LLM Configuration: Default Properties

The following table explains about LLM properties that populates automatically based on the selected provider.

LLM	Property	Value
OpenAI	apiKey	Specify the API key to authenticate access to the OpenAI LLM provider. A secure credential used to authenticate and authorize requests sent to the OpenAI LLM provider. The API key uniquely identifies your application, enables access to OpenAI services, and is required for all API calls. It must be kept confidential and should be stored securely using environment variables or a secret manager.
OpenAI	temperature	Specify the value by which you can control the deterministic or creative nature of the generated text. The temperature property controls the sampling temperature, which influences how creative or deterministic the generated responses are. Higher values make the output more random and diverse, while lower values make it more focused and predictable. Note: Modifying both temperature and topP simultaneously is not recommended, as their interaction can produce unpredictable effects. You can provide the value between 0 to 1.
OpenAI	maxTokens	Specify the maximum number of tokens to generate in the chat completion. The total length of input tokens and generated tokens is limited by the model’s context length. Use for non-reasoning models (for example, gpt-4o, gpt-3.5-turbo). Cannot be used with reasoning models (for example, o1, o3, o4-mini series). Mutually exclusive with maxCompletionTokens - setting both will result in an API error.
OpenAI	responseFormat	Forces the model to always return a valid JSON object and enforces a strict structured response based on the defined schema (schema validation). Format: { "type": "json_object" } Example: { "type": "json_schema", "json_schema": { "name": "user_schema", "schema": { "type": "object", "properties": { "name": { "type": "string" }, "age": { "type": "number" }, "city": { "type": "string" } }, "required": [ "name", "age" ] } } }
OpenAI	retryAttempts	Specify the number of retry attempts for runtime errors. Number of retry attempts for failed LLM requests This setting improves reliability during transient issues, such as network interruptions or temporary service unavailability. When the retry limit is reached, the request fails and returns an exception. Default value: 1 Note: You can try retryAttempts setting with LLM steps that are not useful with other steps
OpenAI	reasoningEffort	reasoningEffort defines how much thinking a reasoning-capable OpenAI model uses to generate a response. Controls how much reasoning the model applies when generating a response. Higher values improve accuracy for complex or multi-step tasks, while lower values increase response speed. Settings for reasoningEffort are: - Low: Uses minimal reasoning; suitable for simple or direct queries. - Medium: Uses balanced reasoning; ideal for most general-purpose tasks. - High: Uses deep reasoning; best for complex logic, analysis, or multi-step problems.
Azure OpenAI	apiKey	Specify the API key to authenticate access to the Azure OpenAI LLM provider. apiKey is the secure credential for authenticating with Azure OpenAI. A secure credential used to authenticate requests to the Azure OpenAI service. This key is obtained from the Keys and Endpoint section of your Azure OpenAI resource in the Azure Portal and must be kept confidential. It enables access to the deployed LLM models and ensures that only authorized applications can invoke the service. This property is mandatory.
Azure OpenAI	deploymentName	Identifies the specific model deployment to target. The identifier for the specific model deployment you want to use within Azure OpenAI. In Azure, models are deployed as named resources, and this value tells the client exactly which deployment to target for generating completions. The deployment name is distinct from the model name itself and must match the one configured in the Azure portal. This property is mandatory.
Azure OpenAI	projectId	An optional identifier to specify which project within your Azure OpenAI usage context should be associated with API requests. When provided, usage is counted against that project’s quota and billing. If not set, the default project context is used instead. This property is mandatory.
Azure OpenAI	temperature	Specify the value by which you can control the deterministic or creative nature of the generated text. The temperature property controls the sampling temperature, which influences how creative or deterministic the generated responses are. Higher values make the output more random and diverse, while lower values make it more focused and predictable. Note: Modifying both temperature and topP simultaneously is not recommended, as their interaction can produce unpredictable effects.
Azure OpenAI	maxTokens	The maximum number of tokens to generate in the chat completion. The total length of input tokens and generated tokens is limited by the model’s context length. Use for non-reasoning models (for example, gpt-4o and gpt-3.5-turbo). Cannot be used with maxCompletionTokens.
Azure OpenAI	responseFormat	Specifies structured output type and schema. Forces the model to always return a valid JSON object and enforces a strict structured response based on the defined schema (schema validation). Syntax: { "type": "json_object" } Example: { "type": "json_schema", "json_schema": { "name": "user_schema", "schema": { "type": "object", "properties": { "name": { "type": "string" }, "age": { "type": "number" }, "city": { "type": "string" } }, "required": [ "name", "age" ] } } }
Azure OpenAI	reasoningEffort	reasoningEffort defines how much thinking a reasoning-capable OpenAI model uses to generate a response. Settings for reasoningEffort are: - Low: Uses minimal reasoning; suitable for simple or direct queries. - Medium: Uses balanced reasoning; ideal for most general-purpose tasks. - High: Uses deep reasoning; best for complex logic, analysis, or multi-step problems.
Google Vertex AI	credentials	URI for Google Cloud credentials to authenticate with Vertex AI. Specifies the URI to the Google Cloud credentials used to authenticate with the Vertex AI service. When provided, the credentials URI is used by Spring AI to build a GoogleCredentials instance that authorizes requests to Vertex AI. This typically points to a Google service account JSON file with appropriate permissions. This property is mandatory.
Google Vertex AI	projectId	The Google Cloud project under which the Vertex AI model operates. Defines the Google Cloud project identifier under which Vertex AI resources are managed and billed. This property must match the project where your Vertex AI deployments are configured, and it ensures that generated requests are scoped to the correct Google Cloud environment. This property is mandatory.
Google Vertex AI	temperature	Specify the value by which you can control the randomness/creativity of model responses. Temperature controls the randomness of the output. Values range from 0.0 to 1.0. Higher values produce more varied responses, while lower values produce more predictable responses. This value is used as the default by the backend when making the request.
Google Vertex AI	maxOutputTokens	Maximum tokens the model can generate in a single response. The maximum number of tokens to consider when sampling. The generative uses combined Top-k and nucleus sampling. Top-k sampling considers the set of topK most probable tokens.
Google Vertex AI	responseMimeType	Specifies how the model returns its response. responseMimeType defines the format of the model’s output. It helps control how the response is structured so it can be easily processed or displayed by your application. - Use text/plain to get a normal text response (default). - Use application/json to get the response in JSON format for easy parsing and structured processing.
Google Gemini	credentials	Authenticates access to the Gemini (Vertex AI) service. These credentials usually point to a service account JSON file that has permission to access Vertex AI. Proper credentials are required for all API requests. This property is mandatory.
Google Gemini	temperature	Specify the value by which you can control the randomness/creativity of model responses. Lower values produce more focused and predictable output, while higher values generate more creative and varied responses. Adjust this setting based on whether accuracy or creativity is more important.
Google Gemini	maxOutputTokens	Defines the maximum number of tokens the model can generate in a single response. This setting limits the response length and helps control usage and cost. Once the limit is reached, the model stops generating additional output.
Google Gemini	responseFormat	Forces the model to always return a valid JSON object and enforces a strict structured response based on the defined schema (schema validation). Syntax: { "type": "json_object" } Example: { "type": "json_schema", "json_schema": { "name": "user_schema", "schema": { "type": "object", "properties": { "name": { "type": "string" }, "age": { "type": "number" }, "city": { "type": "string" } }, "required": [ "name", "age" ] } } }
AWS Bedrock	accessKey	Identifies the AWS account or IAM user. Specifies the AWS access key ID used to authenticate requests to Amazon Bedrock. This key identifies the AWS account or IAM user making the request and must have the required permissions to invoke Bedrock models. This property is mandatory.
AWS Bedrock	secretKey	Authenticates and signs AWS Bedrock requests. Specifies the AWS secret access key associated with the access key ID. This key is used to securely sign API requests to AWS Bedrock and must be kept confidential at all times. This property is mandatory.
AWS Bedrock	region	AWS region used to access Bedrock. Specifies the AWS region where Amazon Bedrock is accessed (for example, us-east-1). This setting determines which regional Bedrock endpoint is used and must match a region where the selected model is available. This property is mandatory.
AWS Bedrock	temperature	Controls how random or creative the model’s responses are. Lower values produce more consistent and focused output, while higher values generate more creative and varied responses. Choose the value based on whether accuracy or creativity is more important for your use case.
AWS Bedrock	maxTokens	Specify the maximum number of tokens (units of text) that a model can generate in a single request. Note: If the max token limit is exceeded then you may see truncation of the input or an error.