Developer Platform

Build across 40+ languages, zero-shot

REST APIs, Python and Node SDKs. Drop-in replacement for Voyage, Cohere, or OpenAI embeddings.

Quick Start

1. Get your API key

2. Install the SDK (optional)

Terminal

pip install bhala
# or
npm install @bhala/sdk

3. Make your first API call

embed.py

from bhala import Bhala

client = Bhala(api_key="bh_sk_...")

# Generate embeddings for any supported language
response = client.embed(
    input="Sawubona, ngingakusiza kanjani?",
    model="bantu-embed-v1",
    dimensions=1024
)

print(response.data[0].embedding[:5])
# [0.0234, -0.0891, 0.1456, 0.0312, -0.0678]

POST /v1/embed

curl -X POST https://api.bhala.ai/v1/embed \
  -H "Authorization: Bearer bh_sk_..." \
  -H "Content-Type: application/json" \
  -d '{
    "input": "Sawubona, ngingakusiza kanjani?",
    "model": "bantu-embed-v1",
    "dimensions": 1024
  }'

Response

{
  "object": "embedding",
  "data": [{
    "embedding": [0.0234, -0.0891, 0.1456, ...],
    "dimensions": 1024
  }],
  "usage": { "total_tokens": 8 }
}

bhala_demo.py

from bhala import Bhala

client = Bhala(api_key="bh_sk_...")

# Generate embeddings for any supported language
embeddings = client.embed(
    input="Sawubona, ngingakusiza kanjani?",
    model="bantu-embed-v1"
)
print(len(embeddings.data[0].embedding))
# → 1024

# Translate English → IsiZulu
result = client.translate(
    text="Your payment was successful.",
    source="en",
    target="zu"
)
print(result.translation)
# → "Inkokhelo yakho iphumelele."

Steer any embedding. Get a signed receipt.

Apply named actions — remove bias, redirect intent, shift sentiment — and get back the shifted vector plus a cryptographic audit record. Every intervention reversible.

Read the docs

POST /v1/embeddings/shift

curl https://api.bhala.ai/v1/embeddings/shift \
  -H "Authorization: Bearer $BHALA_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "The nurse asked the doctor about his prescription.",
    "lang": "en",
    "operators": [
      { "id": "gender_bias", "alpha": -1.0 }
    ]
  }'

# → {
#   "embedding": [ ... 1024 floats ... ],
#   "operators_applied": [
#     { "id": "gender_bias", "alpha": -1.0,
#       "shift_norm": 0.218, "latency_ms": 19 }
#   ],
#   "model": "sci-v3",
#   "audit_id": "aud_01HZ4K..."
# }