LlamaIndex Readers Integration: Lokutor

Transcribe audio files into LlamaIndex Documents with Lokutor speech-to-text (Conv 2.0; English, Spanish, Catalan, Galician, Basque, Portuguese, French, Italian, German), with optional speaker diarization.

pip install llama-index-readers-lokutor
export LOKUTOR_API_KEY=...   # https://app.lokutor.com

from llama_index.readers.lokutor import LokutorReader

reader = LokutorReader(language="en", diarize=True)  # language=None detects it
docs = reader.load_data("meeting.wav")
print(docs[0].text, docs[0].metadata["speakers"])

# index it
from llama_index.core import VectorStoreIndex
index = VectorStoreIndex.from_documents(docs)

• load_data / aload_data take a path, bytes or a file-like object, and return one Document.
• Metadata: file_path, language_code; with diarize=True also speakers; with include_words=True the per-word timings in words.
• Lokutor decodes WAV only (PCM 8 or 16 bit, mono or stereo, 8-48 kHz). Convert MP3/M4A first.
• API errors (bad key, unsupported audio, no balance) raise RuntimeError with Lokutor's message.

API docs: docs.lokutor.com. Licence: Apache-2.0.