🌟 feat(__main__.py): add support for environment variables to configure serve function
The serve function in __main__.py has been refactored to use environment variables for configuration options. The following configuration options now support environment variables: host, workers, port, log_level, log_file, path, open_browser, and remove_api_keys. This change allows for greater flexibility and easier configuration of the serve function by using environment variables instead of command line options.
The environment variable name is changed from LANGFLOW_DATABASE_URL to langflow_database_url to maintain consistency with the naming conventions used in the codebase. This ensures that the code is more readable and easier to understand.
The `cache` option has been changed to use the `LANGFLOW_LANGCHAIN_CACHE` environment variable instead of `LANGCHAIN_CACHE` for clarity and consistency. Similarly, the `database_url` option now uses the `LANGFLOW_DATABASE_URL` environment variable instead of the previous behavior of using a local SQLite database by default. This change allows for more flexibility in configuring the cache and database connections.
🔧 chore(utils.py): refactor `setup_llm_caching` to extract cache setup logic into a separate function for better modularity and readability
The variable `cache_class` has been renamed to `cache_type` to improve clarity and consistency with the naming conventions. The `setup_llm_caching` function has been refactored to extract the cache setup logic into a separate function called `set_langchain_cache`. This improves modularity and readability of the code.
The `database_url` setting is now declared as an optional string (`Optional[str]`) instead of a required string. This change allows the application to handle cases where the `database_url` is not provided, providing more flexibility in the configuration.
The os module is imported to make use of the os.getenv() function. The loading of the environment variables is refactored to check for the presence of the "langflow_database_url" environment variable and use it as the value for the database_url variable if it is not provided. This allows for more flexibility in configuring the database URL. Additionally, the loading of environment variables is refactored to allow overriding with a .env file if the env_file parameter is provided. This allows for easier configuration management.
🔧 chore(__main__.py): import os module and refactor loading of environment variables
The noqa comments have been added to the import statements to ignore the linting errors raised by the linter. This ensures that the linting errors related to the imports are ignored and the code can be properly executed without any issues.
✨ feat(base.py): add error handling when creating database and tables to provide more informative error messages
The code now checks if the `settings.database_url` is provided and raises a `RuntimeError` if it is not. This ensures that the application does not attempt to create a database connection without a valid URL. Additionally, error handling has been added when creating the database and tables. If an exception occurs during the creation process, an error message is logged and a `RuntimeError` is raised with a more informative error message. This helps in identifying and resolving any issues related to the database creation process.
🔄 refactor(template/frontend_node/documentloaders.py): rename web_path variable to file_path for better clarity and consistency in the GutenbergLoader class
The Node class in the vertex/base.py file has been renamed to Vertex class to improve clarity and consistency in the naming conventions used throughout the codebase. In the template/frontend_node/documentloaders.py file, the web_path variable has been renamed to file_path in the GutenbergLoader class to provide better clarity and consistency in the naming conventions used within the class.
The password field is added to the extra_fields list in the VectorStoreFrontendNode class. This allows for the creation of a password input field in the frontend UI for this specific class.
The password field is added to the extra_field2 in the VectorStoreFrontendNode class. This allows the field to be displayed as a password input in the frontend, enhancing security by hiding the input characters.
📝 docs(chat.py): explain the need for building the root node before the rest of the graph
The root node in the graph was not being built before the rest of the graph, which caused issues when certain nodes required parameters that were not connected to them. By adding the missing build step for the root node, we ensure that all necessary connections and parameters are properly set up before building the rest of the graph. This improves the overall functionality and reliability of the chat module.
📝 docs(chat.py): explain the need for building the root node before the rest of the graph
The root node in the graph was not being built before the rest of the graph, which caused issues when certain nodes required parameters that were not connected to them. By adding the missing build step for the root node, we ensure that all necessary connections and parameters are properly set up before building the rest of the graph. This improves the overall functionality and reliability of the chat module.
The debug log statement "Database and tables created" has been removed as it is unnecessary. Instead, a success log statement "Database and tables created successfully" has been added to indicate that the database and tables were created without any errors. This improves the clarity of the log messages and provides better feedback during the database setup process.
The environment variable prefix used in the settings.py file has been updated from "LANGFLOW_" to "langflow_" to maintain consistency with the naming conventions used in the project. This change ensures that all environment variables related to the Langflow backend have a consistent prefix.
Added the documentation link for the PyPDFDirectoryLoader in the config.yaml file. This allows users to access the documentation on how to use the PyPDFDirectoryLoader for loading PDF documents.
The logger.debug() statements are added to provide more visibility into the process of creating the database and tables. This can be helpful for debugging and troubleshooting purposes.
The change is made to improve the logging of exceptions. Using logger.exception instead of logger.error allows the logger to log the full traceback of the exception, providing more detailed information for debugging purposes.
🔧 refactor(documentloaders.py): remove redundant code for adding fields in DocumentLoaderFrontNode class
The code in the DocumentLoaderFrontNode class has been refactored to extract the logic for building directory loader fields into a separate function called build_directory_loader_fields(). This improves code readability and maintainability by separating concerns and reducing duplication. The redundant code for adding fields in the DocumentLoaderFrontNode class has also been removed.
The default value of the "Persist" extra field in the VectorStoreFrontendNode class has been changed from True to False. This change was made to align the default value with the desired behavior of the application.
The output_key field is now only added to the template if the template type is not VectorStoreRetrieverMemory. This ensures that the output_key field is not added unnecessarily for templates of this type.
The `logger.exception(exc)` line has been added to log the exception that occurred. This will help with error handling and debugging by providing more information about the exception that caused the error.
The code now checks if the "retriever" key exists in the params dictionary and if the value has an "as_retriever" method. If it does, the value is replaced with the result of calling the "as_retriever" method. This change ensures that the "retriever" parameter is always an instance of the retriever object, preventing potential attribute errors later in the code.
