[ACTION] Update Google Gemini actions

[dannyroosevelt]

General Improvements

Let's make sure the models are up to date (can we list them dynamically?)

Actions improvement

Action: Chat

Uses the generateContent endpoint with conversation history.

Props:

Prop	Type	Description	API Mapping
model	string/options/async options	The model to use.
prompt	string	The user's input message for the current turn.	contents.parts.text
history	array	The preceding conversation history (user and model turns).	contents (previous items)
systemMessage	string	Optional. Instructions that guide the model's behavior throughout the conversation.	system_instruction
temperature	number	Optional. Controls randomness (0.0 deterministic, 1.0+ creative). Default varies by model.	generationConfig.temperature
maxOutputTokens	integer	Optional. Maximum number of tokens to generate in the response.	generationConfig.maxOutputTokens
topK	integer	Optional. Selects tokens from the top K most probable.	generationConfig.topK
topP	number	Optional. Selects tokens cumulatively reaching probability P.	generationConfig.topP
safetySettings	array	Optional. Configuration for blocking potentially harmful content.	safetySettings
outputSchema	object	Optional. JSON schema defining the structure of the desired output.	generationConfig.response_schema

API Document: https://ai.google.dev/gemini-api/docs/text-generation#chat

---

Action: Create Embeddings

Uses the embedContent or batchEmbedContents endpoints.

Props:

Prop	Type	Description	API Mapping
content	string	The text content to generate embeddings for.	content.parts.text (or in requests)
model	string	The embedding model to use (e.g., "models/embedding-001").	Part of the endpoint URL
taskType	string	Optional. The intended use case (e.g., "RETRIEVAL_QUERY", "RETRIEVAL_DOCUMENT"). Affects embedding quality.	taskType

API Document: https://ai.google.dev/gemini-api/docs/embeddings

---

Action: Summarize

Uses the generateContent endpoint with a specific instruction prompt.

Props:

Prop	Type	Description	API Mapping
model	string/options/async options	The model to use.
content	string	The text content to summarize (provided within the prompt).	contents.parts.text
instruction	string	The prompt instructing the model to summarize the content.	contents.parts.text
maxOutputTokens	integer	Optional. Maximum number of tokens for the generated summary.	generationConfig.maxOutputTokens
temperature	number	Optional. Controls creativity/factuality in the summary (0.0 more factual).	generationConfig.temperature

API Document: https://ai.google.dev/gemini-api/docs/text-generation

---

Action: Translate Text

Uses the generateContent endpoint with a specific instruction prompt.

Props:

Prop	Type	Description	API Mapping
model	string/options/async options	The model to use.
content	string	The text content to translate (provided within the prompt).	contents.parts.text
instruction	string	The prompt instructing the model to translate the content, specifying target language (and optionally source).	contents.parts.text
temperature	number	Optional. Controls translation style (0.0 more literal).	generationConfig.temperature

API Document: https://ai.google.dev/gemini-api/docs/text-generation

---

Action: Classify Items Into Categories

Uses the generateContent endpoint with a specific instruction prompt. Use Output Schema to produce the output in object as follow:

{"category1": ["item1", "item2"], "category2": ["item3", "item4"]}

Props:

Prop	Type	Description	API Mapping
model	string/options/async options	The model to use.
items	array	List of items to be classified (provided within the prompt).	contents.parts.text
categories	array	List of available categories (provided within the prompt).	contents.parts.text
instruction	string	The prompt describing the classification task, items, and categories.	contents.parts.text
temperature	number	Optional. Controls randomness in classification decisions (0.0 more deterministic).	generationConfig.temperature

API Document: https://ai.google.dev/gemini-api/docs/text-generation

---

Action: Chat with Image/Video

Uses the generateContent endpoint with multimodal input (image/video + text prompt).

Props:

Prop	Type	Description	API Mapping
model	string/options/async options	The model to use.
filePath	object	File path in tmp dir to image/video
prompt	string	Instructions .
outputSchema	object	Optional. JSON schema defining the structure of the desired output.	generationConfig.response_schema

API Document: https://ai.google.dev/gemini-api/docs/vision

---

Action: Chat with Audio

Uses the generateContent endpoint with multimodal input (audio + text prompt).

Props:

Prop	Type	Description	API Mapping
model	string/options/async options	The model to use.
filePath	object	File path in tmp dir to the audio
prompt	string	Instructions for transcription (e.g., "Transcribe this audio."). Can specify language if needed.	contents.parts.text
outputSchema	object	Optional. JSON schema defining the structure of the desired output.	generationConfig.response_schema

API Document: https://ai.google.dev/gemini-api/docs/audio