BriefLoop

BRIEFLOOP / GUIDES

Search and task progress

0.22.2: choose search channels, fix the reporting period and follow actual progress.

Current installers: Mac 0.22.2, Windows 0.22.2. Labels and capabilities vary by channel; the installer version is authoritative. Development features are not implied.

Before you start

Use a workspace you can back up. Model tasks require an authenticated runtime and permission to process the chosen materials.

Steps

  1. In Settings → Run and search, choose a preferred channel and allowed supplements. New workspaces prefer Tavily and allow runtime-native search to supplement it. Enable Bocha or Zhipu explicitly and configure their keys.
  2. Choose cross-market coverage checks, gap-only supplementation or preferred-channel only. Tavily, Bocha and Zhipu use separate accounts. Without a search key, choose DuckDuckGo or capabilities already available in your runtime.
  3. Check the cutoff, date range and time zone before starting. The run retains its search policy and time window; later settings apply to new tasks. Scheduled reports recalculate relative dates for each run.
  4. In My reports, follow the current stage, elapsed time, successful and failed searches, saved sources, research rounds, review and revisions. Open saved drafts early; stop or resume the original task when needed.
  5. Open Research budget → Search records for channels, purpose, duration and errors. Filter sources by their discovery channel.
  6. Version 0.22.2 adds compact research-depth, fact-checking and WikiSkill options on the home and new-report screens. The default execution limit is 60 minutes. Deep research defaults to a 10,000-character target and 12,000-character cap; explicit length settings take precedence.
  7. Set simultaneous reports in settings: the default is 4, with a range of 1–16. Scout concurrency is separate. A dedicated waiting page shows pending generation, while the version selector lets you open saved drafts. Queue time and execution time are shown separately.

What success looks like

Stages and counts come from recorded execution and artifacts, without an estimated completion percentage. Failed searches remain visible and consume search-attempt quota.

If something goes wrong

Unmetered runtime-native search is shown as unknown rather than included in controlled API counts. Search snippets, saved sources and successful queries do not establish factual accuracy. There are no dedicated WeChat or Xiaohongshu connectors.

← All guides · Installation options →