--- pretty_name: "Mattermost Data" tags: - webarena-pro - web-agents - benchmark-environment - text --- **Mattermost Data** provides conversation content and channel metadata for WebArena-Pro’s Mattermost environment, a locally hosted team-messaging application for evaluating web agents. It supports channel browsing, search, threaded conversations, and direct/group messaging. Imported message text comes from [Software-related Slack Chats with Disentangled Conversations](https://github.com/preethac/Software-related-Slack-Chats-with-Disentangled-Conversations). The direct source contains disentangled conversations from public software-development Slack communities. Original task messages and benchmark user/channel fixtures are synthetic. We prepared the released content for local use in the environment: - Mapped anonymized source participants deterministically onto local benchmark users. - Retained whole conversations as threads across 24 added engineering channels and a topic-matched backfill for original channels. - Exported content records from the exact SQL snapshot consumed by the application build. Authentication and session tables are not part of the content export. | Statistic | Value | |---|---:| | Imported corpus messages | 31615 | | Total post records including fixtures and system messages | 32389 | | Channel records including direct/default channels | 59 | | User profile records | 19 | | Local release files before documentation | 9 | | Local payload size | 27,189,124 bytes (27.19 MB / 25.93 MiB) | Content JSONL files are under `data/`, original JSON seed layers under `seed_inputs/`, and source-selection scripts under `preparation/`. Post counts include repeated message text and system events; they are not unique-sentence counts. The original source releases are categorized on Zenodo as Other (Open)* (`other-open`): [v1.0.0](https://zenodo.org/records/3627124) and [2.0.0](https://zenodo.org/records/3763432). The metadata does not name MIT, a Creative Commons license, or distinct code/data terms. This release preserves that source declaration without assigning a different license to the corpus or synthetic fixtures. The application is maintained in the [WebArena-Pro repository](https://github.com/McGill-NLP/webarena-pro). ## Code and data licenses | Component | Recorded license or status | |---|---| | Preparation code in `preparation/` | No explicit license declaration was found in the WebArena-Pro source repository for these scripts. No new software license is assigned by this release. | | Data and content | Other (Open)* (`other-open`) is the Zenodo-declared category for the original archived code/data releases. Detailed license conditions and distinct code/data licenses are not identified in that metadata. Original synthetic benchmark fixtures have no separately declared data license. | \* Per original authors. This reproduces the source release or dataset uploader’s declaration; it does not substitute MIT for a Creative Commons license or invent missing reuse conditions. Zenodo distinguishes open access (`access_right: open`) from its license category (`license.id: other-open`). The same `other-open` category appears on both original release records.