Revert "fix: truncate oversized text in embedding requests" (#9227)

Revert "fix: truncate oversized text in embedding requests (#9196)"

This reverts commit a9c342087e022519c63d62fb76b72aed8859539b.

tests/test_embeddings.py

@@ -203,99 +203,3 @@ async def test_openai_embedding_minimum_chunk_failure(default_user):
 
         with pytest.raises(Exception, match="API error"):
             await client.request_embeddings(test_inputs, embedding_config)
-
-
-def test_split_text_in_half():
-    """Test the _split_text_in_half helper function."""
-    from letta.helpers.tpuf_client import _split_text_in_half
-
-    # Test with text that has sentence boundaries
-    long_text = "This is a test sentence. " * 100
-    splits = _split_text_in_half(long_text)
-    assert len(splits) == 2
-    assert len(splits[0]) > 0
-    assert len(splits[1]) > 0
-    # Should split at a sentence boundary
-    assert splits[0].endswith(".")
-
-    # Test with text that has no good break points
-    no_breaks = "a" * 1000
-    splits = _split_text_in_half(no_breaks)
-    assert len(splits) == 2
-    assert len(splits[0]) + len(splits[1]) == 1000
-
-    # Test with empty text
-    splits = _split_text_in_half("")
-    assert splits == []
-
-    # Test with short text (still splits)
-    short_text = "hello world"
-    splits = _split_text_in_half(short_text)
-    assert len(splits) == 2
-
-
-def test_chunked_message_query_deduplication():
-    """Test that chunked messages are deduplicated by message_id in query results."""
-    from unittest.mock import MagicMock
-
-    from letta.helpers.tpuf_client import TurbopufferClient
-
-    # Create a mock result with multiple chunks from the same message
-    mock_result = MagicMock()
-
-    # Simulate 3 rows: 2 chunks from message-1, 1 chunk from message-2
-    # The chunks are ranked by relevance (row order = rank order)
-    row1 = MagicMock()
-    row1.id = "message-1_chunk_1"  # Second chunk of message-1, but ranked first
-    row1.message_id = "message-1"
-    row1.chunk_index = 1
-    row1.text = "chunk 1 text"
-    row1.organization_id = "org-1"
-    row1.agent_id = "agent-1"
-    row1.role = "user"
-    row1.created_at = None
-    row1.conversation_id = None
-
-    row2 = MagicMock()
-    row2.id = "message-2"
-    row2.message_id = "message-2"
-    row2.chunk_index = 0
-    row2.text = "message 2 text"
-    row2.organization_id = "org-1"
-    row2.agent_id = "agent-1"
-    row2.role = "assistant"
-    row2.created_at = None
-    row2.conversation_id = None
-
-    row3 = MagicMock()
-    row3.id = "message-1"  # First chunk of message-1, but ranked third
-    row3.message_id = "message-1"
-    row3.chunk_index = 0
-    row3.text = "chunk 0 text"
-    row3.organization_id = "org-1"
-    row3.agent_id = "agent-1"
-    row3.role = "user"
-    row3.created_at = None
-    row3.conversation_id = None
-
-    mock_result.rows = [row1, row2, row3]
-
-    # Process results with deduplication
-    client = TurbopufferClient.__new__(TurbopufferClient)  # Create without __init__
-    results = client._process_message_query_results(mock_result, deduplicate=True)
-
-    # Should have 2 messages (message-1 deduplicated)
-    assert len(results) == 2
-
-    # First result should be message-1 (from the best-ranked chunk)
-    assert results[0]["id"] == "message-1"
-    assert results[0]["text"] == "chunk 1 text"  # Text from best-ranked chunk
-    assert results[0]["chunk_index"] == 1
-
-    # Second result should be message-2
-    assert results[1]["id"] == "message-2"
-    assert results[1]["text"] == "message 2 text"
-
-    # Test without deduplication
-    results_no_dedup = client._process_message_query_results(mock_result, deduplicate=False)
-    assert len(results_no_dedup) == 3  # All rows returned