Example of RAG using OpenRouter models.

rag-openrouter-example.py

# an example of RAG using OpenRouter

import numpy as np
from openai import OpenAI

# Initialize the OpenRouter client
client = OpenAI(
    base_url="https://openrouter.ai",
    api_key="your_openrouter_api_key_here",
)

# 1. Choose your models from OpenRouter
EMBEDDING_MODEL = "openai/text-embedding-3-large"  # Embedding model
LLM_MODEL = "google/gemini-2.5-flash"  # Inference LLM, can switch to LiquidAI/LFM2.5-1.2B-Instruct or Qwen/Qwen3.5-0.8B

# Mock data simulating your vector database
documents = [
    "OpenRouter supports text completions, embeddings, and reranking APIs.",
    "Python is an interpreted, high-level, general-purpose programming language.",
    "To build a RAG application, you need an embedding model and an LLM.",
]


# Function to generate embeddings using OpenRouter
def get_embedding(text, model=EMBEDDING_MODEL):
    response = client.embeddings.create(input=[text], model=model)
    return response.data[0].embedding


# 2. Store: Generate embeddings for documents (to save in your vector DB)
print("Generating document embeddings...")
doc_embeddings = [get_embedding(doc) for doc in documents]

# 3. Retrieve: Embed user query and find the closest document
query = "How do I make a RAG app on OpenRouter?"
print(f"\nUser Query: {query}")

query_embedding = get_embedding(query)


# Simple cosine similarity function (simulating vector database search)
def cosine_similarity(a, b):
    return np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b))


# Find the most relevant document
similarities = [
    cosine_similarity(query_embedding, doc_emb) for doc_emb in doc_embeddings
]
best_match_idx = np.argmax(similarities)
retrieved_context = documents[best_match_idx]

print(f"Retrieved Context: {retrieved_context}")

# 4. Inference: Pass context and query to the LLM
print(f"\nSending to LLM ({LLM_MODEL})...")
chat_completion = client.chat.completions.create(
    model=LLM_MODEL,
    messages=[
        {
            "role": "system",
            "content": "Answer the query using only the provided context.",
        },
        {
            "role": "user",
            "content": f"Context: {retrieved_context}\n\nQuery: {query}",
        },
    ],
)

print(f"\nLLM Response:\n{chat_completion.choices[0].message.content}")