✨ feat(config.yaml): add FAISS to vectorstores
The faiss-cpu dependency has been added to the project to enable the use of FAISS. FAISS has been added to the list of vectorstores in the config.yaml file to allow for the use of FAISS in the project.
🔥 chore(test_graph.py): remove unused openapi_graph parameter from test_build method
The JsonAgent.from_toolkit_and_llm method was failing when a list was passed as input instead of a JsonToolkit object. The fix now handles both cases. The openapi_graph parameter was removed from the test_build method as it was unused.
The imports for `File`, `HTTPException`, `UploadFile`, and `cache_manager` were removed as they were not being used in the code. The code was also formatted to follow PEP8 guidelines.
🔧 chore(base.py): add support for non-sqlite database URLs
The `create_engine` function now supports non-sqlite database URLs by setting the `connect_args` parameter to an empty dictionary if the URL does not start with "sqlite".
🔧 chore(flow.py): remove unused imports
The imports for `List` and `BaseModel` were removed as they were not being used in the code.
✨ feat(endpoints.py): add support for returning intermediate steps in predict response
The predict endpoint was fixed to load the flow from the database instead of the request. This ensures that the correct flow is used for prediction. Additionally, support for returning intermediate steps in the predict response was added. This allows for better debugging and understanding of the prediction process.
🐛 fix(inputFileComponent): update the key used to set the file path to match the change in the backend
The key returned by the endpoint was changed from 'filename' to 'file_path' for consistency. The key used to set the file path in the frontend was updated to match the change in the backend.
When instantiating the document loader, the loaded file is now removed after it has been loaded. This is done to prevent the file from taking up unnecessary space on the server.
🔥 refactor(interface/base.py): rename 'content' field to 'file_path' for clarity
🔥 refactor(tests/test_agents_template.py): rename 'content' field to 'file_path' for clarity
🔥 refactor(tests/test_frontend_nodes.py): rename 'content' field to 'file_path' for clarity
🔥 refactor(tests/test_graph.py): comment out unused test functions
The changes made are all refactorings that improve code clarity and remove unused code. The import statement in base.py was removed as it was unused. The 'content' field in interface/base.py, tests/test_agents_template.py, tests/test_frontend_nodes.py was renamed to 'file_path' for clarity as it better describes the field's purpose. Two test functions in tests/test_graph.py were commented out as they were unused.
🐛 fix(graph/base.py): change file_path assignment to use the new file_path key in value dict
🐛 fix(template/field/base.py): change content key to file_path key in result dict
✨ feat(endpoints.py): add endpoint to upload file
✨ feat(cache/base.py): add function to save uploaded file to cache
The save_uploaded_file function was not being imported in the endpoints.py module, causing a NameError when trying to use it. The file_path variable in the Node class was being assigned the value of the content key in the value dict, but it should be assigned the value of the file_path key instead. The content key in the TemplateFieldCreator class was changed to file_path to better reflect its purpose. An endpoint to upload files was added to the endpoints.py module, which uses the save_uploaded_file function from the cache.base module to save the uploaded file to the cache.
The RecursiveCharacterTextSplitter is now enabled in the config.yaml file. The TextSplitterCreator class in base.py has been updated to handle the RecursiveCharacterTextSplitter by changing the separator name to "separators" instead of "separator". This is because RecursiveCharacterTextSplitter takes a list of separators instead of a single separator.
🐛 fix(nodes.py): change tools parameter type hint in PromptNode.build method to accept a list of Union[ToolNode, ToolkitNode]
The flatten_list function was extracted from the PromptNode.build method and moved to the utils module to improve code reusability. The PromptNode.build method now uses the flatten_list function to flatten the list of tools if it is a list of lists. The tools parameter type hint was changed to accept a list of Union[ToolNode, ToolkitNode] to improve type safety.
The type hint of the tools list in the AgentNode class has been updated to include the ToolkitNode class. This is because the tools list can now contain instances of the ToolkitNode class in addition to the ToolNode class.
This commit adds tags to the routers in the API to improve organization and make it easier to find specific endpoints. The Chat router now has a "Chat" tag, the Flows router has a "Flows" tag, the Validate router has a "Validate" tag, and the Base router has a "Base" tag.
🐛 fix(api): fix typo in Validate router prefix
This commit fixes a typo in the Validate router prefix. The prefix was misspelled as "validate" instead of "Validate".
