feat(processor_lmstudio): openai-compatible client with health-check and retry

src/journal_bot/processor_lmstudio.py

@@ -0,0 +1,77 @@
+import base64
+import json
+from pathlib import Path
+from typing import Optional
+import httpx
+from pydantic import ValidationError
+from .processor_protocol import ProcessorInput, ProcessorOutput
+
+
+class LMStudioProcessor:
+    """OpenAI-compatible processor against LM Studio local server."""
+
+    def __init__(
+        self,
+        base_url: str,
+        model: str,
+        system_prompt: str,
+        max_retries: int = 2,
+        timeout: float = 120.0,
+    ):
+        self.base_url = base_url.rstrip("/")
+        self.model = model
+        self.system_prompt = system_prompt
+        self.max_retries = max_retries
+        self._http = httpx.Client(timeout=timeout)
+
+    def health_check(self) -> bool:
+        try:
+            r = self._http.get(f"{self.base_url}/models", timeout=2.0)
+            return r.status_code == 200
+        except httpx.HTTPError:
+            return False
+
+    def _user_message(self, payload: ProcessorInput) -> dict:
+        ctx = {
+            "today": payload.today,
+            "weekday": payload.weekday,
+            "received_time": payload.received_time,
+            "persons": payload.persons,
+            "projects": payload.projects,
+            "text": payload.text,
+            "image_embed": payload.image_embed,
+            "image_caption": payload.image_caption,
+        }
+        content: list[dict] = [{"type": "text", "text": json.dumps(ctx, ensure_ascii=False)}]
+        if payload.image_local_path:
+            p = Path(payload.image_local_path)
+            if p.exists():
+                b64 = base64.b64encode(p.read_bytes()).decode("ascii")
+                content.append({
+                    "type": "image_url",
+                    "image_url": {"url": f"data:image/jpeg;base64,{b64}"},
+                })
+        return {"role": "user", "content": content}
+
+    def process(self, payload: ProcessorInput) -> ProcessorOutput:
+        body = {
+            "model": self.model,
+            "messages": [
+                {"role": "system", "content": self.system_prompt},
+                self._user_message(payload),
+            ],
+            "response_format": {"type": "json_object"},
+            "temperature": 0.2,
+        }
+        last_error: Optional[Exception] = None
+        for _ in range(self.max_retries + 1):
+            r = self._http.post(f"{self.base_url}/chat/completions", json=body)
+            r.raise_for_status()
+            content = r.json()["choices"][0]["message"]["content"]
+            try:
+                data = json.loads(content)
+                return ProcessorOutput.model_validate(data)
+            except (json.JSONDecodeError, ValidationError) as e:
+                last_error = e
+                continue
+        raise RuntimeError(f"LM Studio produced invalid output after retries: {last_error}")

tests/test_processor_lmstudio.py

@@ -0,0 +1,87 @@
+import json
+import httpx
+import respx
+import pytest
+from journal_bot.processor_lmstudio import LMStudioProcessor
+from journal_bot.processor_protocol import ProcessorInput
+
+
+@pytest.fixture
+def processor():
+    return LMStudioProcessor(
+        base_url="http://localhost:1234/v1",
+        model="qwen/qwen3-vl-8b",
+        system_prompt="SYS",
+    )
+
+
+def make_input(text="Hallo Welt"):
+    return ProcessorInput(
+        today="2026-06-14",
+        weekday="Sonntag",
+        received_time="14:32",
+        persons=[],
+        projects=[],
+        text=text,
+    )
+
+
+@respx.mock
+def test_health_check_ok(processor):
+    respx.get("http://localhost:1234/v1/models").mock(
+        return_value=httpx.Response(200, json={"data": [{"id": "qwen/qwen3-vl-8b"}]})
+    )
+    assert processor.health_check() is True
+
+
+@respx.mock
+def test_health_check_down(processor):
+    respx.get("http://localhost:1234/v1/models").mock(
+        side_effect=httpx.ConnectError("no")
+    )
+    assert processor.health_check() is False
+
+
+@respx.mock
+def test_process_returns_validated_output(processor):
+    payload = {
+        "target_date": "2026-06-14",
+        "target_path": "05 Daily Notes/2026-06-14.md",
+        "entry_markdown": "## 14:32\nHallo Welt",
+        "clarifications": [],
+        "raw_excluded": [],
+    }
+    respx.post("http://localhost:1234/v1/chat/completions").mock(
+        return_value=httpx.Response(200, json={
+            "choices": [{"message": {"content": json.dumps(payload)}}],
+        })
+    )
+    out = processor.process(make_input())
+    assert out.target_date == "2026-06-14"
+    assert out.entry_markdown.startswith("## 14:32")
+
+
+@respx.mock
+def test_process_retries_on_schema_mismatch(processor):
+    bad = {"choices": [{"message": {"content": json.dumps({"target_date": "bad"})}}]}
+    good = {"choices": [{"message": {"content": json.dumps({
+        "target_date": "2026-06-14",
+        "target_path": "05 Daily Notes/2026-06-14.md",
+        "entry_markdown": "## 14:32\nHallo",
+        "clarifications": [], "raw_excluded": [],
+    })}}]}
+    respx.post("http://localhost:1234/v1/chat/completions").mock(
+        side_effect=[httpx.Response(200, json=bad), httpx.Response(200, json=good)]
+    )
+    out = processor.process(make_input())
+    assert out.target_date == "2026-06-14"
+
+
+@respx.mock
+def test_process_raises_after_max_retries(processor):
+    bad = {"choices": [{"message": {"content": "not json"}}]}
+    respx.post("http://localhost:1234/v1/chat/completions").mock(
+        return_value=httpx.Response(200, json=bad)
+    )
+    with pytest.raises(Exception):
+        processor.process(make_input())