Merge branch 'main' into fix/together-base-url-py

tests/AGENTS.md

@@ -0,0 +1,24 @@
+# Python SDK tests (`tests/`)
+
+pytest suite for the `mem0/` package.
+
+## Commands
+
+```bash
+make install_all    # optional deps; several tests need them
+make test           # pytest tests/
+make test-py-3.9    # pin a Python version (3.9 through 3.12)
+
+pytest tests/llms/test_openai.py::test_generate_response   # single test
+```
+
+## Conventions
+
+- Files are named `test_<module>.py`.
+- Provider tests mirror the source tree: `tests/<category>/<provider_name>/`.
+- pytest-mock for mocks, pytest-asyncio for the async surface.
+- Ruff line length **120**, matching `mem0/`. See [`../mem0/AGENTS.md`](../mem0/AGENTS.md).
+- Mock the provider SDK, never the code under test. A test that asserts the implementation back at itself is worse than no test.
+- Bug fixes need a regression test that fails without the fix. Write it first and watch it fail.
+
+Tests for other packages live with those packages: `mem0-ts/` (jest), `cli/python/tests/` (pytest), `cli/node/` (vitest), and each directory under `integrations/`.

tests/CLAUDE.md

@@ -0,0 +1 @@
+AGENTS.md

tests/llms/test_aws_bedrock.py

@@ -6,7 +6,6 @@ from mem0.configs.llms.aws_bedrock import AWSBedrockConfig
 from mem0.llms.aws_bedrock import AWSBedrockLLM, extract_provider
 from mem0.utils.factory import LlmFactory
 
-
 # ---------------------------------------------------------------------------
 # Helpers
 # ---------------------------------------------------------------------------
@@ -70,6 +69,23 @@ class TestExtractProvider:
         with pytest.raises(ValueError, match="Unknown provider"):
             extract_provider("unknown-vendor.some-model-v1:0")
 
+    def test_application_inference_profile_arn_without_override_raises(self):
+        arn = "arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/abc123xyz"
+        with pytest.raises(ValueError, match="Unknown provider"):
+            extract_provider(arn)
+
+    def test_application_inference_profile_arn_with_explicit_provider(self):
+        arn = "arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/abc123xyz"
+        assert extract_provider(arn, "anthropic") == "anthropic"
+
+    def test_explicit_provider_takes_precedence_over_regex(self):
+        assert extract_provider("anthropic.claude-3-5-sonnet-20240620-v1:0", "amazon") == "amazon"
+
+    def test_explicit_provider_typo_raises(self):
+        arn = "arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/abc123xyz"
+        with pytest.raises(ValueError, match="Unknown provider_override 'anthorpic'"):
+            extract_provider(arn, "anthorpic")
+
 
 # ---------------------------------------------------------------------------
 # AWSBedrockConfig
@@ -114,6 +130,44 @@ class TestAWSBedrockConfig:
         )
         assert config.aws_region == "us-east-2"
 
+    def test_provider_override_defaults_to_none(self):
+        config = AWSBedrockConfig(model="anthropic.claude-3-5-sonnet-20240620-v1:0")
+        assert config.provider_override is None
+
+    def test_provider_override_stored(self):
+        arn = "arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/abc123xyz"
+        config = AWSBedrockConfig(model=arn, provider_override="anthropic")
+        assert config.provider_override == "anthropic"
+
+
+# ---------------------------------------------------------------------------
+# AWSBedrockLLM with application inference profile ARNs
+# ---------------------------------------------------------------------------
+
+class TestApplicationInferenceProfileArn:
+    ARN = "arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/abc123xyz"
+
+    def test_arn_without_provider_override_raises(self, mock_boto3):
+        with pytest.raises(ValueError, match="Unknown provider"):
+            _make_llm(self.ARN, mock_boto3)
+
+    def test_arn_with_provider_override_resolves(self, mock_boto3):
+        llm = _make_llm(self.ARN, mock_boto3, provider_override="anthropic")
+        assert llm.provider == "anthropic"
+        assert llm.supports_tools is True
+
+    def test_plain_model_id_unaffected(self, mock_boto3):
+        llm = _make_llm("anthropic.claude-3-5-sonnet-20240620-v1:0", mock_boto3)
+        assert llm.provider == "anthropic"
+
+    def test_cross_region_inference_profile_unaffected(self, mock_boto3):
+        llm = _make_llm("us.anthropic.claude-haiku-4-5-20251001-v1:0", mock_boto3)
+        assert llm.provider == "anthropic"
+
+    def test_arn_with_misspelled_provider_override_raises(self, mock_boto3):
+        with pytest.raises(ValueError, match="Unknown provider_override 'anthorpic'"):
+            _make_llm(self.ARN, mock_boto3, provider_override="anthorpic")
+
 
