import OpenAI from 'openai';
// Only load .env in Node.js environment (not in Cloudflare Workers)
if (typeof process !== 'undefined' && process.versions && process.versions.node) {
try {
const { config } = await import('dotenv');
config();
} catch (e) {
// dotenv not available in Workers, which is fine
}
}
const apiKey = process.env.OPENAI_API_KEY;
// Only create default client if env var is present
export const openai = apiKey ? new OpenAI({ apiKey }) : null;
// Function to create an OpenAI client with custom credentials
export function createOpenAIClient(credentials) {
if (!credentials || !credentials.openaiKey) {
throw new Error('Missing OpenAI API key');
}
return new OpenAI({ apiKey: credentials.openaiKey });
}
export async function generateEmbedding(text, credentials = null) {
const client = credentials ? createOpenAIClient(credentials) : openai;
if (!client) {
throw new Error('No OpenAI client available - provide credentials or set OPENAI_API_KEY');
}
const response = await client.embeddings.create({
model: 'text-embedding-3-small',
input: text
});
return response.data[0].embedding;
}
/**
* Analyze image and generate comprehensive description using GPT-4V
*/
export async function describeImage(imageUrl, credentials = null) {
const client = credentials ? createOpenAIClient(credentials) : openai;
if (!client) {
throw new Error('No OpenAI client available - provide credentials or set OPENAI_API_KEY');
}
try {
const response = await client.chat.completions.create({
model: "gpt-4o",
messages: [{
role: "user",
content: [
{
type: "text",
text: `Analyze this image and provide:
1. A one-sentence caption
2. A 2-3 sentence summary
3. A detailed description (2-3 paragraphs)
4. List of key objects/elements (comma-separated)
5. List of descriptive tags (comma-separated)
Format as JSON with keys: caption, summary, detailed, objects, tags`
},
{
type: "image_url",
image_url: { url: imageUrl }
}
]
}],
response_format: { type: "json_object" },
max_tokens: 1000
});
const result = JSON.parse(response.choices[0].message.content);
// Ensure all fields exist with defaults
return {
caption: result.caption || "Image uploaded",
summary: result.summary || result.caption || "Image uploaded",
detailed: result.detailed || result.summary || result.caption || "Image uploaded",
objects: Array.isArray(result.objects) ? result.objects :
(result.objects ? result.objects.split(',').map(s => s.trim()) : []),
tags: Array.isArray(result.tags) ? result.tags :
(result.tags ? result.tags.split(',').map(s => s.trim()) : [])
};
} catch (error) {
console.error('Error describing image:', error);
// Fallback description
return {
caption: "Image uploaded - AI description failed",
summary: "Image uploaded - AI description failed",
detailed: "An image was uploaded but could not be automatically described due to an error.",
objects: [],
tags: ["image", "upload"]
};
}
}
/**
* Create searchable chunks from image description
*/
export function createImageChunks(description) {
const chunks = [];
// Caption chunk - short, punchy description
if (description.caption) {
chunks.push({
content: description.caption,
chunk_type: 'caption'
});
}
// Summary chunk - medium description
if (description.summary && description.summary !== description.caption) {
chunks.push({
content: description.summary,
chunk_type: 'summary'
});
}
// Detailed chunk - full description
if (description.detailed && description.detailed !== description.summary) {
chunks.push({
content: description.detailed,
chunk_type: 'description'
});
}
// Objects chunk - for searching by objects in image
if (description.objects && description.objects.length > 0) {
chunks.push({
content: description.objects.join(', '),
chunk_type: 'objects'
});
}
// Tags chunk - for searching by descriptive tags
if (description.tags && description.tags.length > 0) {
chunks.push({
content: description.tags.join(', '),
chunk_type: 'tags'
});
}
return chunks;
}
