feat: use gemini flash for image generation

This commit is contained in:
2025-10-31 13:11:04 -07:00
parent 2c9eb1e360
commit e9de3fc06c
2 changed files with 53 additions and 44 deletions
+27 -26
View File
@@ -3,18 +3,19 @@
* @license Naomi's Public License * @license Naomi's Public License
* @author Naomi Carrigan * @author Naomi Carrigan
*/ */
import Anthropic from "@anthropic-ai/sdk"; import { Anthropic } from "@anthropic-ai/sdk";
import { GoogleGenAI, PersonGeneration } from "@google/genai"; import { GoogleGenAI, } from "@google/genai";
import { AttachmentBuilder } from "discord.js"; import { AttachmentBuilder } from "discord.js";
/** /**
* * Utility class for generating project information and images.
*/ */
export class Ai { export class Ai {
anthropic; anthropic;
gemini; gemini;
/** /**
* @param anthropicKey * Creates a new instance of the Ai class.
* @param geminiKey * @param anthropicKey - The API key for the Anthropic API.
* @param geminiKey - The API key for the Gemini API.
*/ */
constructor(anthropicKey, geminiKey) { constructor(anthropicKey, geminiKey) {
this.anthropic = new Anthropic({ this.anthropic = new Anthropic({
@@ -25,7 +26,9 @@ export class Ai {
}); });
} }
/** /**
* @param prompt * Generates a list of potential project names and a full body anime girl mascot for the project.
* @param prompt - The user's prompt for the project.
* @returns A message create options object containing the project name, description, and image.
*/ */
async generateProjectInfo(prompt) { async generateProjectInfo(prompt) {
const projectRequest = await fetch("https://data.nhcarrigan.com/projects.json"); const projectRequest = await fetch("https://data.nhcarrigan.com/projects.json");
@@ -37,19 +40,20 @@ export class Ai {
return p.name; return p.name;
}). }).
join(", ")}`, prompt); join(", ")}`, prompt);
const image = await this.generateImage(`Your task is to generate a full body anime girl mascot for this project. The image should have a transparent background. Potential names: ${names}. The project description is: ${prompt}`); const image = await this.generateImage(prompt);
if (image === null) { if (image === null) {
return { content: `Project Name: ${names}\nProject Description: ${prompt}\nSorry, I was unable to generate an image for you.` }; return {
content: `Project Name: ${names}\nProject Description: ${prompt}\nSorry, I was unable to generate an image for you.`,
};
} }
return { content: `Project Name: ${names}\nProject Description: ${prompt}`, return {
files: [new AttachmentBuilder(image, { name: "avatar.png" })] }; content: `Project Name: ${names}\nProject Description: ${prompt}`,
files: [new AttachmentBuilder(image, { name: "avatar.png" })],
};
} }
/**
* @param system
* @param prompt
*/
async generateText(system, prompt) { async generateText(system, prompt) {
const response = await this.anthropic.messages.create({ const response = await this.anthropic.messages.create({
// eslint-disable-next-line @typescript-eslint/naming-convention -- SDK requirement.
max_tokens: 1000, max_tokens: 1000,
messages: [ messages: [
{ {
@@ -70,22 +74,19 @@ export class Ai {
join(""); join("");
return text; return text;
} }
/**
* @param prompt
*/
async generateImage(prompt) { async generateImage(prompt) {
const response = await this.gemini.models.generateImages({ const response = await this.gemini.models.generateContent({
config: { config: {
aspectRatio: "3:4", imageConfig: { aspectRatio: "3:4" },
imageSize: "2K", systemInstruction: "Your task is to generate a full body anime girl mascot for this project. This means the full character should be visible. The image should have a white background. NEVER include text in the image, no text anywhere at all. The project description is provided by the user.",
numberOfImages: 1,
outputMimeType: "image/png",
personGeneration: PersonGeneration.ALLOW_ADULT,
}, },
model: "models/imagen-4.0-generate-001", contents: prompt,
prompt: prompt, model: "gemini-2.5-flash-image",
}); });
const base64 = response.generatedImages?.[0]?.image?.imageBytes; const image = response.candidates?.[0]?.content?.parts?.find((p) => {
return Boolean(p.inlineData);
});
const base64 = image?.inlineData?.data;
if (base64 === undefined) { if (base64 === undefined) {
return null; return null;
} }
+26 -18
View File
@@ -5,7 +5,9 @@
*/ */
import { Anthropic } from "@anthropic-ai/sdk"; import { Anthropic } from "@anthropic-ai/sdk";
import { GoogleGenAI, PersonGeneration } from "@google/genai"; import {
GoogleGenAI,
} from "@google/genai";
import { AttachmentBuilder, type MessageCreateOptions } from "discord.js"; import { AttachmentBuilder, type MessageCreateOptions } from "discord.js";
/** /**
@@ -34,14 +36,15 @@ export class Ai {
* @param prompt - The user's prompt for the project. * @param prompt - The user's prompt for the project.
