fix(python): huggingface TEI auth, procedural-memory content handling, and proxy pip auto-install (#6947)

mem0/embeddings/huggingface.py

@@ -1,4 +1,5 @@
 import logging
+import os
 from typing import Literal, Optional
 
 from openai import OpenAI
@@ -17,7 +18,8 @@ class HuggingFaceEmbedding(EmbeddingBase):
         super().__init__(config)
 
         if self.config.huggingface_base_url:
-            self.client = OpenAI(base_url=self.config.huggingface_base_url, api_key=self.config.api_key)
+            api_key = self.config.api_key or os.getenv("HUGGINGFACE_API_KEY") or "hf"
+            self.client = OpenAI(base_url=self.config.huggingface_base_url, api_key=api_key)
             self.config.model = self.config.model or "tei"
         else:
             self.config.model = self.config.model or "multi-qa-MiniLM-L6-cos-v1"

mem0/memory/main.py

@@ -2017,6 +2017,12 @@ class Memory(MemoryBase):
             logger.error(f"Error generating procedural memory summary: {e}")
             raise
 
+        if not procedural_memory:
+            raise ValueError(
+                "The LLM returned no content for the procedural memory summary. "
+                "The model may have declined the request or returned an empty response."
+            )
+
         if metadata is None:
             raise ValueError("Metadata cannot be done for procedural memory.")
 
@@ -3701,11 +3707,17 @@ class AsyncMemory(MemoryBase):
             else:
                 procedural_memory = await asyncio.to_thread(self.llm.generate_response, messages=parsed_messages)
                 procedural_memory = remove_code_blocks(procedural_memory)
-        
+
         except Exception as e:
             logger.error(f"Error generating procedural memory summary: {e}")
             raise
 
+        if not procedural_memory:
+            raise ValueError(
+                "The LLM returned no content for the procedural memory summary. "
+                "The model may have declined the request or returned an empty response."
+            )
+
         if metadata is None:
             raise ValueError("Metadata cannot be done for procedural memory.")
 

mem0/memory/utils.py

@@ -121,7 +121,15 @@ def remove_code_blocks(content: str) -> str:
     - If a code block is detected, it returns only the inner content, stripping out the markers.
     - If no code block markers are found, the original content is returned as-is.
     """
-    if content is None:
+    if isinstance(content, list):
+        parts = []
+        for block in content:
+            if isinstance(block, str):
+                parts.append(block)
+            elif isinstance(block, dict):
+                parts.append(block.get("text", ""))
+        content = "".join(parts)
+    if not isinstance(content, str):
         return ""
     pattern = r"^```[a-zA-Z0-9]*\n([\s\S]*?)\n```$"
     match = re.match(pattern, content.strip())

mem0/proxy/main.py

@@ -1,6 +1,4 @@
 import logging
-import subprocess
-import sys
 import threading
 from typing import List, Optional, Union
 
@@ -11,12 +9,7 @@ import mem0
 try:
     import litellm
 except ImportError:
-    try:
+    raise ImportError("The 'litellm' library is required. Please install it using 'pip install litellm'.")
-        subprocess.check_call([sys.executable, "-m", "pip", "install", "litellm"])
-        import litellm
-    except subprocess.CalledProcessError:
-        print("Failed to install 'litellm'. Please install it manually using 'pip install litellm'.")
-        sys.exit(1)
 
 from mem0 import Memory, MemoryClient
 from mem0.configs.prompts import MEMORY_ANSWER_PROMPT

tests/embeddings/test_huggingface_embeddings.py

@@ -72,7 +72,8 @@ def test_embed_with_custom_embedding_dims(mock_sentence_transformer):
     assert result == [1.0, 1.1, 1.2]
 
 
-def test_embed_with_huggingface_base_url():
+def test_embed_with_huggingface_base_url(monkeypatch):
+    monkeypatch.delenv("HUGGINGFACE_API_KEY", raising=False)
     config = BaseEmbedderConfig(
         huggingface_base_url="http://localhost:8080",
         model="my-custom-model",
@@ -81,20 +82,20 @@ def test_embed_with_huggingface_base_url():
     with patch("mem0.embeddings.huggingface.OpenAI") as mock_openai:
         mock_client = Mock()
         mock_openai.return_value = mock_client
-        
+
         # Create a mock for the response object and its attributes
         mock_embedding_response = Mock()
         mock_embedding_response.embedding = [0.1, 0.2, 0.3]
-        
+
         mock_create_response = Mock()
         mock_create_response.data = [mock_embedding_response]
-        
+
         mock_client.embeddings.create.return_value = mock_create_response
 
         embedder = HuggingFaceEmbedding(config)
         result = embedder.embed("Hello from custom endpoint")
 
