feat: add model config #487
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
11zhouxuan
reviewed
Dec 18, 2024
source/lambda/online/common_logic/langchain_integration/chains/__init__.py
Outdated
Show resolved
Hide resolved
11zhouxuan
approved these changes
Dec 19, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #
🤖 AI-Generated PR Description (Powered by Amazon Bedrock)
Description
This pull request introduces several enhancements and modifications to the language model integration and chat functionality. The primary changes include:
langchain_integrationmodule, including modifications to thechains,chat_models, and the addition of a newmodel_config.pyfile.common_utils/constant.pyfile.common_entry.pyfile, which handles the online entries for the application.ChatBot.tsxfile in the portal's frontend.These changes aim to improve the overall performance, functionality, and integration of the language models with the application's chat and query handling capabilities.
Type of change
File Stats Summary
File number involved in this PR: 10, unfold to see the details:
The file changes summary is as follows:
Claude2ToolCallingChainclass into a genericToolCallingChainclass and dynamically creates subclasses for different models based on configurations.DEFAULTto theLLMModelTypeclass, and includes several other model types for the Anthropic Claude language model.ChatChainclass that sets default model kwargs and handles common logic.chain_classesthat dynamically creates chat chain classes for all models defined inMODEL_CONFIGS.| source/lambda/online/common_logic/langchain_integration/chains/query_rewrite_chain.py | 10 added, 32 removed | The code changes introduce a new
QueryRewriteChainclass that uses the default model specified inMODEL_CONFIGS. It also dynamically creates chain classes for all models inMODEL_CONFIGSfor theQUERY_REWRITE_TYPEtask, and adds them to the global namespace. || source/lambda/online/common_logic/langchain_integration/chains/init.py | 154 added, 120 removed | This code change introduces a new way to dynamically create LLM chain classes based on the model ID and intent type. It adds methods to convert model IDs to valid Python class names, and a factory method
create_for_modelto create a new chain class dynamically. It also reorganizes the imports of different chain types into separate functions and updates the global namespace accordingly. |🤖 AI-Generated PR Description (Powered by Amazon Bedrock)
Description
This pull request introduces several changes to the codebase, primarily focused on enhancing the integration with LangChain and improving the chat functionality. The modifications span multiple files and directories, including the common utilities, LangChain integration components, and the chatbot user interface.
The key changes include:
constant.pyfile to accommodate new constants or modify existing ones.chat_chain.py,query_rewrite_chain.py,rag_chain.py, andtool_calling_chain_api.py, to improve the chat experience and enhance the functionality of the language model.bedrock_models.pyfile, which likely involves changes to the underlying language models used in the application.model_config.py, which potentially introduces configuration settings for the language models.common_entry.pyfile, which may affect the common entry point for the online lambda function.ChatBot.tsxfile in the portal's source code, suggesting updates to the chatbot user interface.Type of change
File Stats Summary
File number involved in this PR: 11, unfold to see the details:
The file changes summary is as follows:
DEFAULTin theLLMModelTypeclass, and adds three new model types:CLAUDE_INSTANCE,CLAUDE_2, andCLAUDE_21.BedrockModelclass that replaces the previous model-specific classes likeClaude2,Claude21, etc. It dynamically creates model classes based on theMODEL_CONFIGSand adds them to the module's global namespace. This change likely aims to simplify and centralize model configuration management.ChatChainthat extendsLLMChainfor chat-based language models. It also defines a dictionarychain_classesthat maps model IDs to their correspondingChatChainsubclasses, which are dynamically created usingLLMChain.create_for_model. The subclasses are then added to the global namespace.create_for_modelmethod, which takes the model ID and task type as arguments. Additionally, it removes the individual class definitions for specific models like Claude and replaces them with a singleQueryRewriteChainclass that uses theLLMModelType.DEFAULTmodel ID.model_id_to_class_namemethod to convert the model ID to a valid Python class name, and acreate_for_modelmethod to create a new chain class dynamically using the model configuration. It also reorganizes the imports of various chain classes and consolidates some of them into separate modules.create_for_modelin theModelclass to create a model instance based on the provided model ID. It also adds a helper methodmodel_id_to_class_nameto convert the model ID to a valid Python class name. Additionally, it refactors the import statements for different model classes and updates the_import_bedrock_modelsfunction to use a more concise approach.🤖 AI-Generated PR Description (Powered by Amazon Bedrock)
Description
This pull request introduces several updates and improvements to the LangChain integration and chat functionality. The main changes include:
model_config.pyfile to centralize model configuration settings.These changes aim to enhance the overall performance, scalability, and maintainability of the LangChain integration and chat functionality within the application.
