feat(harness): port vectorstore dual-key + reader RPC (D11, §5.4)

Ports vectorstore/{rest_client,rest_writer,store,reader,embeddings,records}, evidence/model (leaf dep of records), and cli/_guards (require_vector_write_allowed workstation write-guard). Renamed psdwp3->nsp, verbatim. VectorRestClient gains an api_key property so reader/writer clients carry their distinct keys visibly (spec D11: vector_reader / vector_writer on the same endpoint). scripts/create_vector_reader_rpc.sql is NEW: the reader RPCs (search_similar, list_tables) lived server-side in Supabase and were never versioned. Authored now mirroring the writer allowlist pattern (table allowlist, security definer, revoke from anon/authenticated, grant to vector_reader only). Writer RPC ported verbatim. L1: test_vector_dual_key (7 tests) pins the dual-key construction + the workstation write-guard (exit 4 without writer key).
2026-06-26 22:55:40 +02:00
parent eb3bde90e2
commit 796a39d893
13 changed files with 994 additions and 0 deletions
@@ -0,0 +1,50 @@
+import requests
+
+from nsp.config import EmbeddingsConfig
+
+DOC_PREFIX = "search_document: "
+QUERY_PREFIX = "search_query: "
+
+
+class EmbeddingsError(Exception):
+    pass
+
+
+class OllamaEmbeddings:
+    """Client embeddings via Ollama. Applica i prefissi di task richiesti da nomic v2:
+    ometterli degrada il retrieval in modo silenzioso."""
+
+    def __init__(self, cfg: EmbeddingsConfig):
+        self.cfg = cfg
+
+    def _embed(self, texts: list[str]) -> list[list[float]]:
+        url = f"{self.cfg.base_url.rstrip('/')}/api/embed"
+        out: list[list[float]] = []
+        for i in range(0, len(texts), self.cfg.batch_size):
+            batch = texts[i : i + self.cfg.batch_size]
+            try:
+                resp = requests.post(
+                    url, json={"model": self.cfg.model, "input": batch},
+                    timeout=self.cfg.timeout,
+                )
+                resp.raise_for_status()
+            except requests.RequestException as e:
+                raise EmbeddingsError(
+                    f"Ollama non raggiungibile su {self.cfg.base_url} "
+                    f"(modello {self.cfg.model}): {e}"
+                ) from e
+            embeddings = resp.json().get("embeddings", [])
+            for v in embeddings:
+                if len(v) != self.cfg.dim:
+                    raise EmbeddingsError(
+                        f"dimensione embedding inattesa: {len(v)} != {self.cfg.dim} "
+                        f"(modello {self.cfg.model})"
+                    )
+            out.extend(embeddings)
+        return out
+
+    def embed_documents(self, texts: list[str]) -> list[list[float]]:
+        return self._embed([DOC_PREFIX + t for t in texts])
+
+    def embed_query(self, text: str) -> list[float]:
+        return self._embed([QUERY_PREFIX + text])[0]