OpenAdapter Docs

Three calls — embed, search, chat — and you have retrieval-augmented generation.

OpenAdapter gives you everything you need for RAG in a single account:

Embed the question with /v1/embeddings
Search the collection with /v1/vectors/collections/<name>/search
Chat with the retrieved context as a system message via /v1/chat/completions

Python

import requests

API_KEY = "sk-cv-..."
BASE = "https://api.openadapter.in"
headers = {"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"}

def rag_query(question, collection="docs", top_k=5):
    # 1. Embed
    vec = requests.post(f"{BASE}/v1/embeddings", json={
        "model": "qwen3-embedding-small", "input": [question]
    }, headers=headers).json()["data"][0]["embedding"]

    # 2. Search
    results = requests.post(
        f"{BASE}/v1/vectors/collections/{collection}/search",
        json={"vector": vec, "limit": top_k},
        headers=headers,
    ).json()["results"]
    context = "\n\n".join(f"[{i+1}] {r['payload']['text']}" for i, r in enumerate(results))

    # 3. Chat
    resp = requests.post(f"{BASE}/v1/chat/completions", json={
        "model": "GLM-4.7",
        "messages": [
            {"role": "system", "content": f"Answer using context:\n{context}"},
            {"role": "user", "content": question},
        ],
    }, headers=headers).json()
    return resp["choices"][0]["message"]["content"]

print(rag_query("How do I create a collection?"))

JavaScript

const API_KEY = "sk-cv-...";
const BASE = "https://api.openadapter.in";
const headers = {
  Authorization: `Bearer ${API_KEY}`,
  "Content-Type": "application/json",
};

async function ragQuery(question, collection = "docs", topK = 5) {
  const vec = (await (await fetch(`${BASE}/v1/embeddings`, {
    method: "POST",
    headers,
    body: JSON.stringify({ model: "qwen3-embedding-small", input: [question] }),
  })).json()).data[0].embedding;

  const { results } = await (await fetch(
    `${BASE}/v1/vectors/collections/${collection}/search`,
    { method: "POST", headers, body: JSON.stringify({ vector: vec, limit: topK }) },
  )).json();
  const context = results.map((r, i) => `[${i + 1}] ${r.payload.text}`).join("\n\n");

  const resp = await (await fetch(`${BASE}/v1/chat/completions`, {
    method: "POST",
    headers,
    body: JSON.stringify({
      model: "GLM-4.7",
      messages: [
        { role: "system", content: `Answer using context:\n${context}` },
        { role: "user", content: question },
      ],
    }),
  })).json();
  return resp.choices[0].message.content;
}

Tips

Use the same embedding model at ingest and query time — different dimensions / spaces won't compare.
Top-K of 3–8 is usually enough for chat-style RAG; more chunks dilute the signal.
Include payload metadata (filenames, headings) so the model can cite sources.

RAG Pattern

Python

JavaScript

Tips

On this page