ValOs

This page explains how the agent uses different model layers to observe the market, reason over signals, and turn decisions into actions. The goal is speed, structure, and clear execution.

Runtime Stack
01

Input layer

Telegram commands, wallet activity, and market data feed the agent in real time.

02

Model layer

Models rank signals, evaluate scenarios, and decide whether conditions justify a move.

03

Execution layer

The agent turns approved decisions into actions, monitoring the position after entry.

Three model jobs inside the agent

Each model role handles a different part of the loop so the system stays fast without losing context.

👁

Signal reading

The first layer reads market structure, wallets, volume shifts, and social velocity to detect unusual movement early.

🧠

Reasoning

The reasoning layer compares opportunity, risk, and timing so the agent does not react blindly to noise.

Action shaping

The final layer formats the response into a trade plan, a Telegram update, or an execution-ready instruction.

How a trading decision moves through the system

The models do not work in isolation. They pass context through a fixed sequence so the final action stays coherent.

Step 01

Gather context

ValOs collects token activity, wallet behavior, narrative movement, and user instructions from Telegram.

Step 02

Score the opportunity

The model weighs momentum strength, crowding risk, and market timing before deciding whether to continue.

Step 03

Return action

If the setup is valid, the agent produces an execution path, risk framing, and a user-facing update.

valos / model-flow
$ ingest market-state
+ narratives loaded
+ wallet clusters loaded
$ reason over setup
+ conviction above threshold
$ return action plan
trade direction, sizing, and live monitoring prepared