Skip to content

About

Multi-modal python API

Resources

Stars

2 stars

Watchers

2 watching

Forks

Repository files navigation

whissle — deprecated, use whissle-sdk

This package is no longer a Whissle client. It is a pointer to the one that is.

pip uninstall whissle
pip install whissle-sdk
from whissle_sdk import WhissleClient

w = WhissleClient()                          # reads WHISSLE_API_KEY
w.asr.transcribe("audio.wav", diarize=True)  # speech-to-text, with the metadata heads

Why

whissle wrapped an HTTP API at api.whissle.ai and authenticated with a wh_ token. Neither of those exists for you any more:

  • The host is gone. api.whissle.ai returns 404 on every path this package called — /v1, /asr/transcribe, /backend/api-tokens/validate, /redoc, and the host root. The token page the old README sent you to (browser.whissle.ai/access) does not resolve at all.
  • The token type is not self-serve. wh_ is a legacy gateway device token. It is not something a customer can create. Keys you can create are wsk_ (workspace secret, server-side) and wpk_ (publishable, safe in a browser) — and it is a wsk_ that whissle-sdk takes.

Rather than rebuild a second client against the live platform, this distribution now stands down in favour of the maintained one. Anything you import from it raises an ImportError naming the replacement, so you get told where to go instead of a confusing network failure on your first call.

Moving

You had You want
from whissle import WhissleClient from whissle_sdk import WhissleClient — a different class, different arguments
client.speech_to_text(...) w.asr.transcribe(file, ...)
client.diarize_stt(...) w.asr.transcribe(file, diarize=True, num_speakers=...)
client.llm_text_summarizer(...) no direct equivalent — see below. w.asr.transcribe(file, summarize=...) summarises a recording, not arbitrary text
client.list_asr_models() w.asr.status() — what this deployment loaded
WhissleGatewayClient.transcribe_file(...) w.asr.transcribe(file, ...)
WhissleGatewayClient.transcribe_pcm(...) w.asr.transcribe_pcm(samples, ...)
export WHISSLE_AUTH_TOKEN=wh_… export WHISSLE_API_KEY=wsk_… (create the key in your workspace settings)

whissle-sdk also covers what this package never did: the Voice Agents platform — agents, sessions and traces, streaming chat, browser session tokens, knowledge, tools, compliance, analytics and usage. See its own README for the full surface.

Three things do not carry over. Text machine_translation and text llm_text_summarizer took arbitrary strings and have no equivalent door on the live platform. Speech translation (/asr/translate, /asr/s2s) exists but is deliberately still restricted to the legacy token — those legs compose speech with a language model and have no per-second price, so a workspace key there would be an unbilled door. If you depend on any of the three, talk to us before you migrate.

Name collision

whissle_sdk exports a class also called WhissleClient. It is not this package's WhissleClient and does not take the same arguments. If an editor autocompletes you into whissle, that is this tombstone — the working one is imported from whissle_sdk.

Still installed?

import whissle keeps working and emits a DeprecationWarning; whissle.__version__ and help(whissle) still read. Every client name, and every module that used to hold one, raises ImportError with the migration note above.

Licence

MIT. See LICENSE. Releases up to and including 0.0.1 shipped contradictory metadata — an MIT classifier over GPL-3 licence text. MIT is and always was the intent; 0.2.0 makes the file and the metadata agree.

Maintainers

whissle-sdk must be on PyPI before this release is published, or the install line above points at nothing. Then:

pip install build twine
python -m build
twine check dist/*
twine upload dist/*

Contact

support@whissle.ai

About

Multi-modal python API

Resources

Stars

2 stars

Watchers

2 watching

Forks

Releases

Packages

Contributors

Languages