Initial commit

2026-10-12 09:55:49 +09:00
commit 5eedbdd039
@@ -0,0 +1,68 @@
+import json
+import threading
+from http.server import BaseHTTPRequestHandler, HTTPServer
+
+import pytest
+
+pytest.importorskip("anthropic")
+
+from miniflow import llm  # noqa: E402
+
+
+class FakeAnthropic(BaseHTTPRequestHandler):
+    received: list[dict] = []
+    headers_seen: list[dict] = []
+    stop_reason = "end_turn"
+
+    def do_POST(self):
+        body = json.loads(self.rfile.read(int(self.headers["Content-Length"])))
+        FakeAnthropic.received.append(body)
+        FakeAnthropic.headers_seen.append(dict(self.headers))
+        reply = {
+            "id": "msg_test", "type": "message", "role": "assistant", "model": body["model"],
+            "content": [{"type": "text", "text": " network "}],
+            "stop_reason": FakeAnthropic.stop_reason, "stop_sequence": None,
+            "usage": {"input_tokens": 10, "output_tokens": 2},
+        }
+        payload = json.dumps(reply).encode()
+        self.send_response(200)
+        self.send_header("Content-Type", "application/json")
+        self.send_header("Content-Length", str(len(payload)))
+        self.end_headers()
+        self.wfile.write(payload)
+
+    def log_message(self, *args):
+        pass
+
+
+@pytest.fixture()
+def fake_server(monkeypatch):
+    server = HTTPServer(("127.0.0.1", 0), FakeAnthropic)
+    threading.Thread(target=server.serve_forever, daemon=True).start()
+    FakeAnthropic.received.clear()
+    FakeAnthropic.headers_seen.clear()
+    FakeAnthropic.stop_reason = "end_turn"
+    monkeypatch.setenv("MINIFLOW_LLM", "anthropic")
+    monkeypatch.setenv("ANTHROPIC_API_KEY", "test-key")
+    monkeypatch.setenv("ANTHROPIC_BASE_URL", f"http://127.0.0.1:{server.server_port}")
+    monkeypatch.delenv("ANTHROPIC_MODEL", raising=False)
+    yield server
+    server.shutdown()
+
+
+def test_anthropic_request_and_reply(fake_server):
+    reply, provider = llm.complete({"temperature": 0}, "문의: VPN", "Reply with the label only.", {})
+    assert (reply, provider) == ("network", "anthropic")
+    sent = FakeAnthropic.received[-1]
+    assert sent["model"] == "claude-opus-5-5"
+    assert sent["system"] == "Reply with the label only."
+    assert sent["messages"] == [{"role": "user", "content": "문의: VPN"}]
+    assert sent["fallbacks"] == "default"
+    assert "temperature" not in sent
+    assert "server-side-fallback-2026-07-01" in FakeAnthropic.headers_seen[-1].get("anthropic-beta", "")
+
+
+def test_refusal_is_reported(fake_server):
+    FakeAnthropic.stop_reason = "refusal"
+    with pytest.raises(llm.LLMError, match="declined"):
+        llm.complete({}, "hi", "", {})
@@ -0,0 +1,66 @@
+import json
+import threading
+from http.server import BaseHTTPRequestHandler, HTTPServer
+
+import pytest
+
+from miniflow import llm
+
+
+class FakeOllama(BaseHTTPRequestHandler):
+    received: list[dict] = []
+
+    def do_POST(self):
+        body = json.loads(self.rfile.read(int(self.headers["Content-Length"])))
+        FakeOllama.received.append(body)
+        if body["model"] == "missing-model":
+            self.send_response(404)
+            self.end_headers()
+            self.wfile.write(b'{"error":"model not found"}')
+            return
+        reply = {"message": {"role": "assistant", "content": " network \n"}, "done": True}
+        payload = json.dumps(reply).encode()
+        self.send_response(200)
+        self.send_header("Content-Type", "application/json")
+        self.send_header("Content-Length", str(len(payload)))
+        self.end_headers()
+        self.wfile.write(payload)
+
+    def log_message(self, *args):
+        pass
+
+
+@pytest.fixture()
+def fake_server(monkeypatch):
+    server = HTTPServer(("127.0.0.1", 0), FakeOllama)
+    threading.Thread(target=server.serve_forever, daemon=True).start()
+    FakeOllama.received.clear()
+    monkeypatch.setenv("MINIFLOW_LLM", "ollama")
