01BEYOND THE CLAIM

A headline is only one part of the picture. Inspect the activity, observation window and recorded costs behind it.

02CONSISTENT CONDITIONS

Bring public observations into the same frame. Keep the source and coverage visible before drawing comparisons.

03CONTEXT STAYS VISIBLE

Separate live observations from declared methods. Missing history and measurement limits stay attached to the profile.

SOLANA INTELLIGENCENVIDIA NIM / CONNECTOR PREPARED

GPU MINDSET. ONCHAIN EVIDENCE.

AI AGENTS.
A CLEARER
PICTURE.

Look past the headline.
Render the evidence.

Inspect public Solana activity, compare sources and follow the context behind an agent’s claims.

EXPLORE RASTER

HOW IT WORKS

FROM PUBLIC ACTIVITY TO A CLEARER PICTURE

Observe an agent’s public activity, inspect the inputs and explain the context. A read-only workflow on Solana, with transparent sources and clear measurement limits.

01OBSERVE

START WITH THE SOURCE

Read public Solana activity. Keep the address, transaction, timestamp and observation window attached to the data. No execution permissions.

02MEASURE

MAKE EVERY INPUT COUNT

Inspect balances, recent activity and recorded costs. Compare consistent windows. Missing positions and incomplete histories stay visible.

03EXPLAIN

RENDER THE CONTEXT

Turn observations into an inspectable profile. Keep reported methods separate from direct evidence, with sources and update times beside each conclusion.

SOLANA / READ ONLY

RASTER

Independent intelligence
for AI trading agents.

02 / THE WORKSPACE

ONE ADDRESS.
A CLEARER STARTING POINT.

Observe public Solana activity.
Inspect the source. Keep the context.

RASTERPUBLIC ADDRESS EXPLORERINTERFACE PREVIEW

WALLET INTELLIGENCE

START WITH A PUBLIC ADDRESS.

USE MY WALLET ↗

SOL BALANCE—Current observation

RECENT SIGNATURES—Defined history window

TOKEN ACCOUNTS—Subject to RPC coverage

ACTIVITY WINDOWNO ADDRESS LOADED

Observed activity appears here after inspection.

EVIDENCE STATUS

Public activity OBSERVABLE

Strategy claims REPORTED

Returns & risk UNMEASURED

NO PERFORMANCE RESULTS INVENTED.OPEN THE LIVE WORKSPACE ↗

03 / FOR BUILDERS

A SHARED FRAMEWORK
FOR BETTER EVALUATION.

For builders, evaluators and researchers who want inspectable evidence. Every observation, method and limitation belongs in the same view.

EXPLORE METRICS

01 / 04

AGENT BUILDERS 01

AGENT BUILDERS

Put your public address, observation window and method in one place. Give every claim a source before asking anyone to trust it.

WHAT TO DISCLOSE

Public operating address
Strategy and evaluation window
Model and configuration changes
Data sources and missing coverage
Live, paper or backtest mode

AGENT BUILDERS 01

AGENT EVALUATORS 02

AGENT EVALUATORS

Compare agents only when the windows, costs and evidence support a useful comparison. Start with the questions behind the number.

QUESTIONS TO ASK

Is the activity attributable?
Are costs included?
Is the history complete?
What was measured directly?
Which assumptions remain?

AGENT EVALUATORS 02

RESEARCHERS 03

RESEARCHERS

Inspect the reasoning behind a metric. Keep observations and reported information separate, and make uncertainty visible.

EVIDENCE TO INSPECT

Defined observation windows
Position and exposure history
Different market conditions
Repeatable methodology
Source and last update

RESEARCHERS 03

INFERENCE TEAMS 04

INFERENCE TEAMS

The server-side NVIDIA NIM connector is prepared. Configure credentials and a model to enable explanations of observed activity.

CONNECTOR REQUIREMENTS

Server-side API credentials
Configured NIM model
Public observation context
Source-linked explanations
Inference activation and testing

INFERENCE TEAMS 04

04 / METRICS

MEASURE WHAT MATTERS.

Six dimensions of useful evaluation. A metric is defined before it is measured.

01 RETURNS

What did the agent earn, and over which window?

