fix(ts-oss): isolate entity store from memory store by default (#4829)

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: kartik-mem0 <kartik.labhshetwar@mem0.ai>

mem0-ts/src/client/tests/integration/batch.test.ts

@@ -26,7 +26,7 @@ describeIntegration("MemoryClient Integration — Batch Operations", () => {
 
   beforeAll(async () => {
     cleanup = suppressTelemetryNoise();
-    client = createTestClient();
+    client = await createTestClient();
     memoryIds = await seedTestMemories(client, TEST_USER_ID);
   });
 

mem0-ts/src/client/tests/integration/crud.test.ts

@@ -26,9 +26,9 @@ describeIntegration("MemoryClient Integration — CRUD", () => {
   let cleanup: () => void;
   let memoryIds: string[] = [];
 
-  beforeAll(() => {
+  beforeAll(async () => {
     cleanup = suppressTelemetryNoise();
-    client = createTestClient();
+    client = await createTestClient();
   });
 
   afterAll(async () => {

mem0-ts/src/client/tests/integration/helpers.ts

@@ -18,9 +18,13 @@ export const describeIntegration = API_KEY ? describe : describe.skip;
  * Create a MemoryClient with the real API key.
  * Call this inside beforeAll — not at module scope — so it only
  * runs when the suite is not skipped.
+ *
+ * Returns an initialized client ready for immediate use.
  */
-export function createTestClient(): MemoryClient {
-  return new MemoryClient({ apiKey: API_KEY! });
+export async function createTestClient(): Promise<MemoryClient> {
+  const client = new MemoryClient({ apiKey: API_KEY! });
+  await client.ping();
+  return client;
 }
 
 /**

mem0-ts/src/client/tests/integration/initialization.test.ts

@@ -24,9 +24,9 @@ describeIntegration("MemoryClient Integration — Initialization", () => {
   let client: MemoryClient;
   let cleanup: () => void;
 
-  beforeAll(() => {
+  beforeAll(async () => {
     cleanup = suppressTelemetryNoise();
-    client = createTestClient();
+    client = await createTestClient();
   });
 
   afterAll(() => cleanup());

mem0-ts/src/client/tests/integration/management.test.ts

@@ -27,7 +27,7 @@ describeIntegration("MemoryClient Integration — Users & Project", () => {
 
   beforeAll(async () => {
     cleanup = suppressTelemetryNoise();
-    client = createTestClient();
+    client = await createTestClient();
     await seedTestMemories(client, TEST_USER_ID);
   });
 

mem0-ts/src/client/tests/integration/search.test.ts

@@ -27,7 +27,7 @@ describeIntegration("MemoryClient Integration — Search & History", () => {
 
   beforeAll(async () => {
     cleanup = suppressTelemetryNoise();
-    client = createTestClient();
+    client = await createTestClient();
     memoryIds = await seedTestMemories(client, TEST_USER_ID);
   });
 

mem0-ts/src/oss/src/config/defaults.ts

@@ -22,7 +22,7 @@ export const DEFAULT_MEMORY_CONFIG: MemoryConfig = {
     config: {
       baseURL: "https://api.openai.com/v1",
       apiKey: process.env.OPENAI_API_KEY || "",
-      model: "gpt-4.1-nano-2025-04-14",
+      model: "gpt-5-mini",
       modelProperties: undefined,
     },
   },

mem0-ts/src/oss/src/llms/azure.ts

@@ -18,7 +18,7 @@ export class AzureOpenAILLM implements LLM {
       endpoint: endpoint as string,
       ...rest,
     });
-    this.model = config.model || "gpt-4";
+    this.model = config.model || "gpt-5-mini";
   }
 
   async generateResponse(

mem0-ts/src/oss/src/llms/openai.ts

@@ -11,7 +11,7 @@ export class OpenAILLM implements LLM {
       apiKey: config.apiKey,
       baseURL: config.baseURL,
     });
-    this.model = config.model || "gpt-4.1-nano-2025-04-14";
+    this.model = config.model || "gpt-5-mini";
   }
 
   async generateResponse(

mem0-ts/src/oss/src/llms/openai_structured.ts

@@ -8,7 +8,7 @@ export class OpenAIStructuredLLM implements LLM {
 
   constructor(config: LLMConfig) {
     this.openai = new OpenAI({ apiKey: config.apiKey });
-    this.model = config.model || "gpt-4-turbo-preview";
+    this.model = config.model || "gpt-5-mini";
   }
 
   async generateResponse(

mem0/llms/azure_openai.py

@@ -39,7 +39,7 @@ class AzureOpenAILLM(LLMBase):
 
