KERVALT / BUILDERS
Build AI that stays under your control.
One connection, three checks, and complete data control. Deploy advanced models, run guided workflows, and trace every answer back to its source.
01 / QUICK START
Quick start
Go from key to live query in under five minutes.
Run your first query
Upload documents and ask a question that connects ideas across Tafari, Griot, and Elimu.
Open connection reference02 / CLIENTS
Official clients
Consistent interfaces, familiar patterns, and full type coverage.
03 / CAPABILITIES
Core capabilities
Everything you need to build production-grade AI behind your data perimeter.
Three-way query
Send questions to Tafari, Griot, and Elimu at the same time and combine their answers with reliable source links.
Guided workflows
Connect models, tools, and agents into reliable workflows with built-in guardrails.
Model Hub
Call advanced, open, and specialized models from one place with unified monitoring.
Fast answers
Demand-based growth, reuse, and multi-region routing for fast production workloads.
Rules & guardrails
Policies, input filtering, output records, and personal-detail masking built into every request.
Private deployment
Run Kervalt in your private cloud, on-site, or fully offline with your own encryption keys.
04 / EXAMPLE
Example: a connected question
Ask a question that connects documents, people, and records. Kervalt returns a sourced answer and the path it took.
- Points to exact pages, rows, and connections
- No third-party service data leakage
- Runs entirely on EU bare-metal
from kervalt import Client
client = Client(api_key="kv_...")
response = client.query(
collection="deal_2024_q3",
question=(
"Which portfolio companies have cross-default clauses linked to the German facility agreement?"
),
engines=["tafari", "griot", "elimu"]
)
for citation in response.citations:
print(citation.source, citation.node_id)NEXT STEP
Ready to start building?
Get your keys, run the sample notebook, and deploy your first connected query today.