+    monkeypatch.setenv("OLLAMA_HOST", f"127.0.0.1:{server.server_port}")
+    monkeypatch.setenv("OLLAMA_MODEL", "gemma3:1b")
+    yield server
+    server.shutdown()
+
+
+def test_ollama_request_and_reply(fake_server):
+    reply, provider = llm.complete({"temperature": 0}, "문의: VPN", "Reply with the label only.", {})
+    assert (reply, provider) == ("network", "ollama")
+    sent = FakeOllama.received[-1]
+    assert sent["model"] == "gemma3:1b"
+    assert sent["stream"] is False
+    assert sent["messages"][0] == {"role": "system", "content": "Reply with the label only."}
+    assert sent["messages"][1]["content"] == "문의: VPN"
+
+
+def test_missing_model_gives_pull_hint(fake_server, monkeypatch):
+    monkeypatch.setenv("OLLAMA_MODEL", "missing-model")
+    with pytest.raises(llm.LLMError, match="ollama pull missing-model"):
+        llm.complete({}, "hi", "", {})
+
+
+def test_unreachable_server(monkeypatch):
+    monkeypatch.setenv("MINIFLOW_LLM", "ollama")
+    monkeypatch.setenv("OLLAMA_HOST", "http://127.0.0.1:9")
+    monkeypatch.setenv("OLLAMA_TIMEOUT", "2")
+    with pytest.raises(llm.LLMError, match="Cannot reach Ollama"):
+        llm.complete({}, "hi", "", {})
@@ -0,0 +1,80 @@
+import json
+from pathlib import Path
+
+import pytest
+
+from miniflow import WorkflowError, load_workflow, run_workflow
+from miniflow.template import TemplateError, render
+
+FLOWS = Path(__file__).resolve().parents[1] / "flows"
+
+
+@pytest.fixture(autouse=True)
+def isolated_memory(tmp_path, monkeypatch):
+    monkeypatch.setenv("MINIFLOW_MEMORY_FILE", str(tmp_path / "memory.json"))
+    monkeypatch.setenv("MINIFLOW_LLM", "mock")
+
+
+def run(name, **inputs):
+    return run_workflow(load_workflow(FLOWS / name), inputs, verbose=False)
+
+
+def test_every_sample_flow_is_valid():
+    for path in FLOWS.glob("*.json"):
+        assert load_workflow(path).order()
+
+
+def test_template_keeps_type_for_whole_placeholder():
+    context = {"nodes": {"tool": {"output": {"temp": 20}}}}
+    assert render("{{nodes.tool.output}}", context) == {"temp": 20}
+    assert render("temp={{nodes.tool.output.temp}}", context) == "temp=20"
+    with pytest.raises(TemplateError):
+        render("{{nodes.missing.output}}", context)
+
+
+def test_llm_only():
+    report = run("01-llm-only.json", text="오늘 회의에서 일정이 확정되었습니다.")
+    assert report["status"] == "completed"
+    assert "Mock" in report["result"]
+
+
+def test_tool_data_flows_into_llm():
+    report = run("02-llm-tool.json", city="제주", question="우산이 필요할까?")
+    assert "제주" in report["result"] and "80%" in report["result"]
+
+
+def test_unknown_city_fails_at_tool_node():
+    report = run("02-llm-tool.json", city="평양", question="날씨?")
+    assert report["status"] == "failed"
+    assert report["trace"][-1]["id"] == "weather"
+
+
+def test_memory_survives_between_runs():
+    run("03-save-profile.json", user="minji", preference="실내 전시")
+    report = run("04-recommend-with-memory.json", user="minji", city="부산")
+    assert "실내 전시" in report["result"]
+    again = run("04-recommend-with-memory.json", user="minji", city="강릉")
+    recall = next(step for step in again["trace"] if step["id"] == "recall")
+    assert recall["output"]["last_city"] == "부산"
+
+
+def test_tech_support_ticket():
+    report = run("05-tech-support-ticket.json", user="kim", ticket="비밀번호를 잊어버려서 로그인이 안 돼요")
+    assert report["result"]["category"] == "account"
+    assert report["result"]["faq_id"] == "FAQ-001"
+
+
+def test_cycle_is_rejected(tmp_path):
+    flow = {
+        "nodes": [{"id": "a", "type": "input"}, {"id": "b", "type": "output", "value": "x"}],
+        "edges": [["a", "b"], ["b", "a"]],
+    }
+    path = tmp_path / "cycle.json"
+    path.write_text(json.dumps(flow), encoding="utf-8")
+    with pytest.raises(WorkflowError):
+        load_workflow(path)
+
+
+def test_missing_input_is_reported():
+    with pytest.raises(WorkflowError):
+        run("02-llm-tool.json", city="서울")