         # Model name should match the custom deployment name chosen for it.
         if not self.config.model:
-            self.config.model = "gpt-4.1-nano-2025-04-14"
+            self.config.model = "gpt-5-mini"
 
         api_key = self.config.azure_kwargs.api_key or os.getenv("LLM_AZURE_OPENAI_API_KEY")
         azure_deployment = self.config.azure_kwargs.azure_deployment or os.getenv("LLM_AZURE_DEPLOYMENT")

mem0/llms/azure_openai_structured.py

@@ -18,7 +18,7 @@ class AzureOpenAIStructuredLLM(LLMBase):
 
         # Model name should match the custom deployment name chosen for it.
         if not self.config.model:
-            self.config.model = "gpt-4.1-nano-2025-04-14"
+            self.config.model = "gpt-5-mini"
 
         api_key = self.config.azure_kwargs.api_key or os.getenv("LLM_AZURE_OPENAI_API_KEY")
         azure_deployment = self.config.azure_kwargs.azure_deployment or os.getenv("LLM_AZURE_DEPLOYMENT")

mem0/llms/litellm.py

@@ -16,7 +16,7 @@ class LiteLLM(LLMBase):
         super().__init__(config)
 
         if not self.config.model:
-            self.config.model = "gpt-4.1-nano-2025-04-14"
+            self.config.model = "gpt-5-mini"
 
     def _parse_response(self, response, tools):
         """

mem0/llms/openai.py

@@ -36,7 +36,7 @@ class OpenAILLM(LLMBase):
         super().__init__(config)
 
         if not self.config.model:
-            self.config.model = "gpt-4.1-nano-2025-04-14"
+            self.config.model = "gpt-5-mini"
 
         if os.environ.get("OPENROUTER_API_KEY"):  # Use OpenRouter
             self.client = OpenAI(

mem0/llms/openai_structured.py

@@ -12,7 +12,7 @@ class OpenAIStructuredLLM(LLMBase):
         super().__init__(config)
 
         if not self.config.model:
-            self.config.model = "gpt-4o-2024-08-06"
+            self.config.model = "gpt-5-mini"
 
         api_key = self.config.api_key or os.getenv("OPENAI_API_KEY")
         base_url = self.config.openai_base_url or os.getenv("OPENAI_API_BASE") or "https://api.openai.com/v1"

mem0/memory/setup.py

@@ -32,15 +32,26 @@ def get_user_id():
         return "anonymous_user"
 
 
-def get_or_create_user_id(vector_store):
-    """Store user_id in vector store and return it."""
+def get_or_create_user_id(vector_store=None):
+    """Store user_id in vector store and return it.
+
+    If vector_store is None, simply returns the user_id from config.
+    This ensures telemetry initialization never fails due to missing vector store.
+    """
     user_id = get_user_id()
 
+    # If no vector store provided, just return the user_id
+    if vector_store is None:
+        return user_id
+
     # Try to get existing user_id from vector store
     try:
         existing = vector_store.get(vector_id=user_id)
         if existing and hasattr(existing, "payload") and existing.payload and "user_id" in existing.payload:
-            return existing.payload["user_id"]
+            stored_id = existing.payload["user_id"]
+            # Ensure we never return None from vector store
+            if stored_id is not None:
+                return stored_id
     except Exception:
         pass
 

mem0/memory/telemetry.py

@@ -88,6 +88,12 @@ class AnonymousTelemetry:
         if self.posthog is None:
             return
 
