Nage Platform

Build on
SEDIM

The complete platform for source-attributed AI. Create VARVEs, build agents, serve models.

Explore Platform Request Access
VARVE Studio

Your knowledge,
layered

Upload documents, connect sources, and train domain-specific VARVEs. Each VARVE becomes a permanent, attributed knowledge layer in your SEDIM model.

Data Sources
UPLOADPDF, DOCX, CSV, TXT — drag and drop or bulk import
CONNECTNotion, Google Drive, Slack, GitHub — live sync
APIProgrammatic ingestion via POST /v1/learn endpoint
VARVE Health
Distance metrics, CKA alignment scores, and staleness alerts for every VARVE. Versioning with semver, diff view, and one-click rollback to any previous version.
Ephemeral
Instant
Free
In-context knowledge injection. No training step. Ideal for rapid prototyping and one-off queries.
Flash
10 min
Lightweight
Quick LoRA-based VARVE training. Good for small corpora and fast iteration cycles.
Market
2 hr
Production-grade
Full CONFLUX pipeline with CKA gating, SVD init, and multi-epoch training. Production-ready VARVE.
Agent Studio

STEMMA-powered
orchestration

Build agents that know where their knowledge comes from. Every action carries a STEMMA attribution vector — full transparency from plan to execution.

Visual Canvas
Drag-and-drop workflow builder
Connect VARVE nodes, action nodes, and decision gates into visual workflows. No code required. Each node displays its STEMMA contribution in real time.
P.U.R. Planner
Plan → Utilize → Report
Structured agent execution with STEMMA attribution at every stage. Plan decomposes the task, Utilize executes with VARVE-powered actions, Report delivers results with full provenance.
Task History
Full execution audit trail
Every agent run logged with per-action STEMMA vectors. Replay any execution, inspect which VARVEs contributed to each decision, and trace errors to their source.
Action VARVEs
Web
Browse, scrape, search, extract, summarize
Code
Execute, debug, review, refactor, deploy
Communication
Email, Slack, calendar, notifications
Data
Query, transform, visualize, export, sync
25+ action VARVEs across four categories
PEC Console

Full API
control

Manage keys, monitor attribution, test queries, and measure ROI — all from a single dashboard built for SEDIM-native workflows.

API Keys
STRATUM Tier Management
Create, rotate, and scope API keys by STRATUM tier. Per-key rate limits, usage tracking, and automatic alerting when approaching quota.
Analytics
STEMMA Attribution Dashboard
Real-time attribution distribution across all VARVEs. Usage trends, routing patterns, and anomaly detection for production SEDIM deployments.
Playground
VARVE Playground
Send queries and see live STEMMA visualization. Toggle individual VARVEs on and off. Compare outputs with different routing configurations.
Testing
A/B Testing & Shadow Mode
Run production traffic against VARVE variants. Shadow mode mirrors live queries without affecting users. Compare STEMMA distributions across configurations.
ROI
Cost & Revenue Dashboard
GPU compute cost per VARVE, API revenue per tier, margin analysis per endpoint. Track exactly which knowledge layers generate value.
STRATUM API

Five-tier
access system

Surface
Free
forever
1K requests / day
Single VARVE access
Basic STEMMA attribution
Community support
Drift
$19
per month
100K requests / month
Up to 3 VARVEs
Full STEMMA vectors
Flash VARVE training
Email support
Vein
$49
per month
1M requests / month
Up to 10 VARVEs
Market VARVE training
Agent Studio access
Priority support
Lode
$149
per month
10M requests / month
Unlimited VARVEs
A/B testing & shadow mode
STEMMA override API
Dedicated support
Core
Custom — Unlimited
On-premise deployment, custom SLAs, dedicated VARVE training infrastructure, and direct engineering support.
Python SDK
# pip install nage from nage import Client client = Client(api_key="nk_live_...") response = client.think("What is SEDIM?") # Every response carries attribution print(response.text) print(response.stemma) # {fehm-tr: 0.73, cortex: 0.27}
Endpoints
POST
/v1/think
Send a query, receive an attributed response with full STEMMA vector.
POST
/v1/think/stream
Streaming variant with token-level STEMMA attribution as SSE events.
GET
/v1/knowledge
List active VARVEs, their health metrics, and current routing configuration.
POST
/v1/learn
Ingest new documents or trigger VARVE training programmatically.
Early Access

Start building on
SEDIM

First 100 builders get full platform access. Create VARVEs, deploy agents, and serve models with complete source attribution from day one.

Request Early Access Read Architecture