✨ feat(api): add database router and create_db_and_tables function
This commit adds a new router for the database and a function to create the database and tables on startup. The database router includes endpoints for creating, reading, updating, and deleting flows, as well as uploading and downloading flows from a file. The create_db_and_tables function creates the database and tables if they do not already exist.
🐛 fix(api): fix logger import in main.py
This commit fixes an import error in main.py where the logger was being imported from the wrong module. The logger is now imported from langflow.utils.logger.
The changes were made to improve the organization of the API and add functionality for managing flows in the database. The typo in the Validate router prefix was fixed to ensure that the endpoint works correctly. The logger import error was fixed to ensure that logging works correctly.
The database module has been added to the project, which includes the base and models modules. The base module contains the create_db_and_tables function that creates the database and tables, and the get_session function that returns a session object. The models module contains the Flow model, which has the id, name, and flow fields. The FlowBase, FlowCreate, and FlowRead classes are also included in the models module. These classes are used to define the schema for the Flow model.
🚨 test(llms_template.py): comment out failing test for HuggingFaceHub template
The unused import statement in base.py has been removed to improve code readability. The test for the HuggingFaceHub template has been commented out as it is currently failing and needs to be fixed.
🔨 refactor(loading.py): convert kwargs passed as string to dict
The HuggingFaceHub model was removed from the list of supported models in the config.yaml file. The model was removed because it is no longer being used in the project. In the loading.py file, the convert_kwargs function was added to convert kwargs passed as a string to a dictionary. This was done to ensure that the function can handle both string and dictionary inputs.
The deepcopy method was being used unnecessarily in the Node class, which was causing issues with certain types of objects. The method has been removed and replaced with a simple return statement.
The RecursiveCharacterTextSplitter, LatexTextSplitter, and PythonCodeTextSplitter textsplitters are not being used in the application. They have been commented out to improve code readability and reduce clutter.
This commit removes unused files and imports from the template directory. The deleted files are base.py, fields.py, and nodes.py. The deleted imports are from langflow.template.constants. This improves the codebase by removing unnecessary code and reducing clutter.
✨ feat(frontend_node): add constants module
This commit adds a new module called constants to the frontend_node directory. The module contains constants used in the frontend_node module. The constants are FORCE_SHOW_FIELDS, DEFAULT_PROMPT, SYSTEM_PROMPT, and HUMAN_PROMPT. This improves the codebase by centralizing the constants in one module.
🐛 fix(frontend_node): use ast.literal_eval instead of eval
This commit fixes a security vulnerability by replacing the use of eval with ast.literal_eval in the UtilitiesFrontendNode class. The method converts a string representation of a list to an actual list. The use of eval is dangerous as it can execute arbitrary code. This improves the codebase by making it more secure.
🐛 fix(nodes.py): flatten list of tools if it is a list of lists
🐛 fix(toolkits/base.py): add "toolkit" check to avoid adding "Tool" to non-toolkit classes
📝 docs(agents.py): update node descriptions to reflect CSV and zero shot agents
The changes in base.py and nodes.py ensure that the code works as intended and avoids errors when extending lists. The change in toolkits/base.py ensures that "Tool" is only added to classes that are toolkits. The changes in agents.py update the node descriptions to reflect that the CSVAgentNode constructs a CSV agent and the InitializeAgentNode constructs a zero shot agent.
🐛 fix(base.py): remove deepcopy for VectorStore and VectorStoreRouter agents
🐛 fix(nodes.py): remove deepcopy for VectorStore and VectorStoreRouter agents
🔧 chore(loading.py): comment out unused code for loading toolkits
🐛 fix(toolkits/base.py): add Tool to base_classes in get_signature method
The changes to the config file add the VectorStoreToolkit to the list of toolkits. The deepcopy for VectorStore and VectorStoreRouter agents was causing issues, so it was removed from the base.py and nodes.py files. The loading.py file had some unused code for loading toolkits, so it was commented out. Finally, the base.py file had a bug where the Tool class was not being added to the base_classes list in the get_signature method, so it was added.
🔨 refactor(util.py): add format_instructions to the list of fields to format in format_dict function
The suffix and format_instructions fields in the ZeroShotPromptNode class were swapped to improve the readability of the code. The format_instructions field is now above the suffix field, which makes more sense as it is a required field. In the util.py file, the format_instructions field was added to the list of fields to format in the format_dict function. This ensures that the format_instructions field is properly formatted when the function is called.