+        # Determine distinct_id, skip if None to prevent crashes
+        distinct_id = self.user_id if user_email is None else user_email
+        if distinct_id is None:
+            _logger.debug("Skipping telemetry event %r: no distinct_id available", event_name)
+            return
+
         if properties is None:
             properties = {}
         properties = {
@@ -101,8 +107,10 @@ class AnonymousTelemetry:
             "machine": platform.machine(),
             **properties,
         }
-        distinct_id = self.user_id if user_email is None else user_email
-        self.posthog.capture(distinct_id=distinct_id, event=event_name, properties=properties)
+        try:
+            self.posthog.capture(distinct_id=distinct_id, event=event_name, properties=properties)
+        except Exception as e:
+            _logger.debug("Failed to capture telemetry event %r: %s", event_name, e)
 
     def close(self):
         if self.posthog is not None:
@@ -158,36 +166,52 @@ atexit.register(client_telemetry.close)
 
 
 def capture_event(event_name, memory_instance, additional_data=None):
+    """Capture telemetry event for OSS Memory instances.
+
+    This function is designed to never raise exceptions - telemetry failures
+    should not affect the main application flow.
+    """
     if not MEM0_TELEMETRY:
         return
 
-    oss_telemetry = _get_oss_telemetry()
-    if oss_telemetry is None:
-        return
+    try:
+        oss_telemetry = _get_oss_telemetry()
+        if oss_telemetry is None:
+            return
 
-    event_data = {
-        "collection": memory_instance.collection_name,
-        "vector_size": memory_instance.embedding_model.config.embedding_dims,
-        "history_store": "sqlite",
-        "vector_store": f"{memory_instance.vector_store.__class__.__module__}.{memory_instance.vector_store.__class__.__name__}",
-        "llm": f"{memory_instance.llm.__class__.__module__}.{memory_instance.llm.__class__.__name__}",
-        "embedding_model": f"{memory_instance.embedding_model.__class__.__module__}.{memory_instance.embedding_model.__class__.__name__}",
-        "function": f"{memory_instance.__class__.__module__}.{memory_instance.__class__.__name__}.{memory_instance.api_version}",
-    }
-    if additional_data:
-        event_data.update(additional_data)
+        event_data = {
+            "collection": memory_instance.collection_name,
+            "vector_size": memory_instance.embedding_model.config.embedding_dims,
+            "history_store": "sqlite",
+            "vector_store": f"{memory_instance.vector_store.__class__.__module__}.{memory_instance.vector_store.__class__.__name__}",
+            "llm": f"{memory_instance.llm.__class__.__module__}.{memory_instance.llm.__class__.__name__}",
+            "embedding_model": f"{memory_instance.embedding_model.__class__.__module__}.{memory_instance.embedding_model.__class__.__name__}",
+            "function": f"{memory_instance.__class__.__module__}.{memory_instance.__class__.__name__}.{memory_instance.api_version}",
+        }
+        if additional_data:
+            event_data.update(additional_data)
 
-    oss_telemetry.capture_event(event_name, event_data)
+        oss_telemetry.capture_event(event_name, event_data)
+    except Exception as e:
+        _logger.debug("Failed to capture OSS telemetry event %r: %s", event_name, e)
 
 
 def capture_client_event(event_name, instance, additional_data=None):
+    """Capture telemetry event for hosted MemoryClient instances.
+
+    This function is designed to never raise exceptions - telemetry failures
+    should not affect the main application flow.
+    """
     if not MEM0_TELEMETRY:
         return
 
-    event_data = {
-        "function": f"{instance.__class__.__module__}.{instance.__class__.__name__}",
-    }
-    if additional_data:
-        event_data.update(additional_data)
+    try:
+        event_data = {
+            "function": f"{instance.__class__.__module__}.{instance.__class__.__name__}",
+        }
+        if additional_data:
+            event_data.update(additional_data)
 
-    client_telemetry.capture_event(event_name, event_data, instance.user_email)
+        client_telemetry.capture_event(event_name, event_data, instance.user_email)
+    except Exception as e:
+        _logger.debug("Failed to capture client telemetry event %r: %s", event_name, e)

tests/llms/test_azure_openai.py

@@ -286,8 +286,8 @@ def test_init_with_env_vars(monkeypatch):
             http_client=None,
             default_headers=None,
         )
-        # Should default to "gpt-4.1-nano-2025-04-14" if model is None
-        assert llm.config.model == "gpt-4.1-nano-2025-04-14"
+        # Should default to "gpt-5-mini" if model is None
+        assert llm.config.model == "gpt-5-mini"
 
 
 def test_init_with_default_azure_credential(monkeypatch):

