first

app/clients/deepgram_client.py

@@ -0,0 +1,60 @@
+"""Klient pro Deepgram pre-recorded přepis (REST, přes httpx).
+
+Nepoužíváme těžké SDK — voláme REST endpoint přímo, aby nebyl závislý na verzích.
+"""
+import httpx
+
+from ..config import DEEPGRAM_BASE_URL, UPSTREAM_TIMEOUT_SECONDS
+from ..errors import UpstreamError
+from ..logging_config import get_logger
+
+log = get_logger("audio-transcription.deepgram")
+
+
+async def transcribe(
+    *,
+    api_key: str,
+    audio: bytes,
+    content_type: str,
+    model: str,
+    language: str,
+    diarize: bool,
+    smart_format: bool,
+) -> str:
+    """Přepíše audio Deepgramem a vrátí prostý transkript (channels[0].alternatives[0].transcript)."""
+    params = {
+        "model": model,
+        "language": language,
+        "diarize": "true" if diarize else "false",
+        "smart_format": "true" if smart_format else "false",
+    }
+    headers = {
+        "Authorization": f"Token {api_key}",
+        "Content-Type": content_type or "application/octet-stream",
+    }
+    url = f"{DEEPGRAM_BASE_URL}/listen"
+
+    try:
+        async with httpx.AsyncClient(timeout=UPSTREAM_TIMEOUT_SECONDS) as client:
+            resp = await client.post(url, params=params, headers=headers, content=audio)
+    except httpx.HTTPError as exc:
+        log.error("Deepgram request failed: %s", exc)
+        raise UpstreamError(f"Deepgram request se nezdařil: {exc}") from exc
+
+    if resp.status_code >= 400:
+        # Deepgram vrací chybu v JSON; nikdy nelogujeme klíč (ten je jen v hlavičce).
+        snippet = resp.text[:500]
+        log.warning("Deepgram returned %s: %s", resp.status_code, snippet)
+        raise UpstreamError(
+            f"Deepgram vrátil chybu {resp.status_code}.",
+            detail=snippet,
+        )
+
+    data = resp.json()
+    try:
+        channels = data["results"]["channels"]
+        alternative = channels[0]["alternatives"][0]
+        return alternative.get("transcript", "") or ""
+    except (KeyError, IndexError, TypeError) as exc:
+        log.error("Unexpected Deepgram response shape: %s", exc)
+        raise UpstreamError("Neočekávaná struktura odpovědi z Deepgramu.") from exc

app/clients/openai_client.py

@@ -0,0 +1,87 @@
+"""Klient pro OpenAI — Whisper přepis + Chat Completions (slučování), přes httpx."""
+import httpx
+
+from ..config import OPENAI_BASE_URL, UPSTREAM_TIMEOUT_SECONDS
+from ..errors import UpstreamError
+from ..logging_config import get_logger
+
+log = get_logger("audio-transcription.openai")
+
+
+async def transcribe(
+    *,
+    api_key: str,
+    audio: bytes,
+    filename: str,
+    content_type: str,
+    model: str,
+    language: str | None = None,
+) -> str:
+    """Přepíše audio přes OpenAI Whisper (/audio/transcriptions) a vrátí text."""
+    headers = {"Authorization": f"Bearer {api_key}"}
+    files = {"file": (filename or "audio.mp3", audio, content_type or "application/octet-stream")}
+    form = {"model": model}
+    if language:
+        form["language"] = language
+    url = f"{OPENAI_BASE_URL}/audio/transcriptions"
+
+    try:
+        async with httpx.AsyncClient(timeout=UPSTREAM_TIMEOUT_SECONDS) as client:
+            resp = await client.post(url, headers=headers, data=form, files=files)
+    except httpx.HTTPError as exc:
+        log.error("OpenAI transcription request failed: %s", exc)
+        raise UpstreamError(f"OpenAI Whisper request se nezdařil: {exc}") from exc
+
+    if resp.status_code >= 400:
+        snippet = resp.text[:500]
+        log.warning("OpenAI transcription returned %s: %s", resp.status_code, snippet)
+        raise UpstreamError(
+            f"OpenAI Whisper vrátil chybu {resp.status_code}.",
+            detail=snippet,
+        )
+
+    data = resp.json()
+    return data.get("text", "") or ""
+
+
+async def merge_transcripts(
+    *,
+    api_key: str,
+    model: str,
+    system_prompt: str,
+    text_deepgram: str,
+    text_whisper: str,
+) -> str:
+    """Sloučí dva přepisy do jednoho přes Chat Completions dle system promptu."""
+    headers = {"Authorization": f"Bearer {api_key}"}
+    payload = {
+        "model": model,
+        "messages": [
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": f"Text z Deepgram: {text_deepgram}"},
+            {"role": "user", "content": f"Text z OpenAI Whisper: {text_whisper}"},
+        ],
+    }
+    url = f"{OPENAI_BASE_URL}/chat/completions"
+
+    try:
+        async with httpx.AsyncClient(timeout=UPSTREAM_TIMEOUT_SECONDS) as client:
+            resp = await client.post(url, headers=headers, json=payload)
+    except httpx.HTTPError as exc:
+        log.error("OpenAI chat request failed: %s", exc)
+        raise UpstreamError(f"OpenAI chat (merge) request se nezdařil: {exc}") from exc
+
+    if resp.status_code >= 400:
+        snippet = resp.text[:500]
+        log.warning("OpenAI chat returned %s: %s", resp.status_code, snippet)
+        raise UpstreamError(
+            f"OpenAI chat (merge) vrátil chybu {resp.status_code}.",
+            detail=snippet,
+        )
+
+    data = resp.json()
+    try:
+        return data["choices"][0]["message"]["content"] or ""
+    except (KeyError, IndexError, TypeError) as exc:
+        log.error("Unexpected OpenAI chat response shape: %s", exc)
+        raise UpstreamError("Neočekávaná struktura odpovědi z OpenAI chatu.") from exc