termcode/dist/providers/google.js

Superior terminal AI coding agent with enterprise-grade security, intelligent error recovery, performance monitoring, and plugin system - Advanced Claude Code alternative

import { GoogleGenerativeAI } from "@google/generative-ai";
import { KeyStore } from "../state/keystore.js";
export class GoogleProvider {
    id = "google";
    name = "Google";
    supportsTools = true;
    maxContext = 2000000;
    requiresKey = true;
    client = null;
    async getClient() {
        if (!this.client) {
            const apiKey = await KeyStore.getProviderKey("google");
            if (!apiKey) {
                throw new Error("Google API key not found. Run 'termcode /keys' to add it.");
            }
            this.client = new GoogleGenerativeAI(apiKey);
        }
        return this.client;
    }
    async chat(messages, opts) {
        const client = await this.getClient();
        const model = client.getGenerativeModel({ model: opts.model });
        // Convert messages to Google format
        const systemMessage = messages.find(m => m.role === "system")?.content;
        const conversationMessages = messages
            .filter(m => m.role !== "system")
            .map(m => ({
            role: m.role === "assistant" ? "model" : "user",
            parts: [{ text: m.content }]
        }));
        const chat = model.startChat({
            history: conversationMessages.slice(0, -1),
            systemInstruction: systemMessage,
            generationConfig: {
                temperature: opts.temperature ?? 0.7,
                maxOutputTokens: opts.maxTokens ?? 4000
            }
        });
        const lastMessage = conversationMessages[conversationMessages.length - 1];
        const result = await chat.sendMessage(lastMessage.parts[0].text);
        return result.response.text();
    }
    async embed(texts, opts) {
        const client = await this.getClient();
        const model = client.getGenerativeModel({ model: opts.model });
        const embeddings = [];
        for (const text of texts) {
            const result = await model.embedContent(text);
            embeddings.push(result.embedding.values);
        }
        return embeddings;
    }
    async listModels() {
        return [
            // Chat models
            { id: "gemini-1.5-pro", type: "chat", context: 2000000, costPer1kTokens: 0.00125 },
            { id: "gemini-1.5-flash", type: "chat", context: 1000000, costPer1kTokens: 0.000075 },
            { id: "gemini-1.0-pro", type: "chat", context: 30720, costPer1kTokens: 0.0005 },
            // Embedding models
            { id: "text-embedding-004", type: "embed", context: 2048, costPer1kTokens: 0.00001 },
            { id: "embedding-001", type: "embed", context: 2048, costPer1kTokens: 0.00001 }
        ];
    }
    async healthCheck() {
        try {
            const apiKey = await KeyStore.getProviderKey("google");
            if (!apiKey) {
                return {
                    status: "error",
                    error: "No API key configured"
                };
            }
            // Return healthy with main chat models
            const models = ["gemini-1.5-pro", "gemini-1.5-flash", "gemini-1.0-pro"];
            return {
                status: "healthy",
                models
            };
        }
        catch (error) {
            return {
                status: "error",
                error: error.message
            };
        }
    }
    estimateCost(tokens, model, type) {
        const models = {
            "gemini-1.5-pro": 0.00125,
            "gemini-1.5-flash": 0.000075,
            "gemini-1.0-pro": 0.0005,
            "text-embedding-004": 0.00001,
            "embedding-001": 0.00001
        };
        const costPer1k = models[model] || 0.001; // Default fallback
        return (tokens / 1000) * costPer1k;
    }
}
export const googleProvider = new GoogleProvider();