tests/llms/test_azure_openai_structured.py

@@ -57,7 +57,7 @@ def test_init_with_default_credential(mock_credential, mock_token_provider, mock
     mock_token_provider.return_value = "token-provider"
     llm = AzureOpenAIStructuredLLM(config)
     # Should set default model if not provided
-    assert llm.config.model == "gpt-4.1-nano-2025-04-14"
+    assert llm.config.model == "gpt-5-mini"
     mock_credential.assert_called_once()
     mock_token_provider.assert_called_once_with(mock_credential.return_value, SCOPE)
     mock_azure_openai.assert_called_once()
@@ -92,7 +92,7 @@ def test_init_with_placeholder_api_key_uses_default_credential(
         config = DummyConfig(model=None, azure_kwargs=DummyAzureKwargs(api_key="your-api-key"))
         mock_token_provider.return_value = "token-provider"
         llm = AzureOpenAIStructuredLLM(config)
-        assert llm.config.model == "gpt-4.1-nano-2025-04-14"
+        assert llm.config.model == "gpt-5-mini"
         mock_credential.assert_called_once()
         mock_token_provider.assert_called_once_with(mock_credential.return_value, SCOPE)
         mock_azure_openai.assert_called_once()

tests/test_telemetry.py

@@ -359,3 +359,63 @@ class TestTelemetryEnvVar:
     def test_env_var_parsing(self, value, expected):
         result = value.lower() in ("true", "1", "yes")
         assert result == expected
+
+
+class TestTelemetryNullUserIdHandling:
+    """Verify telemetry doesn't crash when user_id is None.
+
+    This is a regression test for the bug where Memory.from_config() crashed
+    with AssertionError because distinct_id was None in PostHog.capture().
+    """
+
+    def test_capture_event_skips_when_user_id_is_none(self):
+        """AnonymousTelemetry.capture_event should not crash when user_id is None."""
+        with patch.object(telemetry_module, "MEM0_TELEMETRY", True):
+            with patch("mem0.memory.telemetry.Posthog") as mock_posthog_cls:
+                at = telemetry_module.AnonymousTelemetry()
+                at.user_id = None  # Simulate the bug condition
+
+                # This should not raise, even with user_id=None
+                at.capture_event("test.event", {"key": "value"})
+
+                # PostHog.capture should NOT have been called since user_id is None
+                mock_posthog_cls.return_value.capture.assert_not_called()
+
+    def test_capture_event_does_not_crash_on_posthog_error(self):
+        """AnonymousTelemetry.capture_event should catch PostHog exceptions."""
+        with patch.object(telemetry_module, "MEM0_TELEMETRY", True):
+            with patch("mem0.memory.telemetry.Posthog") as mock_posthog_cls:
+                with patch("mem0.memory.telemetry.get_or_create_user_id", return_value="test-user"):
+                    at = telemetry_module.AnonymousTelemetry()
+                    mock_posthog_cls.return_value.capture.side_effect = Exception("PostHog error")
+
+                    # This should not raise, even when PostHog.capture fails
+                    at.capture_event("test.event", {"key": "value"})
+
+    def test_oss_capture_event_does_not_crash_memory_init(self):
+        """capture_event() should never raise, even if everything inside fails."""
+        with patch.object(telemetry_module, "MEM0_TELEMETRY", True):
+            # Make _get_oss_telemetry return a broken telemetry object
+            mock_at = MagicMock()
+            mock_at.capture_event.side_effect = Exception("Telemetry is broken")
+
+            with patch.object(telemetry_module, "_oss_telemetry_instance", mock_at):
+                mock_memory = MagicMock()
+                mock_memory.config.graph_store.config = None
+                mock_memory.api_version = "v1"
+
+                # This should not raise, even when telemetry fails
+                telemetry_module.capture_event("mem0.init", mock_memory)
+
+    def test_client_capture_event_does_not_crash(self):
+        """capture_client_event() should never raise exceptions."""
+        with patch.object(telemetry_module, "MEM0_TELEMETRY", True):
+            mock_client_telemetry = MagicMock()
+            mock_client_telemetry.capture_event.side_effect = Exception("Client telemetry broken")
+
+            with patch.object(telemetry_module, "client_telemetry", mock_client_telemetry):
+                mock_instance = MagicMock()
+                mock_instance.user_email = "test@example.com"
+
+                # This should not raise
+                telemetry_module.capture_client_event("test.event", mock_instance)