + Grafana).
- Prepare k6 load/stress test scenarios and dashboard structure.
- Implement MockProvider integration for LLM stress testing.
Member: Truong Thanh Thang
Done:
- Reviewed 4 pull requests.
- Improved the UI based on the Figma design.
- Added environment list, import, and load features.
- Added generation model selection and updated model parameters across generate endpoints.
Todo:
- Review pull requests.
- Update feedback handling when users hit the rate limit.
- Add an option to use the user's prompt directly.
- Fix the bug where clicking an object tag does not highlight the object, unlike right-clicking the object in the scene.
- Continue improving the UI.
Member: Le Minh Duc
Done:
-
Added usage quota feature, limiting how many LLM apis call a user can make a day
-
Converted existing RESTful asset generate endpoints to ws, added JWT as requirements in request body of 2 main generate endpoints, Added kill process feature endpoint
-
Added venice as provider in monitor feature, added more tracking to monitor feature: agent, model name
-
Fixed the bug where there are mismatchs between how FE and BE store scene causing objects Z-coordinate to unexpectedly change when choosing robot
Todo:
-
Find a way to ultilize 16k asset dataset without overloading front-end (pagination, keyword search, semantic search)
-
Add time estimating feature to time-taking feature like generate
Member: Nguyen Thanh Tuyen
Done:
- Completed planning and research for AES load/stress testing strategy.
Todo:
- Develop asset import feature.
- Develop robot import feature.
Member: Le Minh Duc
Done:
- Planned and discussed with co-workers, tech lead about the estimation feature
- Upserted 16k assets dataset into a new branch, tested the new cache hit with that bigger dataset, tested the compatibility of the dataset with the app
- Fixed the bug where all objects whose the same parent id got collapsed into each other
- Tested the webapp end-to-end to address any leftover bugs
Todo:
- Implement time-estimation feature for asset generation and layout generation
- Further test the compatibility of the 16k dataset, propose a more robust data transformation pipeline for the 16k feature (the orientation of some objects in it is not upright)
Member: Nguyen Thanh Tuyen
Done:
- Completed develop import asset and import robot
- Show list private robot and asset per user
- Check merge conflict from develop and push new branch import-robot-asset
Member: Le Minh Duc
Done:
- Further tested the 16k dataset e2e with layout generation
- Proposed and implemented an improvement to the orchestrator agent
Todo:
- Implement time-estimation feature for asset generation and layout generation
- Further test the compatibility of the 16k dataset, propose a more robust data transformation pipeline for the 16k feature (the orientation of some objects in it is not upright)
- Improve orchestrator agent
Member: Nguyen Phuc Toan
Done:
- Backend Decode Prompt User
- Outputs newly generated assets when a scene is created.