fix: patch bug on TEI embedding lookup

[cpacker]

Please describe the purpose of this pull request.

Same bug we patched before with having to unpack TEI embedding payload. However, I noticed that in LocalIndex the embeddings on a query get created again, so our previous patch doesn't apply, and thus we get a similar error. This patch uses QueryBundle to make sure we don't create another embedding on query, so that we don't have another unpacking problem.

How to test

Use memgpt quickstart

On main, ask "insert "banana" into archival memory", then "search archival memory for "banana"" and you'll get this error:

💭 Processing user request to search archival memory for a specific term.
⚡🧠 [function] updating memory with archival_memory_search
'page'
{'query': 'banana'}
Archival search error unsupported operand type(s) for /: 'dict' and 'int'
Error calling function archival_memory_search: unsupported operand type(s) for /: 'dict' and 'int'
Traceback (most recent call last):
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/memgpt/agent.py", line 521, in handle_ai_response
    function_response = function_to_call(**function_args)
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/memgpt/functions/function_sets/base.py", line 178, in archival_memory_search
    results, total = self.persistence_manager.archival_memory.search(query, count=count, start=page * count)
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/memgpt/memory.py", line 393, in search
    raise e
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/memgpt/memory.py", line 382, in search
    self.cache[query_string] = self.storage.query(query_string, query_vec, top_k=self.top_k)
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/memgpt/connectors/local.py", line 120, in query
    nodes = retriever.retrieve(query)
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/llama_index/core/base_retriever.py", line 54, in retrieve
    nodes = self._retrieve(query_bundle)
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/llama_index/indices/vector_store/retrievers/retriever.py", line 84, in _retrieve
    self._service_context.embed_model.get_agg_embedding_from_queries(
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/llama_index/embeddings/base.py", line 144, in get_agg_embedding_from_queries
    return agg_fn(query_embeddings)
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/llama_index/embeddings/base.py", line 33, in mean_agg
    return list(np.array(embeddings).mean(axis=0))
  File "/home/cpacker/memgpt-test-2/venv/lib/python3.9/site-packages/numpy/core/_methods.py", line 131, in _mean
    ret = ret / rcount
TypeError: unsupported operand type(s) for /: 'dict' and 'int'

⚡🔴 [function] Error: Error calling function archival_memory_search: unsupported operand type(s) for /: 'dict' and 'int'

With this PR, you won't get the error:

Have you tested this PR?

Yes, using memgpt quickstart.

[sarahwooders]

lgtm!