Type of change
File Stats Summary
File number involved in this PR: 12, unfold to see the details:
The file changes summary is as follows:
DEFAULTto theLLMModelTypeclass, likely representing a default model ID for language models.BedrockModelclass that serves as a base class for different language models. It imports theMODEL_CONFIGSfrom themodel_configmodule and dynamically creates model classes based on the configurations. The model classes are added to the module's global namespace for easy access.QueryRewriteChainclass that uses theLLMModelType.DEFAULTmodel and achain_classesdictionary that dynamically creates chain classes for all available models inMODEL_CONFIGS. The previously defined model-specific chain classes are removed.create_for_modelin theModelclass, which dynamically creates a new class for a given model ID using the configurations fromMODEL_CONFIGS. It also refactors the import statements for different model classes and updates themodel_id_to_class_namemethod to handle model IDs with version numbers and vendor prefixes.RagChainclass that serves as a base class for all RAG (Retrieval-Augmented Generation) chains. It simplifies the codebase by consolidating multiple model-specific chain classes into a single class that dynamically creates chains for different models using theMODEL_CONFIGSdictionary. Additionally, it adds a dictionarychain_classesthat maps model IDs to their corresponding chain classes, making it easier to access and instantiate chains for different models.create_for_modelin theLLMChainclass, which dynamically creates a new class for a specific model based on the model ID and intent type. It also adds a helper methodmodel_id_to_class_nameto convert the model ID to a valid Python class name. Additionally, it refactors the import statements for various chain modules, using a centralizedchain_classesdictionary to store the imported classes and updating the global namespace accordingly.🤖 AI-Generated PR Description (Powered by Amazon Bedrock)
Description
This pull request introduces changes to integrate the LangChain library for natural language processing and model management. The main updates include:
model_config.pyto handle the configuration and loading of language models.chat_chain.py,query_rewrite_chain.py,rag_chain.py,tool_calling_chain_api.py) to incorporate LangChain components.bedrock_models.pyfile to support LangChain model integration.common_entry.pyto utilize the new LangChain chains and models.ChatBot.tsxandconst.tsfiles to accommodate the backend changes.The primary motivation behind this change is to leverage the powerful capabilities of LangChain for natural language processing tasks, such as question answering, text generation, and model management. LangChain provides a modular and extensible framework for building applications with language models.
This change does not require any external dependencies beyond the LangChain library itself.
Type of change
File Stats Summary
File number involved in this PR: 12, unfold to see the details:
The file changes summary is as follows:
HISTORY_CHATBOT_IDto the existing set of constants.DEFAULTin theLLMModelTypeclass, likely representing a default model identifier, and adds three constants for different versions of the Claude language model from Anthropic.MODEL_CONFIGSmodule. TheBedrockModelclass serves as a base class for all models, and themodel_classesdictionary maps model IDs to their corresponding instances.QueryRewriteChainclass that uses a default model specified inMODEL_CONFIGS. It also adds all available query rewrite chain classes to the module's global namespace based on the models defined inMODEL_CONFIGS.ChatChainfor LLM chat chains, with a default model type and model parameters. It also adds a dictionarychain_classesthat maps model IDs to their correspondingChatChainsubclasses, which are dynamically created using theMODEL_CONFIGSand added to the global namespace.RagChainthat serves as a base class for all RAG (Retrieval-Augmented Generation) chains, replacing the previousClaude2RagLLMChain. It also removes several model-specific RAG chain classes and instead dynamically creates chain classes for all available models using a dictionary comprehension. The system prompt addition logic is also refactored into a separate function.model_id_to_class_nameto convert model IDs to valid Python class names, and a factory methodcreate_for_modelto dynamically create model classes based on configurations. It also refactors the import statements for bedrock models.model_id_to_class_namemethod to convert model IDs to valid Python class names, and acreate_for_modelfactory method to create a new LLMChain subclass for a specific model and intent type. It also reorganizes the imports of different chain types into separate functions.🤖 AI-Generated PR Description (Powered by Amazon Bedrock)
Description
This pull request introduces changes to the LangChain integration and the chatbot frontend. The main modifications are:
model_config.pyfile to centralize the configuration of LangChain models.These changes aim to enhance the overall performance, scalability, and maintainability of the chatbot application.
