bugfix: rag memleak with weakref (#294)

* fix rag memleak with weakref

* Update api/rag.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

---------

Co-authored-by: chen zhang <chen.zhang@rockbundcapital.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Author: 3 people

api/rag.py

@@ -1,4 +1,5 @@
 import logging
+import weakref
 import re
 from dataclasses import dataclass
 from typing import Any, List, Tuple, Dict
@@ -242,14 +243,17 @@ def __init__(self, provider="google", model=None, use_s3: bool = False):  # noqa
         self.memory = Memory()
         self.embedder = get_embedder()
 
+        self_weakref = weakref.ref(self)
         # Patch: ensure query embedding is always single string for Ollama
         def single_string_embedder(query):
             # Accepts either a string or a list, always returns embedding for a single string
             if isinstance(query, list):
                 if len(query) != 1:
                     raise ValueError("Ollama embedder only supports a single string")
                 query = query[0]
-            return self.embedder(input=query)
+            instance = self_weakref()
+            assert instance is not None, "RAG instance is no longer available, but the query embedder was called."
+            return instance.embedder(input=query)
 
         # Use single string embedder for Ollama, regular embedder for others
         self.query_embedder = single_string_embedder if self.is_ollama_embedder else self.embedder