Google Gemini

LogicGrid talks to the Google Generative Language API directly using Gemini's REST endpoint. No Google SDK required.

• API keys (free tier available): aistudio.google.com
• Models: ai.google.dev/gemini-api/docs/models
• Pricing: ai.google.dev/pricing

Use it

There are two equivalent ways to instantiate the Gemini LLM client.

Option 1 — static factory (recommended)

using LogicGrid.Core.Llm;

var llm = LlmClientBase.Gemini(
    apiKey: Environment.GetEnvironmentVariable("GEMINI_API_KEY")!,
    model:  "gemini-1.5-flash");

Parameter	Type	Default	Notes
apiKey	string	(required)	Google Generative Language API key. Free tier available via AI Studio.
model	string	"gemini-1.5-flash"	Any Gemini model id.

Option 2 — direct construction

using LogicGrid.Core.Providers;

var llm = new GeminiClient(
    apiKey:       Environment.GetEnvironmentVariable("GEMINI_API_KEY")!,
    defaultModel: "gemini-1.5-flash");

Parameter	Type	Default	Notes
apiKey	string	(required)	Same as the factory's apiKey.
defaultModel	string	"gemini-1.5-flash"	The model used when the agent or call site doesn't override it.

The factory and the constructor produce equivalent clients. Use direct construction when you need an injected HttpClient (for retries, proxies, or testing). The caller is responsible for setting the Authorization: Bearer header.

Models currently available

Common picks at the time of writing — always verify against Google's model page for the live list:

Model	Notes
gemini-2.0-flash-exp	Latest experimental. Strong at tool calling.
gemini-1.5-pro	Strongest 1.5-series. 1M-token context window.
gemini-1.5-flash	Fast, cheap, capable.
gemini-1.5-flash-8b	Smallest, cheapest.

Long context

Gemini 1.5 Pro accepts up to 1M tokens — enough for a small codebase or a full book in a single call. Useful for one-shot summarisation where you don't want to chunk and retrieve.

Tool calling

Gemini supports native tool calling. Opt in on agents that need it:

protected override IToolCallingStrategy ToolCallingStrategy
    => new NativeToolCallingStrategy();

Gemini's tool-call protocol is similar to OpenAI's; LogicGrid translates between the two transparently.

Full example — calculator tool over Gemini

using LogicGrid.Core.Agents;
using LogicGrid.Core.Llm;
using LogicGrid.Core.Tools;
using LogicGrid.Tools.Tools;

var llm = LlmClientBase.Gemini(
    apiKey: Environment.GetEnvironmentVariable("GEMINI_API_KEY")!,
    model:  "gemini-1.5-flash");

IAgent math = new Agent<string>(
    name:         "Mathlete",
    description:  "Solves arithmetic.",
    systemPrompt: "Use the calculator tool when the user asks for a number.",
    llm:          llm,
    tools:        new ToolBase[] { new CalculatorTool() });

Console.WriteLine(
    await math.RunAsync(
        input: "What is (17 * 23) + 91?",
        ctx:   new AgentContext().WithLogging()));

Background: Tool calling strategy.

Cost tracking

GeminiClient.Pricing returns per-token rates for known models. Live rates: ai.google.dev/pricing.

Embedding models

Gemini ships its own embedding models (e.g. gemini-embedding-001) with task-type hints like RETRIEVAL_QUERY and RETRIEVAL_DOCUMENT for asymmetric retrieval. A native GeminiEmbeddingClient is planned for a future release. Until then, point OpenAiCompatibleEmbeddingClient at Gemini's OpenAI-compatible endpoint to use Gemini embeddings today.

Troubleshooting

• API key not valid — generate a new one in AI Studio.
• Quota exceeded — you've hit the free-tier rate limit. Wait a minute or upgrade.
• SAFETY blocked — Gemini refused on safety grounds. Try rephrasing or check the safety configuration.