Initial project import

scripts/rag_embedding_providers.py

@@ -41,12 +41,16 @@ def embed_texts_openai_compatible(
     *,
     model: str,
     base_url: str,
+    dimensions: int = 0,
     api_key: str = "",
     timeout_seconds: int = 120,
 ) -> list[list[float]]:
     if not model:
         raise ValueError("embedding_model is required")
-    body = json.dumps({"model": model, "input": texts}, ensure_ascii=False).encode("utf-8")
+    request_payload: dict[str, Any] = {"model": model, "input": texts}
+    if int(dimensions or 0) > 0:
+        request_payload["dimensions"] = int(dimensions)
+    body = json.dumps(request_payload, ensure_ascii=False).encode("utf-8")
     headers = {"Content-Type": "application/json"}
     if api_key:
         headers["Authorization"] = f"Bearer {api_key}"
@@ -67,6 +71,13 @@ def embed_texts_openai_compatible(
         if not isinstance(embedding, list) or not embedding:
             raise ValueError("Embedding response item has no embedding[]")
         vector = [float(value) for value in embedding]
+        if int(dimensions or 0) > 0:
+            if len(vector) < int(dimensions):
+                raise ValueError(
+                    f"Embedding response returned {len(vector)} dimensions, "
+                    f"fewer than requested {int(dimensions)}"
+                )
+            vector = vector[: int(dimensions)]
         by_index[int(index)] = l2_normalize(vector)
     vectors = [by_index[index] for index in range(len(texts)) if index in by_index]
     if len(vectors) != len(texts):
@@ -96,6 +107,7 @@ def embed_texts(
             texts,
             model=model,
             base_url=base_url,
+            dimensions=dimensions,
             api_key=api_key,
             timeout_seconds=timeout_seconds,
         )