-        mock_openai.assert_called_once_with(base_url="http://localhost:8080", api_key=None)
+        mock_openai.assert_called_once_with(base_url="http://localhost:8080", api_key="hf")
         mock_client.embeddings.create.assert_called_once_with(
             input="Hello from custom endpoint",
             model="my-custom-model",
@@ -115,6 +116,26 @@ def test_embed_with_huggingface_base_url_forwards_api_key():
         mock_openai.assert_called_once_with(base_url="http://localhost:8080", api_key="tei-token")
 
 
+def test_embed_with_huggingface_base_url_falls_back_to_env_var(monkeypatch):
+    monkeypatch.setenv("HUGGINGFACE_API_KEY", "env-token")
+    config = BaseEmbedderConfig(huggingface_base_url="http://localhost:8080", model="my-custom-model")
+    with patch("mem0.embeddings.huggingface.OpenAI") as mock_openai:
+        HuggingFaceEmbedding(config)
+
+        mock_openai.assert_called_once_with(base_url="http://localhost:8080", api_key="env-token")
+
+
+def test_embed_with_huggingface_base_url_config_key_beats_env_var(monkeypatch):
+    monkeypatch.setenv("HUGGINGFACE_API_KEY", "env-token")
+    config = BaseEmbedderConfig(
+        huggingface_base_url="http://localhost:8080", model="my-custom-model", api_key="config-token"
+    )
+    with patch("mem0.embeddings.huggingface.OpenAI") as mock_openai:
+        HuggingFaceEmbedding(config)
+
+        mock_openai.assert_called_once_with(base_url="http://localhost:8080", api_key="config-token")
+
+
 def test_embed_batch_sentence_transformer(mock_sentence_transformer):
     config = BaseEmbedderConfig()
     embedder = HuggingFaceEmbedding(config)

tests/embeddings/test_langchain_embeddings.py

@@ -0,0 +1,61 @@
+import pytest
+
+pytest.importorskip("langchain", reason="langchain is an optional extra")
+
+from langchain.embeddings.base import Embeddings  # noqa: E402
+
+from mem0.configs.embeddings.base import BaseEmbedderConfig  # noqa: E402
+from mem0.embeddings.langchain import LangchainEmbedding  # noqa: E402
+
+
+class DummyEmbeddings(Embeddings):
+    def __init__(self):
+        self.queries = []
+
+    def embed_documents(self, texts):
+        return [[0.1, 0.2, 0.3] for _ in texts]
+
+    def embed_query(self, text):
+        self.queries.append(text)
+        return [0.1, 0.2, 0.3]
+
+
+def test_missing_model_raises():
+    with pytest.raises(ValueError, match="`model` parameter is required"):
+        LangchainEmbedding(BaseEmbedderConfig())
+
+
+def test_model_that_is_not_an_embeddings_instance_raises():
+    with pytest.raises(ValueError, match="`model` must be an instance of Embeddings"):
+        LangchainEmbedding(BaseEmbedderConfig(model="text-embedding-3-small"))
+
+
+def test_configured_model_instance_is_kept():
+    model = DummyEmbeddings()
+
+    assert LangchainEmbedding(BaseEmbedderConfig(model=model)).langchain_model is model
+
+
+def test_embed_delegates_to_embed_query():
+    model = DummyEmbeddings()
+    embedder = LangchainEmbedding(BaseEmbedderConfig(model=model))
+
+    assert embedder.embed("hello", "add") == [0.1, 0.2, 0.3]
+    assert model.queries == ["hello"]
+
+
+def test_memory_action_never_reaches_the_model():
+    model = DummyEmbeddings()
+    embedder = LangchainEmbedding(BaseEmbedderConfig(model=model))
+
+    embedder.embed("hello", memory_action="add")
+
+    assert model.queries == ["hello"]
+
+
+def test_embed_batch_falls_back_to_sequential_embed():
+    model = DummyEmbeddings()
+    embedder = LangchainEmbedding(BaseEmbedderConfig(model=model))
+
+    assert embedder.embed_batch(["a", "b"]) == [[0.1, 0.2, 0.3], [0.1, 0.2, 0.3]]
+    assert model.queries == ["a", "b"]

