fix: disable dangerous deserialization by default in FAISS component (#11999)

* fix: disable dangerous deserialization by default in FAISS component

Change the default value of allow_dangerous_deserialization from True
to False to prevent remote code execution via malicious pickle files.

This addresses a security vulnerability where an attacker could upload
a crafted pickle file and trigger arbitrary code execution when the
FAISS component loads the index.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* [autofix.ci] apply automated fixes

* [autofix.ci] apply automated fixes (attempt 2/3)

* fix: set allow_dangerous_deserialization to false in Nvidia Remix starter project and add regression test

- Changed allow_dangerous_deserialization default from true to false in
  Nvidia Remix.json starter project to match the FAISS component security fix
- Added regression tests to ensure the default value does not revert to True

* [autofix.ci] apply automated fixes

* [autofix.ci] apply automated fixes

* fix: skip FAISS test gracefully when langchain_community is not installed

* [autofix.ci] apply automated fixes

* [autofix.ci] apply automated fixes

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: Janardan Singh Kavia <janardankavia@ibm.com>

Author: 4 people

src/backend/base/langflow/initial_setup/starter_projects/Nvidia Remix.json

@@ -2112,7 +2112,7 @@
             "legacy": false,
             "lf_version": "1.4.2",
             "metadata": {
-              "code_hash": "2bd7a064d724",
+              "code_hash": "3e55e36d0692",
               "dependencies": {
                 "dependencies": [
                   {
@@ -2156,7 +2156,7 @@
                 "advanced": true,
                 "display_name": "Allow Dangerous Deserialization",
                 "dynamic": false,
-                "info": "Set to True to allow loading pickle files from untrusted sources. Only enable this if you trust the source of the data.",
+                "info": "Set to True to allow loading pickle files. WARNING: Only enable this if you trust the source of the data. Malicious pickle files can execute arbitrary code on your system.",
                 "list": false,
                 "list_add_label": "Add More",
                 "name": "allow_dangerous_deserialization",
@@ -2167,7 +2167,7 @@
                 "tool_mode": false,
                 "trace_as_metadata": true,
                 "type": "bool",
-                "value": true
+                "value": false
               },
               "code": {
                 "advanced": true,
@@ -2185,7 +2185,7 @@
                 "show": true,
                 "title_case": false,
                 "type": "code",
-                "value": "from pathlib import Path\n\nfrom langchain_community.vectorstores import FAISS\n\nfrom lfx.base.vectorstores.model import LCVectorStoreComponent, check_cached_vector_store\nfrom lfx.helpers.data import docs_to_data\nfrom lfx.io import BoolInput, HandleInput, IntInput, StrInput\nfrom lfx.schema.data import Data\n\n\nclass FaissVectorStoreComponent(LCVectorStoreComponent):\n    \"\"\"FAISS Vector Store with search capabilities.\"\"\"\n\n    display_name: str = \"FAISS\"\n    description: str = \"FAISS Vector Store with search capabilities\"\n    name = \"FAISS\"\n    icon = \"FAISS\"\n\n    inputs = [\n        StrInput(\n            name=\"index_name\",\n            display_name=\"Index Name\",\n            value=\"langflow_index\",\n        ),\n        StrInput(\n            name=\"persist_directory\",\n            display_name=\"Persist Directory\",\n            info=\"Path to save the FAISS index. It will be relative to where Langflow is running.