SHIPPED · v0.15.0

Automated Pipelines and Web Search

New

  • A Social repurpose pipeline that runs without opening New Campaign every time — Pipelines is a dedicated view for watching a source channel, adapting each new post for the destinations you choose, and either saving drafts or scheduling them to publish.

  • Watch a public URL when the channel API can't list posts — Pipelines can poll a public profile or post URL instead of a connected account — useful for personal Facebook profiles, LinkedIn without read approval, and any channel where a public page is the only source you have.

  • Watch local folders and turn every new file into content — The Directories tab is now a folder pipeline: point it at folders on your machine, and each new file that lands there becomes a campaign on the next check — no paste, no upload step.

  • Per-channel overrides inside a folder pipeline — A selected destination shows a summary strip — `EN · Plain text · Balanced` — that opens the same inline settings panel used in New Campaign, so one channel can run in two languages, Markdown, and a more detailed voice while the rest keep the pipeline defaults.

  • Optional web search while campaign content is generated — A Web switch in the campaign composer lets the content AI look up live sources when writing, instead of only the brief and model knowledge.

  • Generation logs that show the search and the sites — When web search runs, Generation Logs open and list each query, then the results as concrete sites — hostname, title when available, and an open link — so you can see what the AI actually looked at.

  • Grok as a local CLI agent — Grok Build is detected alongside Claude, Codex, Gemini, and the other local agents, selectable for campaign generation, redesign, and the rest of the local AI routes.

Improved

  • Jump from a pipeline run to what it produced — A run row in the shared Logs tab now carries Campaign and Distribution actions — the first opens the campaign results for that run's output, the second opens Distribution History already filtered to that pipeline. Runs that produced nothing show no actions.

  • Distribution History tells automated pieces apart from hand-run ones — Pieces written by a pipeline carry a Pipeline badge naming the pipeline and the source it came from, and a new Pipeline filter narrows the whole table — counts included — to a single pipeline's output.

  • One Logs tab for both pipeline kinds — Social and folder runs share the same history, with wording that matches what actually ran ("Scanning the watched folders…" versus "Checking the source channel…").

  • More source channels for pipeline watching — Connected-account sources cover Facebook Pages, X, Instagram, Threads, YouTube, TikTok, Bluesky, and Mastodon where the channel is set up; LinkedIn can still use URL watch when Graph read access isn't available.

  • Pipeline config that matches how campaigns already work — Destination channels, language, content detail, and output format reuse the same options you already set in New Campaign, so a pipeline and a one-off campaign stay consistent.

  • Campaign drafts and presets remember the Web switch — Turning web search on for a draft survives reloads, and Save preset can store it as the default for new campaigns.

  • Open a watched folder from the app — The path under each folder in the pipeline config is a button that reveals it in Finder or Explorer.

Fixed

  • The installed app now uses the right typefaces — Fonts ship with the app instead of being fetched on launch, so the packaged build no longer silently falls back to a system font — and a local-first app makes one less network call at startup.

  • AI CLIs on Windows that were detected but refused to run — Agents installed as a command shim that forwards to a bundled executable — Claude Code among them — reported "Cannot run Windows command shim safely" instead of launching. They now start correctly.

  • Windows agents no longer fail right after a reboot — A cold start could push several CLIs past the version-check timeout at once, showing installed, working agents as errors until you pressed Rescan. Windows now gets a longer window.

  • Pipelines view with an active Social from khoa.lowkey pipeline expanded, showing its Trigger, Transform, Publish to and Output nodes on the pipeline graph
    Pipelines view with an active Social from khoa.lowkey pipeline expanded, showing its Trigger, Transform, Publish to and Output nodes on the pipeline graph
  • Social repurpose Source tab with Facebook selected and a public profile URL pasted into Watch URL, every channel card marked URL watch
    Social repurpose Source tab with Facebook selected and a public profile URL pasted into Watch URL, every channel card marked URL watch
  • Directories tab showing a live Folder watch pipeline for demo-folder, its six watched file types, and the file-to-campaign pipeline graph
    Directories tab showing a live Folder watch pipeline for demo-folder, its six watched file types, and the file-to-campaign pipeline graph
  • Folder watch Configure pipeline modal on the Folders tab with one watched folder, Ignore files that are already there checked, and an hourly check interval
    Folder watch Configure pipeline modal on the Folders tab with one watched folder, Ignore files that are already there checked, and an hourly check interval
  • Folder watch config offering the two grouping modes side by side: Every file is a post versus Every subfolder is a post
    Folder watch config offering the two grouping modes side by side: Every file is a post versus Every subfolder is a post
  • New campaign composer with the Web toggle switched on beside the language and Grok agent pickers
    New campaign composer with the Web toggle switched on beside the language and Grok agent pickers
  • Generation Logs panel during a run with a 6 search badge and log lines confirming web search is on and completed
    Generation Logs panel during a run with a 6 search badge and log lines confirming web search is on and completed
  • Estimated token usage modal breaking down eight outputs by channel with a note that web search adds latency and provider fees
    Estimated token usage modal breaking down eight outputs by channel with a note that web search adds latency and provider fees