Crate ferrin_google 

Ferrin provider for Google Generative AI (Gemini API).

Implements the ferrin-spec traits for generateContent (non-streaming and streaming), embeddings, Gemini image generation, speech (TTS), transcription, Veo video generation, the Files API, batch generation and Live API sessions, plus factories for the Google provider-executed tools.

Examples

use ferrin_google::GoogleSettings;
use ferrin_google::create_google;
use ferrin_spec::LanguageModel;
use ferrin_spec::language_model::CallOptions;
use ferrin_spec::language_model::PromptMessage;

let google = create_google(GoogleSettings::default())?;
let model = google.language_model("gemini-2.5-flash");
let result = model
    .do_generate(CallOptions::new(vec![PromptMessage::user_text("Hello")]))
    .await?;
println!("{:?}", result.content);

Capability matrix and provider options: docs/providers/google.md.

Attribution

Portions of this crate are derived from the Vercel AI SDK (Apache-2.0, Copyright 2023 Vercel, Inc.), translated from TypeScript to Rust and modified. See the NOTICE file in the crate root.

Re-exports

pub use crate::batch::GoogleBatch;

pub use crate::config::GoogleConfig;

pub use crate::config::SharedConfig;

pub use crate::embedding::GoogleEmbeddingModel;

pub use crate::files::GoogleFiles;

pub use crate::image::GoogleImageModel;

pub use crate::language_model::GoogleLanguageModel;

pub use crate::realtime::GoogleRealtimeFactory;

pub use crate::realtime::GoogleRealtimeModel;

pub use crate::speech::GoogleSpeechModel;

pub use crate::tools::GoogleTools;

pub use crate::transcription::GoogleTranscriptionModel;

pub use crate::video::GoogleVideoModel;

Modules

api_types

Wire types of the generateContent responses (requests are built as JSON objects in crate::request).

batch

Batch generation (batchGenerateContent): inline or file-based input, status, results, cancel and list.

capabilities

Model capabilities inferred from the model id.

config

Shared configuration of every Google model and service.

convert_prompt

Conversion of the specification prompt to systemInstruction and contents.

embedding

Gemini embedding model (embedContent / batchEmbedContents).

error

Google error response schema.

files

Files API: resumable upload with processing poll, metadata and delete.

image

Gemini image generation (generateContent with the IMAGE modality).

json_accumulator

Incremental reconstruction of function call arguments streamed as partialArgs (JSON path plus scalar value) into JSON text deltas.

json_schema

Conversion of JSON Schema (draft 7) to the OpenAPI 3.0 schema subset accepted by the Gemini API (responseSchema, function parameters).

language_model

Gemini language model (generateContent / streamGenerateContent).

options

Provider options read from provider_options["google"] (and the configured provider name when it differs).

output

Mapping of generateContent responses to specification content, usage, finish reasons and provider metadata.

prepare_tools

Conversion of tool definitions and tool choice to tools and toolConfig.

realtime

Live API realtime model: ephemeral auth tokens, session setup and the mapping between Live API messages and the specification events.

request

Request body construction (generateContent).

speech

Gemini text-to-speech model (generateContent with AUDIO modality).

stream

Streaming state machine for streamGenerateContent?alt=sse.

tools

Factories for Google provider-executed tools.

transcription

Gemini transcription model (unary, Interactions API POST /interactions).

video

Veo video generation (predictLongRunning operations).

Structs

GoogleProvider

The Google provider: a factory for the models and services.

GoogleSettings

Settings of create_google.

Constants

VERSION

Crate version.

Functions

create_google

Creates a Google Generative AI provider.