\",\n        ),\n        *LCVectorStoreComponent.inputs,\n        BoolInput(\n            name=\"allow_dangerous_deserialization\",\n            display_name=\"Allow Dangerous Deserialization\",\n            info=\"Set to True to allow loading pickle files from untrusted sources. \"\n            \"Only enable this if you trust the source of the data.\",\n            advanced=True,\n            value=True,\n        ),\n        HandleInput(name=\"embedding\", display_name=\"Embedding\", input_types=[\"Embeddings\"]),\n        IntInput(\n            name=\"number_of_results\",\n            display_name=\"Number of Results\",\n            info=\"Number of results to return.\",\n            advanced=True,\n            value=4,\n        ),\n    ]\n\n    @staticmethod\n    def resolve_path(path: str) -> str:\n        \"\"\"Resolve the path relative to the Langflow root.\n\n        Args:\n            path: The path to resolve\n        Returns:\n            str: The resolved path as a string\n        \"\"\"\n        return str(Path(path).resolve())\n\n    def get_persist_directory(self) -> Path:\n        \"\"\"Returns the resolved persist directory path or the current directory if not set.\"\"\"\n        if self.persist_directory:\n            return Path(self.resolve_path(self.persist_directory))\n        return Path()\n\n    @check_cached_vector_store\n    def build_vector_store(self) -> FAISS:\n        \"\"\"Builds the FAISS object.\"\"\"\n        path = self.get_persist_directory()\n        path.mkdir(parents=True, exist_ok=True)\n\n        # Convert DataFrame to Data if needed using parent's method\n        self.ingest_data = self._prepare_ingest_data()\n\n        documents = []\n        for _input in self.ingest_data or []:\n            if isinstance(_input, Data):\n                documents.append(_input.to_lc_document())\n            else:\n                documents.append(_input)\n\n        faiss = FAISS.from_documents(documents=documents, embedding=self.embedding)\n        faiss.save_local(str(path), self.index_name)\n        return faiss\n\n    def search_documents(self) -> list[Data]:\n        \"\"\"Search for documents in the FAISS vector store.\"\"\"\n        path = self.get_persist_directory()\n        index_path = path / f\"{self.index_name}.faiss\"\n\n        if not index_path.exists():\n            vector_store = self.build_vector_store()\n        else:\n            vector_store = FAISS.load_local(\n                folder_path=str(path),\n                embeddings=self.embedding,\n                index_name=self.index_name,\n                allow_dangerous_deserialization=self.allow_dangerous_deserialization,\n            )\n\n        if not vector_store:\n            msg = \"Failed to load the FAISS index.\"\n            raise ValueError(msg)\n\n        if self.search_query and isinstance(self.search_query, str) and self.search_query.strip():\n            docs = vector_store.similarity_search(\n                query=self.search_query,\n                k=self.number_of_results,\n            )\n            return docs_to_data(docs)\n        return []\n"
+                "value": "from pathlib import Path\n\nfrom langchain_community.vectorstores import FAISS\n\nfrom lfx.base.vectorstores.model import LCVectorStoreComponent, check_cached_vector_store\nfrom lfx.helpers.data import docs_to_data\nfrom lfx.io import BoolInput, HandleInput, IntInput, StrInput\nfrom lfx.schema.data import Data\n\n\nclass FaissVectorStoreComponent(LCVectorStoreComponent):\n    \"\"\"FAISS Vector Store with search capabilities.