Google AI Studio (Gemini)

Transcribes and analyzes an audio file or recording, returning spoken content, speaker sentiment, and key topics extracted by Gemini.

Runs a multi-turn chat exchange with Gemini, maintaining conversation history so the model can build on previous context across sequential agent steps.

Parses a PDF, DOCX, or plain-text document and returns structured insights—key clauses, data points, or summaries—extracted by Gemini's document-understanding capabilities.

Produces a spoken-word audio file from a text prompt using Gemini's text-to-speech capabilities, returning an audio file ready for playback or distribution.

Creates an image from a text description using Gemini's image generation, returning a file URL you can store, embed, or pass downstream.

Produces a short video clip from a text or image prompt using Gemini's video generation, returning a file ready for review or downstream publishing.

Sends a single freeform text prompt to a Gemini model and returns the generated response—the simplest way to inject AI generation into any workflow step.

Analyzes a video file using Gemini's multimodal vision, returning scene descriptions, spoken transcripts, visual events, and any extracted structured data.

Retrieves and analyzes a YouTube video by URL using Gemini, returning transcript highlights, topic summaries, and key moments without downloading the file.

Converts text or document content into high-dimensional vector embeddings using Gemini's embedding model, ready for semantic search, similarity scoring, or clustering pipelines.

Counts the token cost of a prompt or document before sending it to Gemini, letting the agent gate expensive calls, enforce budgets, and split oversized inputs automatically.

Retrieves the current catalog of available Gemini models from the API, returning names, version identifiers, and supported capabilities so the agent can select the right model dynamically.

Uploads a file to the Gemini Files API, returning a file URI that can be referenced in subsequent prompts—enabling analysis of large documents, audio, or video without embedding raw bytes in each request.

Sends a prompt to Gemini with a defined function schema; Gemini returns a structured function-call object that the agent executes—enabling reliable JSON extraction and tool orchestration without manual parsing.

Configures Gemini's content-safety thresholds—harassment, hate speech, sexually explicit, and dangerous content—before a generation call, so the agent enforces content policies appropriate to the deployment context.

Fires when a Gemini generation request finishes and a response is ready; can be scoped to a specific model or prompt template identifier.

Fires each time Gemini produces new generated text content, enabling downstream steps to route, store, or act on the output immediately.

Fires when Gemini finishes generating an image, returning the file URL so the agent can store, publish, or route the asset through creative review.

Fires when Gemini completes audio transcription and understanding, delivering the transcript text so the agent can index, summarize, or route it.

Fires when Gemini finishes analyzing a document, returning structured extraction results the agent can write to databases, trigger follow-up actions, or surface in dashboards.

Fires when Gemini returns a set of vector embeddings, signaling that downstream similarity search, clustering, or indexing steps can proceed with the new vectors.

Fires when Gemini returns a structured function-call object, handing the agent a typed JSON payload to execute against the nominated tool or API.