Cenedril Product System

Product architecture

Cenedril is the control layer.

You are the singularity

Local models, assigned agents, owned data, and visible permissions in one operational console.

Cenedril gives operators a visible console for models, agents, data stores, tool access, and research history. The core design principle is simple: keep the system inspectable, shapeable, and close to the work.

Model Rack7/7 ready Agent ProfilesAssigned models Data CollectionsOwned memory
Current Mode LOCAL RUNTIME READY

Models: 7/7 ready · Data: 19 wiki docs + 233 knowledge docs · Agent: bruce — friendly dog

Interface principles

Users should know what the assistant can see before it answers.

Every model, agent, collection, data source, and web connection should have a visible state. The operator can engage, disengage, or change context without losing the thread.

Model rack with readiness stateRuntime visibilityLocal and web source controlsCollection chooserSaved agent profiles
Interface FrameINSPECTABLE
Cenedril product interface screenshot

Capability matrix

One command center. Six durable capabilities.

Models

Add local models, foundation-model APIs, or hybrid routes. Switch task by task.

  • Local llama.cpp style runtime
  • Model readiness states
  • Task-specific model selection

Agents

Create named assistants, assign models, and preserve role-specific instructions.

  • Saved character profiles
  • Model-to-agent assignment
  • Purpose-specific behaviors

Data

Bring the knowledge that matters: documents, PDFs, folders, records, archives.

  • Local collections
  • Research archives
  • Knowledge bases

Permissions

Make data access visible so the operator can govern every research run.

  • Local data engage
  • Collection selection
  • Web access engage

Offline

Run where internet is weak, risky, or unavailable when models and data are local.

  • Field operations
  • Vehicles and basements
  • Remote and disaster zones

Proof

Keep evidence, source state, and operator choices attached to the work.

  • Sources used
  • Conversation history
  • Human confirmation gates
Agent RackASSIGN MODEL
B

bruce — friendly dog

Friendly research assistant · local model · saved profile

llama.cpp
F

Field Analyst

Low-connectivity research and checklist mode

mistral-local
P

Partner Brief

Strategic synthesis for C-suite conversations

hybrid

Multi-model agents

Different assistants can carry different tools.

A legal research agent, field operations agent, sales intelligence agent, and workshop assistant should not all need the same model, memory, or web access. Cenedril makes those choices explicit.

Offline-ready operations

Useful where cloud-first tools stop being useful.

When the relevant models and data are local, Cenedril can support research and decision assistance in places where the internet is unavailable, unreliable, expensive, or inappropriate.

Remote Locations

Run research archives and operating manuals far from stable connectivity.

Vehicles

Use a local knowledge package for mobile teams, maintenance crews, or survey routes.

Disaster Zones

Preserve plans, checklists, contacts, and procedures when networks degrade.

Basements and Facilities

Keep facility-specific documentation available where signal is weak.