Lampline
Copyright 2026 Rohan Salvi
https://github.com/C4rohan/lampline

Created and maintained by Rohan Salvi. Under the Apache License 2.0
(section 4(d)), anyone who redistributes Lampline or a work derived from it
must keep this NOTICE file, or its attribution notices, with what they
distribute.

The engine in engine/ (vpengine) is this project's own code and uses Pipecat
(https://github.com/pipecat-ai/pipecat, BSD-2-Clause, Copyright (c) Daily) as a
library, installed from PyPI and not redistributed here. Model providers
(Anthropic, OpenAI, Google, Azure, AWS, Groq, Mistral and others) are reached
through Pipecat's services for them and their own client libraries, installed
from PyPI. SIP calls go through jambonz (https://jambonz.org, MIT), which you
run separately; this project only speaks its webhook and websocket protocol.

The engine's web app bundles the IBM Plex Sans and IBM Plex Mono fonts (Copyright 2019 IBM Corp.,
SIL Open Font License 1.1, see engine/src/vpengine/static/fonts/OFL.txt).
The web app's built JavaScript bundles open-source libraries (React, React
Flow, Radix UI, lucide, sonner, dagre, Tailwind CSS and their dependencies;
MIT, ISC, BSD and Apache-2.0 licences). Their licence texts ship beside the app
in engine/src/vpengine/static/THIRD-PARTY-LICENSES.txt, generated at build time.

Kokoro speech on this computer comes from pipecat-ai's "kokoro" extra, which
installs kokoro-onnx and, through it, phonemizer and espeak-ng (GPL-3.0). They
are installed from PyPI on your machine and not redistributed here; if you ship
a package or container image that includes them, follow their licence terms.

Moonshine speech to text (https://github.com/moonshine-ai/moonshine, MIT,
Copyright (c) 2025 Moonshine AI) is installed from PyPI as the optional
"moonshine" extra (moonshine-voice), and its English models download from
Moonshine's model hub at run time; neither is redistributed here.

Parakeet speech to text runs through onnx-asr (https://github.com/istupakov/onnx-asr,
MIT), installed from PyPI. Its models are NVIDIA's Parakeet TDT 0.6B v2 and v3
(https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2, CC-BY-4.0, Copyright NVIDIA)
in istupakov's ONNX exports, downloaded from Hugging Face at run time and not
redistributed here. Qwen3-TTS runs on a server you set up; nothing of it is
included here.

The engine reads PDFs with pypdf (BSD-3-Clause), installed from PyPI. For
knowledge search it can download sentence-transformers/all-MiniLM-L6-v2
(Apache-2.0) from Hugging Face at run time; the model is not redistributed here.

--------------------------------------------------------------------------------

Importers (cli/src/vk/importers)

The Dialogflow CX and CX Agent Studio importers read the export formats that
Google documents publicly (Dialogflow CX "JSON package" and CX Agent Studio
app export). They contain no code from Google. The test exports are synthetic,
made-up businesses. Dialogflow and CX Agent Studio are names of Google
products; this project is not affiliated with or endorsed by Google.