\"\"\"\n\n    display_name: str = \"FAISS\"\n    description: str = \"FAISS Vector Store with search capabilities\"\n    name = \"FAISS\"\n    icon = \"FAISS\"\n\n    inputs = [\n        StrInput(\n            name=\"index_name\",\n            display_name=\"Index Name\",\n            value=\"langflow_index\",\n        ),\n        StrInput(\n            name=\"persist_directory\",\n            display_name=\"Persist Directory\",\n            info=\"Path to save the FAISS index. It will be relative to where Langflow is running.\",\n        ),\n        *LCVectorStoreComponent.inputs,\n        BoolInput(\n            name=\"allow_dangerous_deserialization\",\n            display_name=\"Allow Dangerous Deserialization\",\n            info=\"Set to True to allow loading pickle files. WARNING: Only enable this if you trust the source \"\n            \"of the data. Malicious pickle files can execute arbitrary code on your system.\",\n            advanced=True,\n            value=False,\n        ),\n        HandleInput(name=\"embedding\", display_name=\"Embedding\", input_types=[\"Embeddings\"]),\n        IntInput(\n            name=\"number_of_results\",\n            display_name=\"Number of Results\",\n            info=\"Number of results to return.\",\n            advanced=True,\n            value=4,\n        ),\n    ]\n\n    @staticmethod\n    def resolve_path(path: str) -> str:\n        \"\"\"Resolve the path relative to the Langflow root.\n\n        Args:\n            path: The path to resolve\n        Returns:\n            str: The resolved path as a string\n        \"\"\"\n        return str(Path(path).resolve())\n\n    def get_persist_directory(self) -> Path:\n        \"\"\"Returns the resolved persist directory path or the current directory if not set.\"\"\"\n        if self.persist_directory:\n            return Path(self.resolve_path(self.persist_directory))\n        return Path()\n\n    @check_cached_vector_store\n    def build_vector_store(self) -> FAISS:\n        \"\"\"Builds the FAISS object.\"\"\"\n        path = self.get_persist_directory()\n        path.mkdir(parents=True, exist_ok=True)\n\n        # Convert DataFrame to Data if needed using parent's method\n        self.ingest_data = self._prepare_ingest_data()\n\n        documents = []\n        for _input in self.ingest_data or []:\n            if isinstance(_input, Data):\n                documents.append(_input.to_lc_document())\n            else:\n                documents.append(_input)\n\n        faiss = FAISS.from_documents(documents=documents, embedding=self.embedding)\n        faiss.save_local(str(path), self.index_name)\n        return faiss\n\n    def search_documents(self) -> list[Data]:\n        \"\"\"Search for documents in the FAISS vector store.\"\"\"\n        path = self.get_persist_directory()\n        index_path = path / f\"{self.index_name}.faiss\"\n\n        if not index_path.exists():\n            vector_store = self.build_vector_store()\n        else:\n            vector_store = FAISS.load_local(\n                folder_path=str(path),\n                embeddings=self.embedding,\n                index_name=self.index_name,\n                allow_dangerous_deserialization=self.allow_dangerous_deserialization,\n            )\n\n        if not vector_store:\n            msg = \"Failed to load the FAISS index.\"\n            raise ValueError(msg)\n\n        if self.search_query and isinstance(self.search_query, str) and self.search_query.strip():\n            docs = vector_store.similarity_search(\n                query=self.search_query,\n                k=self.number_of_results,\n            )\n            return docs_to_data(docs)\n        return []\n"
               },
               "embedding": {
                 "_input_type": "HandleInput",

