[caiweihan]

What are you adding, and what does it do?

Adds Jev Account Filter to Data and Evaluation (data/data.yaml) with English/Chinese descriptions and a screenshot hotlinked from the project's own pinned revision. No generated page is edited or committed.

Affiliation: self-submission by caiweihan, the maintainer.

• Edited only data/data.yaml; ran python scripts/build.py --validate and the local build.
• Links point to the original project, not another list.
• English description starts with a capital letter and ends with a period.
• Image is hotlinked from the project's own published screenshot.

Project and Jev workflow

Jev Account Filter is a Python CLI for screening video-platform accounts for dataset collection and research without downloading or watching complete videos. Our team uses it for sign-language account selection; category rules and the model/API are explicit configuration.

• Pass 1 sends only first-page video titles to Jev Noul questions; application code sorts the probabilities and selects Top-N (five by default).
• Pass 2 sends account name, biography, and selected titles/JPEG thumbnails together. Jev returns an independent account-category Choice, separate account/video Noul probabilities, and a best-candidate Choice.
• CSV output, raw typed responses, request previews and an offline HTML viewer make the result reviewable. Collection and inference are separate, and successful items are skipped on resume.

Evidence and limitations

• Reviewed public revision: c46e437, on 2026-10-17.
• Implementation: request building and typed-answer validation, HTTP client, Jev 1.14 configuration.
• Earlier in this AI-assisted development session, three public example accounts completed both stages: six successful live OpenRouter requests, five thumbnails per second-stage request. This is an integration check, not an accuracy benchmark.
• The published team experiment reports 2,140 Douyin accounts, eight hours of data collection plus one hour of Jev inference, and $0.565 total API cost. Reported 94.49% accuracy uses 198 existing second-review labels and assumes unreviewed original labels are correct; it is a conditional estimate, not independent full-dataset validation. These figures are in the public README.
• Python 3.11+, a browser for collection, and a user-supplied OpenRouter key are required for the documented Jev path. Screening sends selected account metadata, titles and images to OpenRouter/its model provider and incurs API charges. Already collected data can be screened without revisiting the platform.
• Source is public, but no project-wide source-code license is declared at the reviewed revision; the third-party icon license is not a license for the application. This submission does not change or imply licensing terms.

Affiliation and AI assistance

Self-submission by the project maintainer, caiweihan, on behalf of the team using the tool. The implementation, documentation, submission preparation and checks were substantially assisted by OpenAI Codex. No independent human review or third-party evaluation of this submission is claimed. No paid placement, referral links, credentials, or private account datasets are included.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.