 # ---------------------------------------------------------------------------
 # LlmFactory

tests/llms/test_vllm.py

@@ -4,6 +4,7 @@ import pytest
 
 from mem0 import AsyncMemory, Memory
 from mem0.configs.llms.base import BaseLlmConfig
+from mem0.configs.llms.vllm import VllmConfig
 from mem0.llms.vllm import VllmLLM
 
 
@@ -15,6 +16,59 @@ def mock_vllm_client():
         yield mock_client
 
 
+def test_base_url_explicit_arg_wins_over_env(mock_vllm_client, monkeypatch):
+    monkeypatch.setenv("VLLM_BASE_URL", "http://from-env:8000/v1")
+    config = VllmConfig(vllm_base_url="http://from-arg:8000/v1")
+    VllmLLM(config)
+
+    assert config.vllm_base_url == "http://from-arg:8000/v1"
+
+
+def test_base_url_env_wins_over_default(mock_vllm_client, monkeypatch):
+    monkeypatch.setenv("VLLM_BASE_URL", "http://from-env:8000/v1")
+    config = VllmConfig()
+    VllmLLM(config)
+
+    assert config.vllm_base_url == "http://from-env:8000/v1"
+
+
+def test_base_url_default_when_neither_set(mock_vllm_client, monkeypatch):
+    monkeypatch.delenv("VLLM_BASE_URL", raising=False)
+    config = VllmConfig()
+    VllmLLM(config)
+
+    assert config.vllm_base_url == "http://localhost:8000/v1"
+
+
+def test_openai_client_receives_explicit_base_url(monkeypatch):
+    monkeypatch.delenv("VLLM_BASE_URL", raising=False)
+    with patch("mem0.llms.vllm.OpenAI") as mock_openai:
+        VllmLLM(VllmConfig(vllm_base_url="http://from-arg:8000/v1"))
+
+    assert mock_openai.call_args.kwargs["base_url"] == "http://from-arg:8000/v1"
+
+
+def test_openai_client_receives_base_url_from_env(monkeypatch):
+    monkeypatch.setenv("VLLM_BASE_URL", "http://from-env:8000/v1")
+    with patch("mem0.llms.vllm.OpenAI") as mock_openai:
+        VllmLLM(VllmConfig())
+
+    base_url = mock_openai.call_args.kwargs["base_url"]
+    assert base_url == "http://from-env:8000/v1"
+    assert base_url != "http://localhost:8000/v1"
+
+
+def test_openai_client_receives_base_url_from_env_via_base_llm_config(monkeypatch):
+    """Original #6256 repro: BaseLlmConfig (not VllmConfig) plus VLLM_BASE_URL only."""
+    monkeypatch.setenv("VLLM_BASE_URL", "http://remote-vllm:8000/v1")
+    with patch("mem0.llms.vllm.OpenAI") as mock_openai:
+        VllmLLM(BaseLlmConfig(model="Qwen/Qwen2.5-32B-Instruct"))
+
+    base_url = mock_openai.call_args.kwargs["base_url"]
+    assert base_url == "http://remote-vllm:8000/v1"
+    assert base_url != "http://localhost:8000/v1"
+
+
 def test_generate_response_without_tools(mock_vllm_client):
     config = BaseLlmConfig(model="Qwen/Qwen2.5-32B-Instruct", temperature=0.7, max_tokens=100, top_p=1.0)
     llm = VllmLLM(config)

