Core
Local inference server
Runs GGUF models inside your infrastructure with OpenAI-format-compatible streaming. Configure context, threads, fast attention, and KV cache from the panel.
01 · Inference
The engine that answers.
Models, memory, and concurrency: where your node's real capacity is defined.
Core
Text and vision
Models with a vision component and multimodal projection. Accept images via URL or embedded data to read documents, screenshots, and photos.
Performance
Execution lanes
base, gpu0, gpu1, dual, and cpu lanes to distribute loads and isolate critical tasks from ordinary traffic.
Performance
Model profiles
Full lifecycle: draft, test, and publish. What one engineer experiments with becomes stable configuration for everyone else.
Performance
Multi-model stack
Saved, restorable combinations of several models, with memory estimation before loading.
Advanced
Frontier mode
Suspends the usual stack, loads a larger model, and restores configuration after idle time without cutting active inference.
02 · Voice
Listen and speak without leaving the building.
Transcription and synthesis that don't send audio to any external service.
Core
Local speech to text
In-node audio transcription, with unload on idle to return the GPU to primary tasks when unused.
Core
Text to speech
On-demand audio generation for announcements, accessibility, training content, and products that need to speak.
03 · Documents
From file to queryable text.
The foundation of almost every institutional use case: turning paper into usable knowledge.
Core
File service and OCR
Converts PDF, Office documents, images, and spreadsheets into plain text ready to query, summarize, or extract data.
Performance
Download queues and weights
Model storage management: download queues, weight verification, and control of which versions stay available.
04 · Platform
So your applications can consume the node.
Interfaces and optional modules. Everything optional is labeled as such.
Core
Inference API
HTTPS endpoints with scope-limited access keys to connect internal systems, automations, or your own products.
Optional per project
Unreal / GameDev integration
Connect the node to game and simulation dev environments for characters, assisted dialogue, and dynamic content.
Optional per project
SIP telephony
PBX integration for phone voice scenarios on the same transcription and synthesis capacity.
Optional Phase 3
Sync between nodes
Configuration and catalog coordination across several nodes in the same organization.
Hybrid
Explicit external providers
Ability to route specific workloads to external services when the organization decides and documents it.
Performance
Multilingual interface
Panel and documentation in Spanish, English, and Portuguese for diverse teams and regions.
05 · Governance and security
Who can do what—and a record of it.
Concrete controls with clear responsibilities: neither absolute promises nor a black box.
Core
System and application users
Human and service identities with roles and least privilege, managed from the node itself.
Core
Authentication and MFA
Time-based second factor and session management to reduce risk from compromised credentials.
Core
Scoped API keys
Per-integration keys, permission-limited, revocable, with usage logging.
Core
Security audit
Logs of access, configuration changes, and relevant events to reconstruct what happened.
Performance
Inference logging
History of system queries and responses for operational traceability and usage analysis.
Core
Network and firewall
Port, rule, and external exposure management from the panel, with TLS termination.
06 · Operations
See the node—don't guess.
Telemetry, services, and power: what you need to sustain the system over time.
Core
Hardware HUD
GPU and VRAM per card, power draw, CPU, memory, temperatures, disks, and service status in one view.
Core
Service management
Status and control of system processes, with health checks for each engine.
Performance
Power
Node power consumption and its correlation with workload—direct input for total cost calculation.
Advanced
Internal benchmarks
First-token latency and generation speed measured on the client's actual hardware.
Core
Offline operation
With models installed, local flows keep working without internet egress.
Core
Phased installation
Staged deployment on a clean server, with verification and training at each phase.