* @returns A message create options object containing the project name, description, and image. * @returns A message create options object containing the project name, description, and image.
*/ */
public async generateProjectInfo(prompt: string): public async generateProjectInfo(
Promise<MessageCreateOptions> { prompt: string,
): Promise<MessageCreateOptions> {
const projectRequest = await fetch( const projectRequest = await fetch(
"https://data.nhcarrigan.com/projects.json", "https://data.nhcarrigan.com/projects.json",
); );
const projectResponse const projectResponse
// eslint-disable-next-line @typescript-eslint/consistent-type-assertions -- Fetch does not accept a generic. // eslint-disable-next-line @typescript-eslint/consistent-type-assertions -- Fetch does not accept a generic.
= (await projectRequest.json()) as Array<{ name: string }>; = (await projectRequest.json()) as Array<{ name: string }>;
const names = await this.generateText( const names = await this.generateText(
`Your task is to generate a project name based on the user's description. Provide ONLY a list of 1-5 fitting names, and an explanation for why you chose them. Note that project names should be unique. Here's a list of all existing project names: ${projectResponse. `Your task is to generate a project name based on the user's description. Provide ONLY a list of 1-5 fitting names, and an explanation for why you chose them. Note that project names should be unique. Here's a list of all existing project names: ${projectResponse.
map((p) => { map((p) => {
@@ -50,12 +53,16 @@ export class Ai {
join(", ")}`, join(", ")}`,
prompt, prompt,
); );
const image = await this.generateImage(`Your task is to generate a full body anime girl mascot for this project. The image should have a transparent background. NEVER include text in the image. The project description is: ${prompt}`); const image = await this.generateImage(prompt);
if (image === null) { if (image === null) {
return { content: `Project Name: ${names}\nProject Description: ${prompt}\nSorry, I was unable to generate an image for you.` }; return {
content: `Project Name: ${names}\nProject Description: ${prompt}\nSorry, I was unable to generate an image for you.`,
};
} }
return { content: `Project Name: ${names}\nProject Description: ${prompt}`, return {
files: [ new AttachmentBuilder(image, { name: "avatar.png" }) ] }; content: `Project Name: ${names}\nProject Description: ${prompt}`,
files: [ new AttachmentBuilder(image, { name: "avatar.png" }) ],
};
} }
private async generateText(system: string, prompt: string): Promise<string> { private async generateText(system: string, prompt: string): Promise<string> {
@@ -83,18 +90,19 @@ export class Ai {
} }
private async generateImage(prompt: string): Promise<Buffer | null> { private async generateImage(prompt: string): Promise<Buffer | null> {
const response = await this.gemini.models.generateImages({ const response = await this.gemini.models.generateContent({
config: { config: {
aspectRatio: "3:4", imageConfig: { aspectRatio: "3:4" },
imageSize: "2K", systemInstruction:
numberOfImages: 1, `Your task is to generate a full body anime girl mascot for this project. This means the full character should be visible. The image should have a white background. NEVER include text in the image, no text anywhere at all. The project description is provided by the user.`,
outputMimeType: "image/png",
personGeneration: PersonGeneration.ALLOW_ADULT,
}, },
model: "models/imagen-4.0-generate-001", contents: prompt,
prompt: prompt, model: "gemini-2.5-flash-image",
}); });
const base64 = response.generatedImages?.[0]?.image?.imageBytes; const image = response.candidates?.[0]?.content?.parts?.find((p) => {
return Boolean(p.inlineData);
});
const base64 = image?.inlineData?.data;
if (base64 === undefined) { if (base64 === undefined) {
return null; return null;
} }