The chatbot knowledge runtime for multi-tenant AI — multi-signal HyperSearch, fail-closed namespaces, BYOK (Claude/OpenAI), free Ollama path. Built for Aria-class concierges and CS bots.
# Search in Spanish, find English results curl -X POST https://fluxvector.dev/v1/vectors/query \ -H "Authorization: Bearer $FV_KEY" \ -H "Content-Type: application/json" \ -d '{"collection": "products", "text": "zapatos comodos para correr"}' # Response {"results": [ {"id": "1", "text": "Trail running shoes", "score": 0.91}, {"id": "2", "text": "Waterproof hiking boots", "score": 0.84} ], "took_ms": 42}
Create a collection, upsert text, query. Embeddings happen server-side. No OpenAI key. No ML pipeline.
# Store English text curl -X POST https://fluxvector.dev/v1/vectors/upsert \ -H "Authorization: Bearer $FV_KEY" \ -H "Content-Type: application/json" \ -d '{ "collection": "products", "vectors": [ {"id": "1", "text": "Trail running shoes"}, {"id": "2", "text": "Waterproof hiking boots"}, {"id": "3", "text": "Casual leather sandals"} ] }' # Search in Spanish curl -X POST https://fluxvector.dev/v1/vectors/query \ -H "Authorization: Bearer $FV_KEY" \ -H "Content-Type: application/json" \ -d '{"collection": "products", "text": "zapatos comodos para correr"}'
{"upserted": 3}
// Embeddings generated server-side
// No OpenAI key needed
{"results": [
{"id": "1",
"text": "Trail running shoes",
"score": 0.91},
{"id": "2",
"text": "Waterproof hiking boots",
"score": 0.84}
], "took_ms": 42}
import requests BASE = "https://fluxvector.dev/v1" H = {"Authorization": f"Bearer {KEY}"} # Store English text requests.post(f"{BASE}/vectors/upsert", json={ "collection": "products", "vectors": [ {"id": "1", "text": "Trail running shoes"}, {"id": "2", "text": "Waterproof hiking boots"}, {"id": "3", "text": "Casual leather sandals"}, ] }, headers=H) # Search in Spanish r = requests.post(f"{BASE}/vectors/query", json={ "collection": "products", "text": "zapatos comodos para correr", }, headers=H) print(r.json())
// Embeddings generated server-side // No OpenAI key needed {"upserted": 3} {"results": [ {"id": "1", "text": "Trail running shoes", "score": 0.91}, {"id": "2", "text": "Waterproof hiking boots", "score": 0.84} ], "took_ms": 42}
const BASE = "https://fluxvector.dev/v1"; const h = { Authorization: `Bearer ${KEY}` }; // Store English text await fetch(`${BASE}/vectors/upsert`, { method: "POST", headers: h, body: JSON.stringify({ collection: "products", vectors: [ { id: "1", text: "Trail running shoes" }, { id: "2", text: "Waterproof hiking boots" }, { id: "3", text: "Casual leather sandals" }, ], }), }); // Search in Spanish const r = await fetch(`${BASE}/vectors/query`, { method: "POST", headers: h, body: JSON.stringify({ collection: "products", text: "zapatos comodos para correr", }), }); const data = await r.json();
// Embeddings generated server-side // No OpenAI key needed {"upserted": 3} {"results": [ {"id": "1", "text": "Trail running shoes", "score": 0.91}, {"id": "2", "text": "Waterproof hiking boots", "score": 0.84} ], "took_ms": 42}
Stored in English. Searched in Spanish. Found the right result. No configuration.
Semantic search without the complexity, the surprise bills, or the vendor lock-in.
Send text, not vectors. We handle embedding server-side with multiple models optimized for quality. No OpenAI key needed.
Hybrid (vector + keyword) by default. "mode": "fusion" fuses up to 7 signals with re-ranking. Anti-hallucination confidence scoring built in.
Store in Spanish, query in English. Multilingual model handles cross-lingual retrieval natively.
Monthly fee, that's it. Not per-query, per-embedding, per-dimension. Know what you pay before you start.
Runs on our servers in Germany. Embeddings and search never leave them. Your data is never used to train models.
Always warm. No serverless spin-up delays. Sub-second responses from the first query.
No asterisks.
| FluxVector | Pinecone | Qdrant | |
|---|---|---|---|
| Built-in embeddings | Included free | Add-on (extra cost) | No |
| Hybrid BM25 + vector | multi-signal HyperSearch | Basic sparse | Yes |
| Self-hosted | No (hosted only) | No | Yes |
| Cold starts | None | Seconds (serverless) | None |
| 100+ languages | Native | Depends on your model | Depends on your model |
| Pricing model | Flat monthly | Per-query + storage | Free cloud + paid tiers |
| Developer console | Built-in | Yes | Yes |
| Free tier | 10K vectors, no expiry | Limited (single index) | 1GB free cloud |
Start free. Upgrade when you need to.
Every plan: email support at [email protected]. FluxVector is a hosted service.
Pro and Scale support Bring Your Own Key (BYOK) — vault or request-scoped Claude, OpenAI, Groq, DeepSeek, Moonshot. We retrieve; your model phrases. Free plan uses local Ollama. LLM keys stay in the server-side vault; the browser widget only carries a read-only widget key.
curl -X POST https://fluxvector.dev/v1/vectors/query \ -H "Authorization: Bearer $FV_KEY" \ -H "Content-Type: application/json" \ -d '{"collection": "docs", "text": "your first query"}'
Get your API key in 10 seconds. No credit card.
Get API Key