tests/embeddings/test_ollama_embeddings.py

@@ -1,3 +1,6 @@
+import builtins
+import importlib
+import sys
 from unittest.mock import Mock, patch
 
 import pytest
@@ -94,3 +97,25 @@ def test_embed_batch_count_mismatch_raises(mock_ollama_client):
 
     with pytest.raises(ValueError, match="returned 1 embeddings for 2 texts"):
         embedder.embed_batch(["first text", "second text"])
+
+
+def test_missing_ollama_raises_actionable_import_error(monkeypatch):
+    """Missing ollama raises a catchable ImportError naming the install command, never input()/sys.exit()."""
+    monkeypatch.delitem(sys.modules, "mem0.embeddings.ollama", raising=False)
+    monkeypatch.delitem(sys.modules, "ollama", raising=False)
+
+    real_import = builtins.__import__
+
+    def fake_import(name, *args, **kwargs):
+        if name == "ollama":
+            raise ImportError("No module named 'ollama'")
+        return real_import(name, *args, **kwargs)
+
+    monkeypatch.setattr(builtins, "__import__", fake_import)
+
+    with patch("builtins.input") as mock_input, patch("sys.exit") as mock_exit:
+        with pytest.raises(ImportError, match="pip install ollama"):
+            importlib.import_module("mem0.embeddings.ollama")
+
+    mock_input.assert_not_called()
+    mock_exit.assert_not_called()

tests/memory/test_memory_utils.py

@@ -192,3 +192,16 @@ class TestProcessTelemetryFilters:
 class TestRemoveCodeBlocks:
     def test_none_content_returns_empty_string(self):
         assert remove_code_blocks(None) == ""
+
+    def test_list_content_from_langchain_aimessage_is_joined(self):
+        assert remove_code_blocks([{"type": "text", "text": "hello"}]) == "hello"
+
+    def test_list_content_joins_multiple_blocks_and_bare_strings(self):
+        content = [{"type": "text", "text": "hello "}, "world", {"type": "thinking", "thinking": "ignored"}]
+        assert remove_code_blocks(content) == "hello world"
+
+    def test_list_content_still_strips_code_fences(self):
+        assert remove_code_blocks([{"type": "text", "text": "```json\n{}\n```"}]) == "{}"
+
+    def test_unsupported_content_type_returns_empty_string(self):
+        assert remove_code_blocks(42) == ""

tests/test_memory.py

@@ -1235,9 +1235,10 @@ class TestHybridSearchWarning:
     def test_warning_for_store_without_keyword_search(
         self, mock_vs_factory, mock_emb, mock_llm, mock_sqlite, _cap, caplog
     ):
-        from mem0.vector_stores.base import VectorStoreBase
         import logging
 
+        from mem0.vector_stores.base import VectorStoreBase
+
         class StoreWithoutKeywordSearch(VectorStoreBase):
             def create_col(self, *a, **kw): pass
             def insert(self, *a, **kw): pass
@@ -1271,9 +1272,10 @@ class TestHybridSearchWarning:
     def test_no_warning_for_store_with_keyword_search(
         self, mock_vs_factory, mock_emb, mock_llm, mock_sqlite, _cap, caplog
     ):
-        from mem0.vector_stores.base import VectorStoreBase
         import logging
 
+        from mem0.vector_stores.base import VectorStoreBase
+
         class StoreWithKeywordSearch(VectorStoreBase):
             def keyword_search(self, query, top_k=5, filters=None):
                 return []
@@ -1574,3 +1576,92 @@ async def test_async_procedural_memory_default_path_without_langchain(mock_llm_f
 
