Brainpolo / Embeddings Infrastructure

Local high-speed inference for smart embeddings

1024-dimensional vector embeddings served from our infrastructure. Per-paragraph caching. Intelligent compaction. Property-preserving encryption. Your data never leaves.

Sign up API reference
$pip install polyembedpypi
Python SDK
HTTP
from polyembed import PolyEmbed

client = PolyEmbed(api_key="pe_...")

vec = client.embed(
    "Classified procurement report",
    input_type="document",
    shielded=True,
)
✓ 1024 floats · shielded · cached 
POST /api/v1/embed/
{
  "text": "Classified procurement report Q4",
  "input_type": "document",
  "shielded": true
}
200 {"embedding": [0.0229, ...], "dimensions": 1024, "shielded": true}
Specifications
1024
Dimensions
<1ms
Cache hit
4,096
Max tokens
24h
Cache TTL
SAP
Shield algo
Architecture
01

Local inference

Text never touches a third-party API. No data sharing agreements. No external network calls during inference.

02

Intelligent compaction

Long documents auto-split into paragraphs, embedded individually, combined via token-weighted averaging. No truncation. No information loss.

03

Shielded mode

Property-preserving encryption (Fuchsbauer et al. 2022). Embeddings preserve similarity ordering but cannot be reversed. Per-team isolation.

04

Team-scoped access

API keys scoped to teams with per-key rate limits and expiry. Owner, admin, member roles. Full usage tracking per key.

Create an account to generate API keys and start embedding.

Create account