src/lfx/src/lfx/_assets/component_index.json

@@ -29,7 +29,7 @@
           "icon": "FAISS",
           "legacy": false,
           "metadata": {
-            "code_hash": "2bd7a064d724",
+            "code_hash": "3e55e36d0692",
             "dependencies": {
               "dependencies": [
                 {
@@ -85,7 +85,7 @@
               "advanced": true,
               "display_name": "Allow Dangerous Deserialization",
               "dynamic": false,
-              "info": "Set to True to allow loading pickle files from untrusted sources. Only enable this if you trust the source of the data.",
+              "info": "Set to True to allow loading pickle files. WARNING: Only enable this if you trust the source of the data. Malicious pickle files can execute arbitrary code on your system.",
               "list": false,
               "list_add_label": "Add More",
               "name": "allow_dangerous_deserialization",
@@ -98,7 +98,7 @@
               "trace_as_metadata": true,
               "track_in_telemetry": true,
               "type": "bool",
-              "value": true
+              "value": false
             },
             "code": {
               "advanced": true,
@@ -116,7 +116,7 @@
               "show": true,
               "title_case": false,
               "type": "code",
-              "value": "from pathlib import Path\n\nfrom langchain_community.vectorstores import FAISS\n\nfrom lfx.base.vectorstores.model import LCVectorStoreComponent, check_cached_vector_store\nfrom lfx.helpers.data import docs_to_data\nfrom lfx.io import BoolInput, HandleInput, IntInput, StrInput\nfrom lfx.schema.data import Data\n\n\nclass FaissVectorStoreComponent(LCVectorStoreComponent):\n    \"\"\"FAISS Vector Store with search capabilities.\"\"\"\n\n    display_name: str = \"FAISS\"\n    description: str = \"FAISS Vector Store with search capabilities\"\n    name = \"FAISS\"\n    icon = \"FAISS\"\n\n    inputs = [\n        StrInput(\n            name=\"index_name\",\n            display_name=\"Index Name\",\n            value=\"langflow_index\",\n        ),\n        StrInput(\n            name=\"persist_directory\",\n            display_name=\"Persist Directory\",\n            info=\"Path to save the FAISS index. It will be relative to where Langflow is running.\",\n        ),\n        *LCVectorStoreComponent.inputs,\n        BoolInput(\n            name=\"allow_dangerous_deserialization\",\n            display_name=\"Allow Dangerous Deserialization\",\n            info=\"Set to True to allow loading pickle files from untrusted sources. \"\n            \"Only enable this if you trust the source of the data.\",\n            advanced=True,\n            value=True,\n        ),\n        HandleInput(name=\"embedding\", display_name=\"Embedding\", input_types=[\"Embeddings\"]),\n        IntInput(\n            name=\"number_of_results\",\n            display_name=\"Number of Results\",\n            info=\"Number of results to return.\",\n            advanced=True,\n            value=4,\n        ),\n    ]\n\n    @staticmethod\n    def resolve_path(path: str) -> str:\n        \"\"\"Resolve the path relative to the Langflow root.\n\n        Args:\n            path: The path to resolve\n        Returns:\n            str: The resolved path as a string\n        \"\"\"\n        return str(Path(path).resolve())\n\n    def get_persist_directory(self) -> Path:\n        \"\"\"Returns the resolved persist directory path or the current directory if not set.\"\"\"\n        if self.persist_directory:\n            return Path(self.resolve_path(self.persist_directory))\n        return Path()\n\n    @check_cached_vector_store\n    def build_vector_store(self) -> FAISS:\n        \"\"\"Builds the FAISS object.\"\"\"\n        path = self.get_persist_directory()\n        path.mkdir(parents=True, exist_ok=True)\n\n        # Convert DataFrame to Data if needed using parent's method\n        self.ingest_data = self._prepare_ingest_data()\n\n        documents = []\n        for _input in self.ingest_data or []:\n            if isinstance(_input, Data):\n                documents.append(_input.to_lc_document())\n            else:\n                documents.append(_input)\n\n        faiss = FAISS.from_documents(documents=documents, embedding=self.embedding)\n        faiss.save_local(str(path), self.index_name)\n        return faiss\n\n    def search_documents(self) -> list[Data]:\n        \"\"\"Search for documents in the FAISS vector store.\"\"\"\n        path = self.get_persist_directory()\n        index_path = path / f\"{self.index_name}.faiss\"\n\n        if not index_path.exists():\n            vector_store = self.build_vector_store()\n        else:\n            vector_store = FAISS.load_local(\n                folder_path=str(path),\n                embeddings=self.embedding,\n                index_name=self.index_name,\n                allow_dangerous_deserialization=self.allow_dangerous_deserialization,\n            )\n\n        if not vector_store:\n            msg = \"Failed to load the FAISS index.\"\n            raise ValueError(msg)\n\n        if self.search_query and isinstance(self.search_query, str) and self.search_query.strip():\n            docs = vector_store.similarity_search(\n                query=self.search_query,\n                k=self.number_of_results,\n            )\n            return docs_to_data(docs)\n        return []\n"
