Skip to content

v0.1.11

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 07 Oct 02:03
v0.1.11
4bd8c28
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

TextFlowKit v0.1.11

New: opt-in native live transcription

  • Shared StreamingSession Python API with incremental events, bounded PCM input, normal finish/flush, cancellation, and owned native-worker cleanup.
  • Loopback-only WebSockets: /stream on the developer HTTP server and /api/stream on the local browser UI.
  • Packaged browser microphone example with AudioWorklet resampling and explicit Start / Stop and finish / Cancel lifecycle.
  • Enable with the streaming extra and TEXTFLOWKIT_STREAMING=1. Live native support is Windows x86_64 only; existing standalone cross-platform support is unchanged.
  • Direct pinned native library binding avoids the upstream Python SDK and its telemetry module.

Existing file/URL transcription, CLI, MCP, durable jobs, and default standalone Whistle engine remain intact. Fonts remain0.1.6 and their original published distributions are reused.

README, manuals, installation and adapter guides, architecture documentation, changelog, and the Cloudflare landing page now describe0.1.11. A new streaming architecture drawing distinguishes ephemeral live sessions from durable file/URL jobs.

Evidence and limits

Final local Windows suite: 1960 passed, 18 skipped, 2 warnings in 187.52s (0:03:07); skips are not passes. Ruff, actionlint, release guards, build and Twine checks passed. Real native English streaming ran for65s and300s, plus an assembled UI WebSocket run for12s and a five-second installed-wheel core check. Those feeds used paced prerecorded PCM, not a physical browser microphone. Physical microphone capture, non-English native runs, and native live Linux/macOS are not claimed. Live sessions are ephemeral and cannot resume after restart.

A clean native Windows YouTube run on the merged release commit4bd8c281a6911182c0fde76f57e0e65e72e3c56d completed2026-10-07T01:55:09Z (2026-10-06 locally): Whisper tiny/CPU, three timestamped segments, expected whole word elephants recognized. This covers that one URL, not every recognized platform or the default Whistle engine; it is not GitHub-hosted YouTube evidence.

The public website is documentation only, not a public transcription or streaming service. Current live-platform and microphone verification limits are stated in the manuals.

Install

pip install "textflowkit[export,mcp,http,streaming]==0.1.11"

See the streaming guide and user manual.

Assets 7
Loading