browser-agent/NOTICE at main · MadeAgents/browser-agent · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
NOTICE

This project incorporates code from the following open source project:

================================================================================

BrowserGym
https://github.com/ServiceNow/BrowserGym

Copyright 2024 ServiceNow

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

Description:
BrowserGym is a Gym environment for web task automation in the Chromium browser.
This project uses BrowserGym's core functionalities, experiments utilities, and
WebArena benchmark components.


Version: 0.14.1
Commit ID: 79ac68e843fb173c26f7cc21900e4bcf7c77c16b


================================================================================


Additional modifications made by OPPO Research Institute - Topo Lab in 2025:

Summary of Changes:
- Extended core action space with new high-level functions for robust web interaction.
- Enhanced environment and observation utilities to support structured HTML processing and accessibility tree enrichment.
- Introduced a custom agent post-processor (AgentOCCAM) for action refinement and safety filtering.
- Improved experiment loop to support advanced logging, checkpointing, and evaluation metrics.
- Updated WebArena and VisualWebArena task implementations to enable flexible goal specification and multi-step reasoning.
- Added new utility modules for HTML parsing, element selection, and DOM traversal.

Modified Files:
1. browsergym/core/src/browsergym/core/action/functions.py
   - Added take_note() and calculate() functions to support multi-step reasoning
   - Enhanced action primitives for agent capabilities

2. browsergym/core/src/browsergym/core/action/highlevel.py
   - Updated action imports to include new functions (take_note, calculate)
   - Modified action set configurations

3. browsergym/core/src/browsergym/core/env.py
   - Integrated agentoccam_processor module for enhanced observation extraction
   - Added extract_tree_from_page functionality
   - Modified observation space and extraction logic

4. browsergym/core/src/browsergym/utils/obs.py
   - Added overlay_som() function for Set-of-Marks visualization
   - Enhanced observation processing and rendering capabilities

5. browsergym/experiments/src/browsergym/experiments/loop.py
   - Added extract_action_content() function for action tag parsing
   - Enhanced experiment loop logic and action processing

6. browsergym/visualwebarena/src/browsergym/visualwebarena/task.py
   - Modified task initialization and configuration logic
   - Updated goal building and image processing functionality

7. browsergym/webarena/src/browsergym/webarena/task.py
   - Adjusted task setup and configuration loading
   - Modified task prompt and evaluation logic

New Files Added:
1. browsergym/core/src/browsergym/core/agentoccam_processor.py
   - Implements accessibility tree extraction and processing
   - Provides TreeNode class and tree manipulation utilities
   - Supports advanced observation extraction from browser pages

2. browsergym/core/src/browsergym/core/utils.py
   - Defines core data structures and type definitions
   - Implements utility functions for data conversion (png_bytes_to_numpy)
   - Provides TypedDict definitions for accessibility tree and DOM nodes

3. browsergym/core/src/browsergym/core/html_tools/ (New Module)
   - __init__.py: Module initialization and exports
   - html_parser.py: HtmlParser class for DOM tree parsing and manipulation
   - identifier.py: IdentifierTool class for generating unique element labels
   - prompt.py: HtmlPrompt class for formatted HTML representation generation
   - utils.py: Utility functions for XPath generation and element mapping
   - configs/: Configuration files for HTML parsing strategies

All modifications are clearly marked at the beginning of each file with modification
notices including date and description of changes.