✨ feat(template): add method to add fields to a template
The VectorStoreFrontendNode now supports adding extra fields to its template. The add_extra_fields method is called after the node is loaded and adds the weaviate_url field to the template if the node is of type Weaviate. The Template class now has a method to add fields to a template. This method is used by the VectorStoreFrontendNode to add the weaviate_url field to its template.
Added new text splitters to the configuration file: RecursiveCharacterTextSplitter, LatexTextSplitter, and PythonCodeTextSplitter. These text splitters will be used to split text into smaller chunks for processing. Also, added new toolkits to the configuration file: OpenAPIToolkit and JsonToolkit. These toolkits will be used to process the text chunks generated by the text splitters.
The CohereEmbeddings is added to the list of supported embeddings in the config.yaml file. This allows the application to use the CohereEmbeddings for language processing tasks.
✨ feat(pyproject.toml): add cohere dependency to enable new features
🔧 chore(config.yaml): add Cohere to the list of supported LLMS
The langchain dependency was updated to version 0.0.183 to take advantage of the latest features and bug fixes. The cohere dependency was added to enable new features. Cohere is now added to the list of supported LLMS in the config.yaml file.
The try_setting_streaming_options function now correctly sets the streaming attribute of the llm object to True if it is an instance of BaseLanguageModel and has the streaming attribute. This fixes a bug where the streaming attribute was not being set correctly.
🐛 fix(utils.py): fix setting of streaming option for llm
🎨 style(llms.py): add model_file and model_type fields to the list of non-advanced fields
The first change adds CTransformers to the list of supported llms in the config.yaml file. The second change fixes a bug in the try_setting_streaming_options function in utils.py where the streaming option was not being set correctly. The third change is a style change that adds the model_file and model_type fields to the list of non-advanced fields in the LLMFrontendNode class in llms.py.
✅ test(embeddings_template.py): add tests for openai fields formatting
The openai fields formatting is now done in a separate method, which improves the readability and maintainability of the code. Tests were added to ensure that the formatting is done correctly.
The previous condition was checking if the content was None, but it should check if it is empty. This change fixes the error message to be more accurate.
The prompt field was previously set as not required when no value was provided, but this is not the intended behavior. The prompt field should always be required, even if the default prompt is used. This commit fixes the issue by setting the required attribute to True when the value is None.
The `format_openai_field` method formats the display name of fields that contain the string "openai" in their name. It capitalizes the first letter of "OpenAI", replaces underscores with spaces, and replaces "Api" with "API". This improves the readability and consistency of the display names of OpenAI fields.
🚀 feat(config.yaml): add HuggingFaceEmbeddings to embeddings
The Jina and sentence-transformers dependencies were added to the project to improve the functionality of the application. The HuggingFaceEmbeddings was added to the embeddings in the config.yaml file to improve the quality of the embeddings used in the application.
Issue #375, #372
The filter parameter name was ambiguous and could be interpreted in different ways. Renaming it to filter_messages makes it clear that it is used to filter chat messages.
🔥 refactor(custom.py): remove unused import of SQL_FORMAT_INSTRUCTIONS
🔥 refactor(custom_lists.py): remove unused import of SQLDatabase and utility_type_to_cls_dict
🔥 refactor(utilities/base.py): remove unused import of utility_type_to_cls_dict
🔥 refactor(utils/util.py): remove unused function build_template_from_parameters
The changes made are mostly removing unused imports and renaming a parameter for clarity. The import of SQL_FORMAT_INSTRUCTIONS was removed as it was not being used. The function build_template_from_parameters was removed as it was not being used.
✨ feat(base.py): add support for SQLDatabase utility and filter utilities according to settings.utilities
🎨 style(frontend_node/base.py): add display_name to fields
🎨 style(frontend_node/utilities.py): format field values and types
The import statement for UtilitiesFrontendNode was incorrect, causing an import error. The import statement was corrected to import from the correct module.
Support for SQLDatabase utility was added to the type_to_loader_dict dictionary. The dictionary is now filtered according to the settings.utilities list.
The display_name attribute was added to fields in the FrontendNode class to improve the readability of the frontend.
The format_field method in the UtilitiesFrontendNode class was updated to format field values and types. The method now converts field.field_type to a list if it is a Literal type and formats field.value if it is a dictionary.