The logic for updating memory keys in the `run.py` file has been refactored to use a loop and a try-except block. Instead of individually assigning values to `input_key`, `output_key`, and `memory_key`, the keys and attributes are now stored in lists. The loop iterates over the lists and attempts to set the attribute values using `setattr()`. If an attribute does not exist, a `ValueError` is caught and a debug log message is printed. This refactoring improves code readability and maintainability.
This pull request addresses a bug related to WebSocket functionality
that was causing an empty error to be displayed. The bug resulted in
incorrect error-handling behavior within the WebSocket implementation,
leading to confusion and difficulties in identifying and resolving
issues.
By addressing this bug and improving the WebSocket error handling
mechanism, this pull request significantly enhances the overall
stability and reliability of the application.
The missing return statement caused the function to not return any value when the node_type is not "VertexAI". Adding the return statement ensures that the function returns the instantiated class object with the provided parameters in all cases.
🔧 fix(loading.py): remove empty lines
The type error for the service_account import in llm.py is ignored to prevent a linting error. In loading.py, empty lines were removed for code cleanliness.
The ChatVertexAI integration is temporarily commented out due to a bug that causes an AttributeError. This bug needs to be resolved before the integration can be activated again.
🔀 chore(custom_lists.py): add ChatVertexAI to the import statements for better modularity and extensibility
🔀 chore(custom_lists.py): add ChatVertexAI to the llm_type_to_cls_dict for better compatibility and flexibility
🔀 chore(llms.py): change required field for credentials to be optional for better user experience
🔀 chore(llms.py): add advanced and show fields for specific fields related to VertexAI for better configurability
The import statements in `custom_lists.py` have been reformatted to improve readability and maintainability. The `ChatVertexAI` class has been added to the import statements to enhance modularity and extensibility.
The `ChatVertexAI` class has been added to the `llm_type_to_cls_dict` dictionary in `custom_lists.py` to improve compatibility and flexibility.
In `llms.py`, the `required` field for the `credentials` field has been changed to be optional for a better user experience.
The `advanced` and `show` fields have been added to specific fields related to VertexAI in `llms.py` to provide better configurability.
The logic to find the matched_type in the Edge class has been simplified by removing unnecessary nested loops and using a single generator expression. This improves the readability and efficiency of the code.
🔧 chore(loading.py): call initialize_vertexai function when node_type is "VertexAI"
The `llm.py` file now includes a new function `initialize_vertexai` that initializes the VertexAI credentials if a `credentials` parameter is provided. This allows for the usage of VertexAI credentials in the application. In `loading.py`, the `initialize_vertexai` function is called when the `node_type` is "VertexAI", ensuring that the VertexAI credentials are properly initialized for that specific node type.
The condition for showing fields in the VertexAI class has been simplified to exclude specific field names. This improves readability and maintainability of the code.
📝 chore(llms.py): improve field configuration for VertexAI template and modify field type for "credentials"
The `add_extra_fields` method is modified to add an additional field called "credentials" for the VertexAI template. The field is of type "file" and is required. It allows the user to upload a JSON file as credentials. The `format_openai_field` method is also updated to handle the new "credentials" field.
🐛 fix(base.py): add optional 'name' parameter to the format method in FieldFormatter
In loading.py, the error message when failing to build a connection to the database is now formatted in a more readable way. This improves the clarity of the error message and makes it easier to identify the cause of the issue.
In base.py, the format method in FieldFormatter now accepts an optional 'name' parameter. This allows for more flexibility when formatting the field and provides the ability to include the field name in the formatting process if needed.
The ConversationEntityMemory integration is now documented in the config.yaml file. This provides a link to the documentation for using the ConversationEntityMemory module with SQLite as the underlying storage. This change improves the clarity and accessibility of the documentation for this integration.
The code now catches an additional exception when the object does not have a 'conn' field. This improves the error handling by providing a more specific error message when building a connection to the database fails.
The 'from_method_nodes' dictionary in the 'MemoryCreator' class in 'base.py' has been updated to include the 'SQLiteEntityStore' class and its '__init__' method. This change allows the 'SQLiteEntityStore' class to be used as a memory node in the LangFlow application.
✨ feat(memories.py): add show attribute to entity_store field in MemoryFrontendNode class
The condition to add extra fields in the MemoryFrontendNode class has been fixed to correctly check if any of the base classes are in a list of base message classes. Additionally, the show attribute has been added to the entity_store field in the MemoryFrontendNode class to control its visibility.
The "kwargs" and "args" parameters are now excluded from the generated template dictionary. This ensures that only the relevant parameters are included in the template, improving the accuracy and usefulness of the generated template.
When processing the data graph, if the langchain_object is of an unknown type, a ValueError is now raised. This provides better error handling and helps identify any unexpected langchain_object types that may be encountered during processing.
✨ feat(base.py): add support for creating MemoryFrontendNode from a specific method in ZepChatMessageHistory class
The `MemoryCreator` class now supports creating `MemoryFrontendNode` from a specific method in the `ZepChatMessageHistory` class. This is achieved by adding the `from_method_nodes` dictionary with the method name as the key and the class name as the value. The `build_template_from_method` function is used to create the `MemoryFrontendNode` from the specified method. This enhancement allows for more flexibility in creating `MemoryFrontendNode` instances.
✨ feat(base.py): add support for ZepRetriever in from_method_nodes dictionary
The build_template_from_method call for the MultiQueryRetriever in the get_signature method was using the incorrect method name "from_llm". It has been fixed to use the correct method name "from_method_nodes". Additionally, support for the ZepRetriever has been added to the from_method_nodes dictionary to allow for its usage in the RetrieverCreator class.
