Initial project import

2026-08-14 09:40:51 +03:00
parent 00040e5ce4
commit d7099bf80d
146 changed files with 30509 additions and 1055 deletions
@@ -0,0 +1,102 @@
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[2]
+sys.path.insert(0, str(ROOT / "scripts"))
+
+import search_1c_code_vectors as search_cli  # noqa: E402
+
+
+def test_search_code_vectors_sends_model_and_strict_query_embedding(monkeypatch) -> None:
+    calls: list[tuple[str, dict]] = []
+
+    monkeypatch.setattr(
+        search_cli,
+        "embed_texts",
+        lambda texts, **kwargs: [[1.0, 0.0, 0.0]],
+    )
+
+    def fake_adapter_call(
+        adapter_url: str,
+        method: str,
+        payload: dict,
+        *,
+        timeout_seconds: int = 180,
+    ) -> dict:
+        calls.append((method, payload))
+        return {"status": "ok", "matches": []}
+
+    monkeypatch.setattr(search_cli, "adapter_call", fake_adapter_call)
+
+    result = search_cli.search_code_vectors(
+        adapter_url="http://adapter/rpc",
+        base_id="upo_test",
+        query="где рассчитывается налог",
+        embedding_provider="openai-compatible",
+        embedding_model="test-code-model",
+        embedding_base_url="http://embeddings",
+        dimensions=3,
+    )
+
+    assert result["status"] == "ok"
+    assert result["client_embedding"]["stored_embedding_model"] == "openai-compatible:test-code-model@d3"
+    assert calls == [
+        (
+            "metadata.code_vector.search",
+            {
+                "base_id": "upo_test",
+                "query": "где рассчитывается налог",
+                "query_embedding": [1.0, 0.0, 0.0],
+                "embedding_model": "openai-compatible:test-code-model@d3",
+                "limit": 10,
+                "scan_limit": 2000,
+                "verify": True,
+                "strict": True,
+            },
+        )
+    ]
+
+
+def test_qwen3_search_uses_query_instruction_but_keeps_lexical_query_plain(monkeypatch) -> None:
+    embedded_texts: list[str] = []
+    calls: list[tuple[str, dict]] = []
+
+    def fake_embed_texts(texts: list[str], **kwargs) -> list[list[float]]:
+        embedded_texts.extend(texts)
+        return [[0.5, 0.5]]
+
+    monkeypatch.setattr(search_cli, "embed_texts", fake_embed_texts)
+
+    def fake_adapter_call(
+        adapter_url: str,
+        method: str,
+        payload: dict,
+        *,
+        timeout_seconds: int = 180,
+    ) -> dict:
+        calls.append((method, payload))
+        return {"status": "ok", "matches": []}
+
+    monkeypatch.setattr(search_cli, "adapter_call", fake_adapter_call)
+
+    result = search_cli.search_code_vectors(
+        adapter_url="http://adapter/rpc",
+        base_id="upo_test",
+        query="обработка проведения документа",
+        embedding_provider="openai-compatible",
+        embedding_model="qwen3-embedding-0.6b",
+        embedding_base_url="http://embeddings",
+    )
+
+    assert embedded_texts == [
+        "Instruct: Given a natural-language software task, retrieve the relevant "
+        "1C Enterprise BSL source-code fragment that implements or explains it\n"
+        "Query:обработка проведения документа"
+    ]
+    assert calls[0][1]["query"] == "обработка проведения документа"
+    assert result["client_embedding"]["query_instruction"] == (
+        search_cli.DEFAULT_QWEN3_CODE_RETRIEVAL_INSTRUCTION
+    )