tests/test_client.py

@@ -8,6 +8,7 @@ import httpx
 import pytest
 import requests
 
+from mem0.client.main import AsyncMemoryClient
 from mem0.client.types import GetAllMemoryOptions, SearchMemoryOptions
 
 
@@ -181,6 +182,150 @@ class TestGetAllTypedOptionsParity:
         )
 
 
+class TestGetAllPageSizeQueryParams:
+    """MEM-6227: page and page_size must reach the API as query params independently."""
+
+    def test_get_all_page_size_alone_lands_in_query_params(self, mock_memory_client):
+        """get_all(page_size=...) without page should send page_size as a query param."""
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        mock_memory_client.client.post.return_value = mock_response
+
+        mock_memory_client.get_all(filters={"user_id": "u1"}, page_size=50)
+
+        mock_memory_client.client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+            params={"page_size": 50},
+        )
+
+    def test_get_all_page_alone_lands_in_query_params(self, mock_memory_client):
+        """get_all(page=...) without page_size should send page as a query param."""
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        mock_memory_client.client.post.return_value = mock_response
+
+        mock_memory_client.get_all(filters={"user_id": "u1"}, page=2)
+
+        mock_memory_client.client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+            params={"page": 2},
+        )
+
+    def test_get_all_page_and_page_size_together(self, mock_memory_client):
+        """get_all(page=..., page_size=...) should send both as query params."""
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        mock_memory_client.client.post.return_value = mock_response
+
+        mock_memory_client.get_all(filters={"user_id": "u1"}, page=2, page_size=50)
+
+        mock_memory_client.client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+            params={"page": 2, "page_size": 50},
+        )
+
+    def test_get_all_neither_page_nor_page_size_sends_no_query_params(self, mock_memory_client):
+        """get_all() without page or page_size should not pass a params kwarg."""
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        mock_memory_client.client.post.return_value = mock_response
+
+        mock_memory_client.get_all(filters={"user_id": "u1"})
+
+        mock_memory_client.client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+        )
+
+    def test_async_get_all_page_size_alone_lands_in_query_params(self):
+        asyncio.run(self._assert_async_get_all_page_size_alone())
+
+    async def _assert_async_get_all_page_size_alone(self):
+        client = AsyncMemoryClient.__new__(AsyncMemoryClient)
+        client.async_client = MagicMock()
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        client.async_client.post = AsyncMock(return_value=mock_response)
+
+        with patch("mem0.client.main.capture_client_event"):
+            await client.get_all(filters={"user_id": "u1"}, page_size=50)
+
+        client.async_client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+            params={"page_size": 50},
+        )
+
+    def test_async_get_all_page_alone_lands_in_query_params(self):
+        asyncio.run(self._assert_async_get_all_page_alone())
+
+    async def _assert_async_get_all_page_alone(self):
+        """get_all(page=0) without page_size should still send page=0 as a query param."""
+        client = AsyncMemoryClient.__new__(AsyncMemoryClient)
+        client.async_client = MagicMock()
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        client.async_client.post = AsyncMock(return_value=mock_response)
+
+        with patch("mem0.client.main.capture_client_event"):
+            await client.get_all(filters={"user_id": "u1"}, page=0)
+
+        client.async_client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+            params={"page": 0},
+        )
+
+    def test_async_get_all_page_and_page_size_together(self):
+        asyncio.run(self._assert_async_get_all_page_and_page_size_together())
+
+    async def _assert_async_get_all_page_and_page_size_together(self):
+        """get_all(page=0, page_size=0) should send both as query params despite being falsy."""
+        client = AsyncMemoryClient.__new__(AsyncMemoryClient)
+        client.async_client = MagicMock()
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        client.async_client.post = AsyncMock(return_value=mock_response)
+
+        with patch("mem0.client.main.capture_client_event"):
+            await client.get_all(filters={"user_id": "u1"}, page=0, page_size=0)
+
+        client.async_client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+            params={"page": 0, "page_size": 0},
+        )
+
+    def test_async_get_all_neither_page_nor_page_size_sends_no_query_params(self):
+        asyncio.run(self._assert_async_get_all_neither_page_nor_page_size())
+
+    async def _assert_async_get_all_neither_page_nor_page_size(self):
+        client = AsyncMemoryClient.__new__(AsyncMemoryClient)
+        client.async_client = MagicMock()
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"results": []}
+        mock_response.raise_for_status.return_value = None
+        client.async_client.post = AsyncMock(return_value=mock_response)
+
+        with patch("mem0.client.main.capture_client_event"):
+            await client.get_all(filters={"user_id": "u1"})
+
+        client.async_client.post.assert_called_once_with(
+            "/v3/memories/",
+            json={"filters": {"user_id": "u1"}},
+        )
+
+
 class TestUpdateExpirationDate:
     """Tests for update expiration_date payload handling."""
 

