egune 1.0.2

Egune Library

Egune Python Client Library

Async Python client for the Chibolegune API, providing easy access to speech, text, OCR, embedding, and video generation services.

Installation

pip install .

Features

• Speech-to-Text (STT) with dual-channel and streaming support.
• Text-to-Speech (TTS) with actor presets, voice cloning, and emotional synthesis.
• Punctuation restoration for unpunctuated text.
• Language detection for Mongolian and English.
• Latin to Cyrillic transliteration.
• Text embedding for semantic similarity.
• OCR for extracting text from images.
• Kimo for Traditional <-> Cyrillic conversion.
• Video generation from prompts and images.

Quickstart Example

import asyncio
from egune import EguneClient

async def main():
    async with EguneClient("http://localhost:8000") as client:
        # Speech-to-Text
        with open("audio.wav", "rb") as f:
            audio_bytes = f.read()
        stt_result = await client.stt(audio=audio_bytes, model="citrinet", audio_dtype="int16")
        print("STT:", stt_result.full_text)

        # Text-to-Speech
        tts_result = await client.tts("234")
        print("TTS audio URL:", tts_result.audio_url)

        # Punctuation
        punctuated = await client.punctuate("sain uu")
        print("Punctuated:", punctuated)

        # Language Detection
        lang = await client.detect_language(text="Hello world")
        print("Language:", lang.language)

        # Latin to Cyrillic
        cyrillic = await client.latin_to_cyrillic("sain uu")
        print("Cyrillic:", cyrillic)

        # Embedding
        emb = await client.embed(text="Hello world")
        print("Embedding (first 5):", emb.embedding[:5])

        # OCR
        with open("image.png", "rb") as f:
            image_bytes = f.read()
        ocr = await client.ocr(image=image_bytes)
        print("OCR blocks:", ocr.blocks)

        # Kimo
        kimo_text = await client.kimo("2", direction="traditional->cyrillic")
        print("Kimo:", kimo_text)

        # Video Generation
        video_job = await client.generate_video(image_path="face.png", prompt="A person is talking")
        print("Video job status:", video_job.status)

asyncio.run(main())

REST Endpoints (Summary)

• /stt/ — Speech-to-text (audio bytes or MinIO path)
• /tts/ — Text-to-speech
• /punctuator/ — Restore punctuation
• /language-detector/ — Detect language
• /latin-to-cyrillic/ — Convert Latin to Cyrillic
• /embedder/ — Text embedding
• /ocr/ — OCR from image
• /kimo/ — Traditional <-> Cyrillic conversion
• /video-generation/ — Video generation

See the docs/endpoints.md for full details.

Advanced: ZMQClient

For low-latency, internal service-to-service calls, use egune.ZMQClient:

from egune import ZMQClient

async def zmq_example():
    async with ZMQClient("tcp://localhost:5600") as client:
        meta, data = await client.call("whisper", {"dtype": "int16"}, audio_bytes)