+              "value": "from pathlib import Path\n\nfrom langchain_community.vectorstores import FAISS\n\nfrom lfx.base.vectorstores.model import LCVectorStoreComponent, check_cached_vector_store\nfrom lfx.helpers.data import docs_to_data\nfrom lfx.io import BoolInput, HandleInput, IntInput, StrInput\nfrom lfx.schema.data import Data\n\n\nclass FaissVectorStoreComponent(LCVectorStoreComponent):\n    \"\"\"FAISS Vector Store with search capabilities.\"\"\"\n\n    display_name: str = \"FAISS\"\n    description: str = \"FAISS Vector Store with search capabilities\"\n    name = \"FAISS\"\n    icon = \"FAISS\"\n\n    inputs = [\n        StrInput(\n            name=\"index_name\",\n            display_name=\"Index Name\",\n            value=\"langflow_index\",\n        ),\n        StrInput(\n            name=\"persist_directory\",\n            display_name=\"Persist Directory\",\n            info=\"Path to save the FAISS index. It will be relative to where Langflow is running.\",\n        ),\n        *LCVectorStoreComponent.inputs,\n        BoolInput(\n            name=\"allow_dangerous_deserialization\",\n            display_name=\"Allow Dangerous Deserialization\",\n            info=\"Set to True to allow loading pickle files. WARNING: Only enable this if you trust the source \"\n            \"of the data. Malicious pickle files can execute arbitrary code on your system.\",\n            advanced=True,\n            value=False,\n        ),\n        HandleInput(name=\"embedding\", display_name=\"Embedding\", input_types=[\"Embeddings\"]),\n        IntInput(\n            name=\"number_of_results\",\n            display_name=\"Number of Results\",\n            info=\"Number of results to return.\",\n            advanced=True,\n            value=4,\n        ),\n    ]\n\n    @staticmethod\n    def resolve_path(path: str) -> str:\n        \"\"\"Resolve the path relative to the Langflow root.\n\n        Args:\n            path: The path to resolve\n        Returns:\n            str: The resolved path as a string\n        \"\"\"\n        return str(Path(path).resolve())\n\n    def get_persist_directory(self) -> Path:\n        \"\"\"Returns the resolved persist directory path or the current directory if not set.\"\"\"\n        if self.persist_directory:\n            return Path(self.resolve_path(self.persist_directory))\n        return Path()\n\n    @check_cached_vector_store\n    def build_vector_store(self) -> FAISS:\n        \"\"\"Builds the FAISS object.\"\"\"\n        path = self.get_persist_directory()\n        path.mkdir(parents=True, exist_ok=True)\n\n        # Convert DataFrame to Data if needed using parent's method\n        self.ingest_data = self._prepare_ingest_data()\n\n        documents = []\n        for _input in self.ingest_data or []:\n            if isinstance(_input, Data):\n                documents.append(_input.to_lc_document())\n            else:\n                documents.append(_input)\n\n        faiss = FAISS.from_documents(documents=documents, embedding=self.embedding)\n        faiss.save_local(str(path), self.index_name)\n        return faiss\n\n    def search_documents(self) -> list[Data]:\n        \"\"\"Search for documents in the FAISS vector store.\"\"\"\n        path = self.get_persist_directory()\n        index_path = path / f\"{self.index_name}.faiss\"\n\n        if not index_path.exists():\n            vector_store = self.build_vector_store()\n        else:\n            vector_store = FAISS.load_local(\n                folder_path=str(path),\n                embeddings=self.embedding,\n                index_name=self.index_name,\n                allow_dangerous_deserialization=self.allow_dangerous_deserialization,\n            )\n\n        if not vector_store:\n            msg = \"Failed to load the FAISS index.\"\n            raise ValueError(msg)\n\n        if self.search_query and isinstance(self.search_query, str) and self.search_query.strip():\n            docs = vector_store.similarity_search(\n                query=self.search_query,\n                k=self.number_of_results,\n            )\n            return docs_to_data(docs)\n        return []\n"
             },
             "embedding": {
               "_input_type": "HandleInput",

src/lfx/src/lfx/components/FAISS/faiss.py

@@ -31,10 +31,10 @@ class FaissVectorStoreComponent(LCVectorStoreComponent):
         BoolInput(
             name="allow_dangerous_deserialization",
             display_name="Allow Dangerous Deserialization",
-            info="Set to True to allow loading pickle files from untrusted sources. "
-            "Only enable this if you trust the source of the data.",
+            info="Set to True to allow loading pickle files. WARNING: Only enable this if you trust the source "
+            "of the data. Malicious pickle files can execute arbitrary code on your system.",
             advanced=True,
-            value=True,
+            value=False,
         ),
         HandleInput(name="embedding", display_name="Embedding", input_types=["Embeddings"]),
         IntInput(