Quickstart

flinq is OpenAI-compatible. The fastest path is to take a working OpenAI integration, change the base_url and the model name, and run it. The whole integration is under ten lines.

1. Get a key

   Create a key in the flinq console. It looks like flq_... and goes in the Authorization: Bearer header. New accounts start with EUR 5 of free credit. The examples below use the placeholder flq_your_key_here against the production API at https://api.flinq.ai.

2. Send your first request

   curl

   curl https://api.flinq.ai/v1/embeddings \
     -H "Authorization: Bearer flq_your_key_here" \
     -H "Content-Type: application/json" \
     -d '{
       "model": "flinq-embed-2.0",
       "input": "Beton C25/30 XC4"
     }'

   Python (OpenAI SDK)

   from openai import OpenAI
   
   client = OpenAI(base_url="https://api.flinq.ai/v1", api_key="flq_your_key_here")
   r = client.embeddings.create(
       model="flinq-embed-2.0",
       input="Beton C25/30 XC4",
   )
   print(len(r.data[0].embedding))  # 2048

   Node (openai)

   import OpenAI from 'openai';
   
   const client = new OpenAI({
     baseURL: 'https://api.flinq.ai/v1',
     apiKey: 'flq_your_key_here',
   });
   
   const r = await client.embeddings.create({
     model: 'flinq-embed-2.0',
     input: 'Beton C25/30 XC4',
   });
   console.log(r.data[0].embedding.length); // 2048

3. Choose a dimension

   The model is Matryoshka-trained: a single forward pass produces a 2048-dim vector whose leading slice stays meaningful when truncated. Ask for dimensions of 1024, 512, 256 or 128 to trade a little accuracy for smaller vectors, a smaller index and faster nearest-neighbor search. flinq truncates and re-normalizes for you.

   r = client.embeddings.create(
       model="flinq-embed-2.0",
       input="Beton C25/30 XC4",
       dimensions=256,
   )

4. Mark your queries

   Embed your corpus as documents (the default) and your search queries with input_type: "query". A task such as lv_matching selects the query instruction the model was trained with; it is not billed.

   q = client.embeddings.create(
       model="flinq-embed-2.0",
       input="Mauerwerk Kalksandstein, d = 17,5 cm",
       extra_body={"input_type": "query", "task": "lv_matching"},
   )

Try it live

Send requests against every endpoint, with schemas and example payloads, in the interactive API reference.

Where to go next

• Models: flinq-embed-2.0, flinq-rerank-2.0 and flinq-embed-1.5-beta, their dimensions, tasks and limits.
• Embeddings: queries and documents, image embeddings, Matryoshka dimensions, and the two failure modes flinq is built to fix.
• Hybrid search: why the BM25 lexical leg matters for AEC.
• Examples: runnable recipes for LV matching, cost prediction, and RAG with pgvector.