Project

Namespaces

Public Vars

com.rpl.agent-o-rama.langchain4j

Convenience functions for working with LangChain4j model APIs in agent nodes.

This namespace provides a small subset of common functionality wrapped in Clojure-friendly functions. It focuses on the most frequently used

operations for chat models, tool calling, and JSON response formatting.

For advanced LangChain4j features not covered here, you can directly use the LangChain4j Java API within your agent node functions.

basic-chat

(basic-chat model prompt)

Performs a simple chat interaction with a model using a string prompt.

This is the simplest way to interact with a chat model. The prompt is automatically converted to a UserMessage and sent to the model.

Args:

  • model - ChatModel instance (obtained from get-agent-object)
  • prompt - String prompt to send to the model

Returns:

  • String - The model’s response text

Example:

(let [model (aor/get-agent-object agent-node "openai-model")]
  (lc4j/basic-chat model "What is the capital of France?"))
;; => "The capital of France is Paris.")

chat

(chat model request)

Performs a chat interaction with a model using a structured request.

This function provides more control over the chat interaction, supporting both simple message sequences and full ChatRequest objects with advanced configuration options.

Args:

  • model - ChatModel instance (obtained from get-agent-object)
  • request - Either:
    • Sequential collection of messages (strings or message objects)
    • ChatRequest object with full configuration

Returns:

  • ChatResponse - Full response object with metadata and tool calls

Example:

(let [model (aor/get-agent-object agent-node "openai-model")]
  ;; With message sequence
  (lc4j/chat model [(UserMessage. "Hello") (UserMessage. "How are you?")])
  ;; With ChatRequest
  (lc4j/chat model (lc4j/chat-request
                    [(UserMessage. "Calculate 2+2")]
                    {:tools [calculator-tool]
                     :temperature 0.1})))

chat-request

(chat-request messages)(chat-request messages {:keys [frequency-penalty max-output-tokens model-name presence-penalty response-format stop-sequences temperature tool-choice tools top-k top-p]})

Creates a ChatRequest object for advanced model interactions.

This function builds a structured request with full control over model parameters, tool usage, and response formatting. String messages are

automatically converted to UserMessage objects.

Args:

  • messages - Collection of messages (strings or message objects)
  • options - Optional map with configuration:
    • :frequency-penalty - Number, penalty for frequent tokens
    • :max-output-tokens - Integer, maximum tokens to generate
    • :model-name - String, specific model to use
    • :presence-penalty - Number, penalty for presence of tokens
    • :response-format - ResponseFormat object for structured outputs, such as with json-response-format
    • :stop-sequences - Collection of strings that stop generation
    • :temperature - Number, randomness in generation (0.0-2.0)
    • :tool-choice - Keyword, tool usage strategy (:auto or :required)
    • :tools - Collection of tool info created with tool-info
    • :top-k - Integer, top-k sampling parameter
    • :top-p - Number, nucleus sampling parameter (0.0-1.0)

Returns:

  • ChatRequest - Request object for use with chat

Example:

(lc4j/chat-request
  [(UserMessage. "Calculate the area of a circle with radius 5")]
  {:tools [calculator-tool geometry-tool]
   :temperature 0.1
   :tool-choice :required
   :response-format (lc4j/json-response-format "calculation" math-schema)
   :max-output-tokens 500})

json-response-format

(json-response-format name schema)

Creates a JSON response format configuration for structured model outputs.

This function configures the model to return responses in a specific JSON format, useful for structured data extraction and API integrations.

Args:

  • name - String name for the JSON schema
  • schema - JSON schema object defining the expected response structure. The com.rpl.agent-o-rama.langchain4j.json namespace provides helpers for creating JSON schema objects.

Returns:

  • ResponseFormat - Configuration object for use in chat-request

Example:

(let [math-schema (lj/object
                    {"result" (lj/number "The calculated result")
                     "steps" (lj/array (lj/string) "Calculation steps")})
      response-format (lc4j/json-response-format "math-calc" math-schema)]
  (lc4j/chat model
    (lc4j/chat-request
      [(UserMessage. "Calculate 15 * 23")]
      {:response-format response-format})))