Type of change
File Stats Summary
File number involved in this PR: 12, unfold to see the details:
The file changes summary is as follows:
DEFAULTto theLLMModelTypeclass, representing a default model identifier.ToolCallingBaseChainfor tool calling chains, replacing the previous model-specific classes likeClaude2ToolCallingChain. It dynamically creates chain classes for all available models using theMODEL_CONFIGSinstead of hardcoding them.QueryRewriteBaseChainfor query rewriting task chains, and dynamically creates chain classes for all available models defined inMODEL_CONFIGS. It removes the hard-coded subclasses for specific models like Claude variants, and instead generates the subclasses programmatically based on the model configurations.BedrockBaseModelthat extends theModelclass, and defines a dictionarymodel_classesthat maps model IDs to their corresponding model classes. The model classes are dynamically created using theBedrockBaseModel.create_for_modelmethod and theMODEL_CONFIGSimported frommodel_config.create_for_modelto dynamically create a model class based on the provided model ID. It also adds a helper methodmodel_id_to_class_nameto convert the model ID to a valid Python class name. Additionally, it simplifies the import of bedrock models by using a singlemodel_classesimport instead of individual imports.RagBaseChainfor RAG (Retrieval-Augmented Generation) LLM chains, which dynamically creates chain classes for different LLM models based on the configurations inMODEL_CONFIGS. It replaces the previous approach of defining individual chain classes for each model.ChatBaseChainfor handling chat-based language model interactions. It defines a default model ID, task type, and model parameters. Several existing chat chain classes are modified to inherit fromChatBaseChaininstead ofClaude2ChatChain. Additionally, a dictionarychain_classesis created to dynamically generate chat chain classes for each model configuration defined inMODEL_CONFIGS.🤖 AI-Generated PR Description (Powered by Amazon Bedrock)
Description
This pull request introduces several enhancements and modifications to the codebase, primarily focused on improving the integration with LangChain, a framework for building applications with large language models. The changes include updates to various utility functions, chain implementations, and chat model configurations.
Some notable changes are:
constant.pyandprompt_utils.pyfiles in thecommon_utilsmodule, likely related to handling constants and prompt generation.chat_chain.py,conversation_summary_chain.py,query_rewrite_chain.py,rag_chain.py, andtool_calling_chain_api.py, which may involve improvements or bug fixes related to these specific chains.model_config.pyin thelangchain_integrationmodule, which is likely responsible for managing configurations related to language models.bedrock_models.pyfile in thechat_modelsmodule, potentially introducing changes to the handling of Bedrock language models.common_entry.pyfile in theonline_entriesmodule, which may affect the common entry point for online functionality.ChatBot.tsxfile in theportal/src/pages/chatbotdirectory, suggesting updates to the chatbot user interface.const.tsfile in theportal/src/utilsdirectory, which may involve modifications to constant values used throughout the application.Type of change
File Stats Summary
File number involved in this PR: 14, unfold to see the details:
The file changes summary is as follows:
all_knowledge_rag_toolvariable, using double quotes for the dictionary key instead of single quotes.HISTORY_CHATBOT_ID, use it to store and retrieve the chatbot ID from localStorage, and remove the stored chatbot ID when manually changing the chatbot option.ChatBaseChainthat inherits fromLLMChainand defines common properties and methods for chat-based LLM chains. It also importsMODEL_CONFIGSandLLMModelTypeandLLMTaskTypefrom other modules. Additionally, it creates a dictionarychain_classesthat maps model IDs to their corresponding chat chain classes, which are dynamically created using thecreate_for_modelmethod ofChatBaseChain.BedrockBaseModelthat extends theModelclass and uses theLLMModelType.DEFAULTmodel ID. It also imports theMODEL_CONFIGSfrommodel_config. Instead of defining individual model classes likeClaude2,ClaudeInstance, etc., it creates a dictionarymodel_classesthat maps model IDs fromMODEL_CONFIGSto their respective model classes dynamically created usingBedrockBaseModel.create_for_model().RagBaseChainfor RAG (Retrieval-Augmented Generation) LLM chains, which uses a default model type. It also removes specific classes for different model types (e.g., Claude2RagLLMChain, Claude21RagLLMChain) and instead dynamically creates chain classes for all available model configurations using a dictionary comprehension.QueryRewriteBaseChainis created with a default model type.chain_classesis created to dynamically generate query rewrite chain classes for all available models. || source/lambda/online/common_logic/langchain_integration/chains/init.py | 143 added, 121 removed | This code change introduces a dynamic class creation mechanism for LLMChains based on model configurations. It adds a new method
create_for_modelthat generates a new class for a specific model ID and intent type. The class name is derived from the model ID using a naming convention. The new class inherits from the base LLMChain class and has pre-defined attributes likeintent_type,model_id, anddefault_model_kwargsbased on the model configuration. Additionally, it refactors the import statements for different chain types into separate functions and adds a newchain_classesmodule for each chain type. || source/lambda/online/common_logic/langchain_integration/chat_models/init.py | 97 added, 43 removed | The code changes introduce a new factory method
create_for_modelin theModelclass to dynamically create a model class based on the provided model ID. It also adds support for several new models, including variations of Claude and LLaMa models for different regions (US, EU, APAC). Themodel_classesmodule is imported instead of importing individual model classes directly. Additionally, themodel_id_to_class_namemethod is added to convert model IDs to valid Python class names. || source/lambda/online/common_logic/langchain_integration/chains/tool_calling_chain_api.py | 7 added, 44 removed | The code changes introduce a new base class
ToolCallingBaseChainthat replaces the previousClaude2ToolCallingChainclass, and dynamically creates subclasses for different model configurations using a dictionary comprehension. |