import OpenAI from "openai";
// Configuration
const OPENROUTER_API_KEY = process.env.OPENROUTER_API_KEY;
const EMBEDDING_MODEL =
process.env.QMD_EMBEDDING_MODEL || "openai/text-embedding-3-small";
// Validate API key
if (!OPENROUTER_API_KEY) {
console.error(
"Warning: OPENROUTER_API_KEY not set. Embeddings will not be generated."
);
}
// OpenRouter client (OpenAI-compatible API)
const client = OPENROUTER_API_KEY
? new OpenAI({
baseURL: "https://openrouter.ai/api/v1",
apiKey: OPENROUTER_API_KEY,
defaultHeaders: {
"HTTP-Referer": "https://github.com/qmd",
"X-Title": "QMD Knowledge Base",
},
})
: null;
// Embedding dimensions by model
const MODEL_DIMENSIONS: Record<string, number> = {
"openai/text-embedding-3-small": 1536,
"openai/text-embedding-3-large": 3072,
"cohere/embed-english-v3.0": 1024,
"cohere/embed-multilingual-v3.0": 1024,
};
export function getEmbeddingDimensions(): number {
return MODEL_DIMENSIONS[EMBEDDING_MODEL] || 1536;
}
export function isEmbeddingsEnabled(): boolean {
return !!client;
}
/**
* Generate embeddings for one or more texts
* @param texts Array of text strings to embed
* @returns Array of embedding vectors (number arrays)
*/
export async function embed(texts: string[]): Promise<number[][]> {
if (!client) {
throw new Error("OPENROUTER_API_KEY not configured");
}
if (texts.length === 0) {
return [];
}
// OpenRouter/OpenAI has a limit on batch size
const BATCH_SIZE = 100;
const results: number[][] = [];
for (let i = 0; i < texts.length; i += BATCH_SIZE) {
const batch = texts.slice(i, i + BATCH_SIZE);
try {
const response = await client.embeddings.create({
model: EMBEDDING_MODEL,
input: batch,
});
// Sort by index to maintain order
const sorted = response.data.sort((a, b) => a.index - b.index);
results.push(...sorted.map((d) => d.embedding));
} catch (error) {
console.error(`Error generating embeddings for batch ${i}:`, error);
throw error;
}
}
return results;
}
/**
* Generate embedding for a single text
* @param text Text string to embed
* @returns Embedding vector
*/
export async function embedSingle(text: string): Promise<number[]> {
const [embedding] = await embed([text]);
return embedding;
}
/**
* Compute cosine similarity between two vectors
*/
export function cosineSimilarity(a: number[], b: number[]): number {
if (a.length !== b.length) {
throw new Error(
`Vector dimension mismatch: ${a.length} vs ${b.length}`
);
}
let dot = 0;
let normA = 0;
let normB = 0;
for (let i = 0; i < a.length; i++) {
dot += a[i] * b[i];
normA += a[i] * a[i];
normB += b[i] * b[i];
}
const denominator = Math.sqrt(normA) * Math.sqrt(normB);
if (denominator === 0) return 0;
return dot / denominator;
}
/**
* Get current embedding model name
*/
export function getModelName(): string {
return EMBEDDING_MODEL;
}
