MiroThinker is a deep research agent framework designed for complex, multi-step tasks requiring external tool use. The system orchestrates LLM reasoning with a modular suite of MCP (Model Context Protocol) tools including web search, code execution, vision, audio, and reasoning capabilities. It features sophisticated context management through failure experience summarization and sliding window truncation to handle long conversations. The agent iteratively generates tool calls, executes them via MCP servers, and integrates results until producing a final answer. The repository includes benchmark evaluation scripts for GAIA, BrowseComp, HLE, and other research benchmarks, trace collection utilities for model improvement, and a Gradio demo for interactive use. Architecture centers on an orchestrator managing the agent loop, a pipeline for step execution, tool executors handling MCP server interactions, and LLM abstractions supporting Anthropic and OpenAI APIs with prompt caching.
E-RayZer is a self-supervised 3D reconstruction system that takes unordered multi-view images as input and outputs estimated camera poses, target-view renders, Gaussian point clouds, and turntable videos. The system uses a transformer-based architecture to estimate camera poses and scene representations, then renders novel views using differentiable Gaussian splatting. It includes a Gradio demo interface for easy inference, along with training and evaluation scripts for the DL3DV dataset. The model is trained without ground truth camera poses or depth information, learning to reconstruct 3D scenes purely from multi-view image consistency.
SAIL-Recon is a feed-forward Transformer-based system for large-scale Structure-from-Motion (SfM) and visual localization. It processes input images or video to estimate camera poses, depth maps, and generate 3D point clouds without iterative optimization. The architecture combines a Vision Transformer backbone with an alternating attention aggregator for temporal modeling and specialized heads for camera, depth, and track predictions. The system includes both command-line and Gradio-based demo interfaces for end-to-end 3D reconstruction, along with evaluation scripts for standard benchmarks (TUM, 7Scenes, Tank&Temple, etc.). While research-oriented, it is packaged as a usable application with clear entry points and modular components.
The OmniParser repository provides a system for parsing GUI screenshots into structured elements (text and icons) to enable vision-based GUI agents. It combines object detection (YOLO), OCR (PaddleOCR), and vision-language models (Florence2) to detect and describe screen elements, outputting bounding boxes and structured data. The system includes a FastAPI server for parsing as a service, Gradio and Streamlit demos for interactive use, and agent frameworks that integrate with LLMs (like GPT-4o, Claude) to perform automated computer tasks within a Windows VM environment (via the OmniBox component). The architecture is modular, with core parsing utilities in the `util` directory, agent tools and loops in `omnitool/gradio`, and VM management in `omnitool/omnibox`. Key reusable components include the model loading and processing utilities (`util/utils.py`), a tool abstraction pattern for defining agent actions (`omnitool/gradio/tools/`), and a visualization utility for annotating screenshots (`util/box_annotator.py`). The system demonstrates non-obvious skill in integrating multiple ML models for comprehensive screen understanding and providing a standardized interface for agent-LLM-tool interaction. While actively maintained (last commit July 2026), the repository shows signs of technical drift, including dependency version pinning that may become outdated and deprecation warnings in the demo notebook (e.g., timm layers, Florence2 GenerationMixin). The reliance on external services (Hugging Face models, LLM APIs) and the complexity of the Windows VM setup also introduce maintenance considerations. Nonetheless, the codebase delivers on its promise of enabling vision-based GUI agents and offers several patterns reusable in similar projects.