     assert result["results"][0]["event"] == "ADD"
     memory.llm.generate_response.assert_called_once()
+
+
+@pytest.mark.asyncio
+@patch("mem0.memory.main.VectorStoreFactory")
+@patch("mem0.memory.main.EmbedderFactory")
+@patch("mem0.memory.main.LlmFactory")
+async def test_async_procedural_memory_langchain_list_content(mock_llm_factory, mock_emb, mock_vs):
+    """Regression #6150: a LangChain block list must be extracted, not dropped."""
+    mock_vs.return_value = MagicMock()
+    mock_emb.return_value = MagicMock()
+    mock_emb.return_value.embed.return_value = [0.1] * 1536
+    mock_llm_factory.return_value = MagicMock()
+
+    from mem0.memory.main import AsyncMemory
+
+    memory = AsyncMemory(MemoryConfig())
+    memory.vector_store = MagicMock()
+    memory.vector_store.insert = MagicMock()
+
+    mock_langchain_llm = MagicMock()
+    mock_response = MagicMock()
+    mock_response.content = [{"type": "text", "text": "- deploy with the release script"}]
+    mock_langchain_llm.invoke.return_value = mock_response
+
+    await memory._create_procedural_memory(
+        [{"role": "user", "content": "how do we deploy"}],
+        metadata={"user_id": "test_user"},
+        llm=mock_langchain_llm,
+    )
+
+    stored_data = memory.vector_store.insert.call_args[1]["payloads"][0]["data"]
+    assert stored_data == "- deploy with the release script"
+
+
+@pytest.mark.asyncio
+@patch("mem0.memory.main.VectorStoreFactory")
+@patch("mem0.memory.main.EmbedderFactory")
+@patch("mem0.memory.main.LlmFactory")
+async def test_async_procedural_memory_empty_content_raises(mock_llm_factory, mock_emb, mock_vs):
+    """Regression #6150: a refusal (content=None) must raise, not store an empty memory."""
+    mock_vs.return_value = MagicMock()
+    mock_emb.return_value = MagicMock()
+    mock_llm_factory.return_value = MagicMock()
+
+    from mem0.memory.main import AsyncMemory
+
+    memory = AsyncMemory(MemoryConfig())
+    memory.vector_store = MagicMock()
+    memory.vector_store.insert = MagicMock()
+
+    mock_langchain_llm = MagicMock()
+    mock_response = MagicMock()
+    mock_response.content = None
+    mock_langchain_llm.invoke.return_value = mock_response
+
+    with pytest.raises(ValueError, match="no content for the procedural memory summary"):
+        await memory._create_procedural_memory(
+            [{"role": "user", "content": "how do we deploy"}],
+            metadata={"user_id": "test_user"},
+            llm=mock_langchain_llm,
+        )
+
+    memory.vector_store.insert.assert_not_called()
+
+
+@patch("mem0.utils.factory.EmbedderFactory.create")
+@patch("mem0.utils.factory.VectorStoreFactory.create")
+@patch("mem0.utils.factory.LlmFactory.create")
+@patch("mem0.memory.storage.SQLiteManager")
+def test_sync_procedural_memory_empty_content_raises(
+    mock_sqlite, mock_llm_factory, mock_vector_factory, mock_embedder_factory
+):
+    """Regression #6150: the sync path must raise on empty LLM content too."""
+    mock_embedder_factory.return_value = MagicMock()
+    mock_vector_store = MagicMock()
+    mock_vector_factory.return_value = mock_vector_store
+    mock_llm_factory.return_value = MagicMock()
+    mock_sqlite.return_value = MagicMock()
+
+    memory = Memory(MemoryConfig())
+    memory.llm.generate_response = Mock(return_value=None)
+
+    with pytest.raises(ValueError, match="no content for the procedural memory summary"):
+        memory._create_procedural_memory(
+            [{"role": "user", "content": "how do we deploy"}],
+            metadata={"user_id": "test_user"},
+        )
+
+    mock_vector_store.insert.assert_not_called()

tests/test_proxy.py

@@ -1,4 +1,7 @@
+import builtins
+import importlib
 import inspect
+import sys
 from unittest.mock import Mock, patch
 
 import pytest
@@ -138,3 +141,20 @@ def test_completions_create_messages_default_does_not_leak_between_calls(mock_me
         f"Completions.create(messages=...) must default to None to avoid the "
         f"B006 shared-default-list bug; got {messages_default!r}."
     )
+
+
+def test_missing_litellm_raises_actionable_import_error(monkeypatch):
+    monkeypatch.delitem(sys.modules, "mem0.proxy.main", raising=False)
+    monkeypatch.delitem(sys.modules, "litellm", raising=False)
+
+    real_import = builtins.__import__
+
+    def fake_import(name, *args, **kwargs):
+        if name == "litellm":
+            raise ImportError("No module named 'litellm'")
+        return real_import(name, *args, **kwargs)
+
+    monkeypatch.setattr(builtins, "__import__", fake_import)
+
+    with pytest.raises(ImportError, match="pip install litellm"):
+        importlib.import_module("mem0.proxy.main")