This handler (com.rierino.handler.langchain4j.LangChainModelEventHandler) provides ability to use LangChain models from various providers
Handler Parameters
Parameter
Definition
Example
Default
model.state
Name of the state manager with model configurations
genai_model
-
model.domain
Name of the model domain to include
chat
-
saga.handler
Name of the saga event handler that executes tool sagas
-
saga
saga.state
Name of the state manager with saga definitions (used as tools)
-
saga
Models
Models used by this event handler are stored in a regular state manager with the following model parameters:
Class: Java class name for the LLM provider model (e.g. dev.langchain4j.model.openai.OpenAiChatModel)
Methods: List of methods and their parameters to call for initializing LLM provider model (e.g. { "apiKey": ["TBD"] } ).
Memory: State manager that should serve as the memory for assistants.
Memory Size: Maximum number of messages to retain in assistant's memory.
Method parameters can use value and secret injection using ${{}} and #{{}} notation, similar to event runner configuration loaders, allowing use of Secret entries for securing credentials.
Actions
LLMChat
Performs a chat interaction with a target LLM model provider. Event metadata fields applicable for this action are as follows:
Field
Definition
Example
Default
Domain
ID of the model to use
chatgpt_chat
Input Element
Json path for the input in request event payload
message
-
Output Element
Json path for the output in response event payload
output
-
Input element can include the following fields:
Base64 contents can always be passed as data URI as well, with the correct data prefixes.
With event metadata parameters as:
Parameter
Definition
Example
Default
Message Pattern
Used for tool sagas, allowing merging data from original event to saga call (with arguments as agent generated input and payload as original payload)
merge(arguments, {user: payload.user})
-
Full Result
Whether response should include full AI call details (e.g. token counts, tool executions)
true
false
Json Response
Whether model response should be automatically parsed as a json object
true
false
LLMGenerateImage
Performs an image generation with a target LLM model provider. Event metadata fields applicable for this action are as follows:
Field
Definition
Example
Default
Domain
ID of the model to use
dalle_gen
Input Element
Json path for the input in request event payload
{prompt: "", imageCount: ""}
-
Output Element
Json path for the output in response event payload
imageBase64
-
With event metadata parameters as:
Parameter
Definition
Example
Default
Full Result
Whether response should include full AI call details (e.g. token counts)
true
false
LLMEditImage
Performs an image edit with a target LLM model provider. Event metadata fields applicable for this action are as follows:
