vortex.client — org.clojars.lukaszkorecki/vortex 1.0.0.19

->client clj

(->client {:keys [project location]})

->config clj

(->config {:keys [system-instruction response-schema]})

Create config for content generation. Options:

• :system-instruction - string with default system instructions attached to the client in every request/conversation
• :response-schema - JSON schema for the response, intended for single-shot mode. See vortex.schema for converting Malli schemas to GenAI's format

->content clj

(->content {:keys [role text]})

context->message clj

(context->message context)

create clj

(create {:keys [project location model system-instruction response-schema]})

Create a GenAI (aka Vertex aka Gemini on Vertex) client. Required opts:

• project - your GCP project
• model - model name, as per API docs e.g gemini-2.5-flash

Optional:

• system-instruction - system instructions for the model. Set global instructions for all uses of the model
• response-schema - forces the model to reply using structured data, uses Malli schema for definition

NOTE: response schema forces the model to reply using JSON. This is required for generate-content (one-shot structured data) and also works for send-message (structured replies on every chat turn). When a schema is set, send-message parses each reply and stores the resulting map under the :text key of the model history turn - so for chat-with-schema :text is an object, not a string. See the README for details. Omit response-schema if you want plain-text chat replies. NOTE: chat doesn't support tool calling or other grounding features yet.

generate-content* clj

(generate-content* client {:keys [model config input]})

history->contents clj

(history->contents history)

Convert a persisted history vec into a java.util.List<Content>.

IGenAI clj protocol

generate-content clj

(generate-content client {:keys [input]})

One-shot generate content using the client, with pre-baked system instructions and schema

send-message clj

(send-message client {:keys [message history context]})

Send message to a chat session, requires history, chat context

invoke-chat clj

(invoke-chat client {:keys [model contents config]})

Multi-turn SDK call. Sends contents (a List<Content>) and returns the raw GenerateContentResponse.

invoke-generate clj

(invoke-generate client {:keys [model input config]})

One-shot SDK call. Sends input (a String) and returns the raw GenerateContentResponse.

send-message* clj

(send-message* client
               {:keys [model config context history message
                       structured-responses?]})