tests/vector_stores/test_e2e_threshold.py

@@ -249,9 +249,23 @@ REDIS_HOST = os.environ.get("REDIS_HOST", "localhost")
 REDIS_PORT = int(os.environ.get("REDIS_PORT", "6379"))
 
 
+def _redis_search_available(host, port, timeout=2):
+    """RediSearch (the FT.* command family) is an optional module, not part of vanilla Redis."""
+    if not _tcp_reachable(host, port, timeout=timeout):
+        return False
+    try:
+        import redis
+
+        client = redis.Redis(host=host, port=port, socket_connect_timeout=timeout, socket_timeout=timeout)
+        modules = {m["name"].lower() for m in client.module_list()}
+        return "search" in modules
+    except Exception:
+        return False
+
+
 @pytest.mark.skipif(
-    not _tcp_reachable(REDIS_HOST, REDIS_PORT),
-    reason=f"Redis not reachable at {REDIS_HOST}:{REDIS_PORT}",
+    not _redis_search_available(REDIS_HOST, REDIS_PORT),
+    reason=f"RediSearch module not available at {REDIS_HOST}:{REDIS_PORT} (requires Redis Stack, not vanilla Redis)",
 )
 class TestRedisThreshold:
     def test_threshold_filtering(self):

tests/vector_stores/test_score_normalization.py

@@ -239,9 +239,23 @@ REDIS_HOST = os.environ.get("REDIS_HOST", "localhost")
 REDIS_PORT = int(os.environ.get("REDIS_PORT", "6379"))
 
 
+def _redis_search_available(host, port, timeout=2):
+    """RediSearch (the FT.* command family) is an optional module, not part of vanilla Redis."""
+    if not _tcp_reachable(host, port, timeout=timeout):
+        return False
+    try:
+        import redis
+
+        client = redis.Redis(host=host, port=port, socket_connect_timeout=timeout, socket_timeout=timeout)
+        modules = {m["name"].lower() for m in client.module_list()}
+        return "search" in modules
+    except Exception:
+        return False
+
+
 @pytest.mark.skipif(
-    not _tcp_reachable(REDIS_HOST, REDIS_PORT),
-    reason=f"Redis not reachable at {REDIS_HOST}:{REDIS_PORT}",
+    not _redis_search_available(REDIS_HOST, REDIS_PORT),
+    reason=f"RediSearch module not available at {REDIS_HOST}:{REDIS_PORT} (requires Redis Stack, not vanilla Redis)",
 )
 class TestRedis:
     """Redis returns cosine distance [0,2]. Conversion: score = max(0, 1 - dist)."""