BETA

MMarkellos Ecosystem
AI LabPrivate Local Workspace

CONNECTIONS

Connections are optional. Local agents remain on Ollama unless you explicitly use one of these controls.

MARKELLOS ECOSYSTEM · LOCAL LAB

Learning Lab

Learn → Practice → Evaluate → Improve

Back to markellosecosystem LOCAL ONLY

Appearance Settings

Appearance

Dark Light

Accent Color

Background Color

Text Font

Important Info

This laboratory runs locally with Ollama. Conversation history stays in local SQLite storage.

No analytics or external AI service is integrated into this app.

© Markellos Markides. All rights reserved.

ZIP version: v2.0.16

MODEL ADVISOR · PHASE 1

Model Guidance

General, vendor-neutral starting points for choosing an installed local model. Guidance does not change your selection.

Recommended Model Size
Required Capabilities

Why this recommendation?

Hardware considerations

Informational profiles

Hardware guidance

System RAM, available GPU VRAM, CPU performance, quantization, context length, and the individual model's memory requirements all affect compatibility and speed. Parameter count alone cannot determine memory needs or execution performance.

Hardware Compatibility: Not Verified

Recommended model sizes are indicative starting points, not strict requirements. Larger models are not necessarily better. Model architecture, specialization, quantization and task-specific performance also influence suitability.

Evaluate a completed Training Room session

Only completed session records are evaluated. Original agent responses remain unchanged. Scores are qualified model assessments.

Completed session Local Evaluator model Criteria (one per line, up to eight)

Evaluation runs

Selected assessment

Select a run to inspect scores, findings, and provenance.

Training Topic

Agents have no access to files or commands.

No training sessions yet. Enter a topic above.

© Markellos Markides. All rights reserved.ZIP version: v2.0.16

WORKSPACE

Rooms & agents

◉ Checking local model...

Inference Engine: Ollama

Add Model · Ollama

Ollama model name

Downloads from the internet into Ollama's local model store. Check available disk space and model licensing before installing. Model files may be several GB.

Model installation is optional.

Advanced connectionsPersonal provider and ChatGPT advisor Optional

Personal Fine-Tuning ProviderYour provider account

Connect and review your own provider account. The provider bills you directly; this Lab does not hold a shared GPU account or payment method.

RunPod API tokens and SSH public keys are credentials that only enable a direct connection to your own RunPod account or an already-running server. By themselves, they cannot create a server, start a job, or charge your account. Use the official RunPod Console to connect to and manage your own RunPod servers; the Lab cannot create a provider resource or charge your account.

Open RunPod Console ↗

OpenAI API keys are not accepted, stored, or used by this Lab. Instead, you may optionally connect with ChatGPT through your browser. This user-approved connection does not use an API key. Connecting alone does not send prompts, make an AI request, or create a charge.

Open OpenAI API Platform ↗

ChatGPT Advisor

Connect your ChatGPT account to ask a question. Connecting does not send a prompt or invoke a model.

Checking local ChatGPT connection status…

ChatGPT Advisor

Connected

Only the question you submit is sent to ChatGPT. Local rooms, histories, files and agents stay local.

Your Question

ChatGPT Answer

Local Advisor history

Saved on this PC only. It is never sent automatically to ChatGPT.

Loading local Advisor history…

Provider Region preference Personal spending currency Your maximum spending limit

This is your planning limit in the currency selected above. RunPod's listed GPU prices are shown in USD; no currency conversion is performed.

Maximum runtime (minutes) I will use my own provider account and accept its charges. I understand selected training data would go directly to my provider, not to a Markellos server.

Set a plan, then connect your own provider account. No paid job is started automatically.

Optional: save a RunPod access token for this Windows user

Your token is encrypted locally with Windows protection. It is never shown again by the Lab. It is used only when you explicitly check your own provider connection; saving it never starts a paid job.

RunPod access token

Use only a token from your own RunPod account. Do not use a password, SSH private key, or a shared token.

Store this token encrypted for my current Windows user.

Checking local credential status…

Check your RunPod account

Checks your encrypted token and reads current GPU availability and listed prices from your RunPod account. It cannot create a Pod, upload data, or charge your account.

No connection check has been run.

What is Fine-Tuning? A simple guide ?

OPTIONAL CONNECTIONS

Internet connections & fine-tuning

OPTIONAL CONNECTION

ChatGPT Advisor

OPTIONAL CONNECTION

Personal provider