feat: LLM module — Ollama client with transcript refinement
This commit is contained in:
@@ -0,0 +1,44 @@
|
|||||||
|
import httpx
|
||||||
|
|
||||||
|
SYSTEM_PROMPT = """Du bist ein präziser Schreibassistent.
|
||||||
|
Du bekommst einen rohen Sprachtranskript und optionale Instruktionen des Nutzers.
|
||||||
|
Deine Aufgabe:
|
||||||
|
1. Bereinige den Text (Füllwörter, Wiederholungen, Tippfehler)
|
||||||
|
2. Strukturiere ihn mit Markdown-Überschriften wenn sinnvoll
|
||||||
|
3. Erzeuge einen passenden deutschen Titel als H1
|
||||||
|
4. Beachte Instruktionen des Nutzers wenn vorhanden
|
||||||
|
5. Antworte NUR mit dem fertigen Markdown — kein Kommentar, keine Erklärung
|
||||||
|
|
||||||
|
Format:
|
||||||
|
# Titel
|
||||||
|
|
||||||
|
Inhalt...
|
||||||
|
"""
|
||||||
|
|
||||||
|
|
||||||
|
class OllamaClient:
|
||||||
|
def __init__(self, base_url: str = "http://localhost:11434"):
|
||||||
|
self.base_url = base_url
|
||||||
|
|
||||||
|
async def list_models(self) -> list[str]:
|
||||||
|
async with httpx.AsyncClient() as client:
|
||||||
|
r = await client.get(f"{self.base_url}/api/tags")
|
||||||
|
r.raise_for_status()
|
||||||
|
return [m["name"] for m in r.json().get("models", [])]
|
||||||
|
|
||||||
|
async def refine(
|
||||||
|
self,
|
||||||
|
raw_text: str,
|
||||||
|
instructions: str = "",
|
||||||
|
model: str = "gemma3:12b",
|
||||||
|
) -> str:
|
||||||
|
prompt = f"Transkript:\n{raw_text}"
|
||||||
|
if instructions.strip():
|
||||||
|
prompt += f"\n\nInstruktionen:\n{instructions.strip()}"
|
||||||
|
async with httpx.AsyncClient(timeout=120) as client:
|
||||||
|
r = await client.post(
|
||||||
|
f"{self.base_url}/api/generate",
|
||||||
|
json={"model": model, "prompt": prompt, "system": SYSTEM_PROMPT, "stream": False},
|
||||||
|
)
|
||||||
|
r.raise_for_status()
|
||||||
|
return r.json()["response"]
|
||||||
@@ -0,0 +1,37 @@
|
|||||||
|
import pytest
|
||||||
|
from unittest.mock import AsyncMock, patch, MagicMock
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_refine_calls_ollama():
|
||||||
|
from llm import OllamaClient
|
||||||
|
mock_response = MagicMock()
|
||||||
|
mock_response.json.return_value = {"response": "# Titel\n\nInhalt."}
|
||||||
|
mock_response.raise_for_status = MagicMock()
|
||||||
|
|
||||||
|
with patch("httpx.AsyncClient") as MockClient:
|
||||||
|
instance = MockClient.return_value.__aenter__.return_value
|
||||||
|
instance.post = AsyncMock(return_value=mock_response)
|
||||||
|
client = OllamaClient(base_url="http://localhost:11434")
|
||||||
|
result = await client.refine(
|
||||||
|
raw_text="Das ist ein test.",
|
||||||
|
instructions="Mach eine Zusammenfassung.",
|
||||||
|
model="gemma3:12b",
|
||||||
|
)
|
||||||
|
assert "Inhalt" in result
|
||||||
|
instance.post.assert_called_once()
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_list_models_returns_list():
|
||||||
|
from llm import OllamaClient
|
||||||
|
mock_response = MagicMock()
|
||||||
|
mock_response.json.return_value = {"models": [{"name": "gemma3:12b"}, {"name": "mistral:7b"}]}
|
||||||
|
mock_response.raise_for_status = MagicMock()
|
||||||
|
|
||||||
|
with patch("httpx.AsyncClient") as MockClient:
|
||||||
|
instance = MockClient.return_value.__aenter__.return_value
|
||||||
|
instance.get = AsyncMock(return_value=mock_response)
|
||||||
|
client = OllamaClient(base_url="http://localhost:11434")
|
||||||
|
models = await client.list_models()
|
||||||
|
assert "gemma3:12b" in models
|
||||||
Reference in New Issue
Block a user