NET PERFORMANCE: A complete equity history with deposits, withdrawals and costs accounted for.
RISK-ADJUSTED RETURNS: Calculations and their assumptions disclosed alongside the result.
WINDOW CONSISTENCY: Results examined across defined, comparable periods.

02 RISK

How severe were the losses and fluctuations?

MAXIMUM DRAWDOWN: Peak-to-trough change measured over a consistent equity history.
VOLATILITY: How widely returns varied, beyond the average result.
EXPOSURE: Positions and concentration, with missing history made explicit.

03 EXECUTION REALITY

Does the result survive the costs of execution?

SLIPPAGE: Quoted versus executed price, when trade-level attribution is available.
FEES / FUNDING: Realized costs, with coverage and omissions clearly identified.
EXECUTION TIMING: The gap between a decision and its attributable execution.

04 METHODOLOGY

What conditions produced the observations?

LIVE / PAPER / BACKTEST: The operating mode attached to each reported result.
SOURCE / WINDOW: A defined source, timestamp and observation period.
ATTRIBUTION: Evidence linking activity to the agent being evaluated.

05 ROBUSTNESS

Would the method hold beyond one favorable test?

ASSET COVERAGE: Which assets and venues were tested; limited coverage remains visible.
OUT-OF-SAMPLE TESTING: Periods outside the strategy’s original development window.
REGIME SENSITIVITY: How the method responds to changing market conditions.

06 AGENT STATUS

Is the profile still describing the agent that runs today?

ACTIVE / PAUSED: Recent activity observed without assuming who controls the wallet.
VERSION HISTORY: Changes to the strategy and model are kept in context.
COVERAGE STATUS: Missing observations are labeled rather than estimated.

CURRENT WORKSPACE: PUBLIC OBSERVATIONS ARE AVAILABLE. RETURNS, RISK AND ATTRIBUTABLE AGENT PERFORMANCE REMAIN UNMEASURED.

05 / EVIDENCE

DIFFERENT SOURCES.
DIFFERENT CERTAINTY.

Follow the source before following the conclusion.

01 / OBSERVED

Read from the chain.

Public balances, recent signatures, transaction outcomes and timestamps retrieved from Solana RPC. Source, time window and coverage travel with the observation.

02 / REPORTED

Attributed to the builder.

Strategy descriptions and model choices remain attributed to their source. A declaration is context; it is not an independently verified performance result.

03 / UNMEASURED

Uncertainty stays visible.

Missing equity histories, incomplete costs and unclear attribution are labeled. No current balance is turned into a claim about returns or agent quality.

06 / COMPARE

COMPARE CONTEXT.
BEFORE CONCLUSIONS.

Two public addresses. One observation window. Keep incomplete evidence visible instead of turning it into a ranking.

OPEN COMPARE

OBSERVATIONADDRESS AADDRESS B

SOL balanceAwaiting addressAwaiting address

Recent activityAwaiting addressAwaiting address

Source coverageNot inspectedNot inspected

Agent returnsUnmeasuredUnmeasured

INTERFACE PREVIEW / NO ADDRESSES LOADED

07 / ARCHITECTURE

PUBLIC DATA.
PREPARED INFERENCE.

SOLANARASTERNVIDIA NIM

Solana provides read-only observations. RASTER puts sources and coverage in context. The NVIDIA NIM connector is prepared; inference requires server credentials and a configured model.

INDEPENDENT PRODUCT / NO NVIDIA PARTNERSHIP OR ENDORSEMENT

IMPLEMENTED API / THREE ROUTES

INSPECT THE INTERFACE.

GET/api/status

Reports Solana configuration and whether the NVIDIA connector is configured. Server credentials are never returned.

GET/api/agent

Reads a valid public Solana address via the address query parameter. Observations include source coverage and timestamps. Unavailable measurements stay labeled.

POST/api/nvidia

Prepared explanation route accepting an address in JSON. Currently unavailable until NVIDIA NIM is configured; it returns status 503 instead of invented inference.

08 / QUESTIONS

THE DETAILS,
IN THE OPEN.

What can I use today?

Open the app, paste a public Solana address and inspect observed balances, recent signatures and available account information. Save addresses in a browser-local watchlist or compare two addresses. RPC coverage can vary; unavailable data is labeled.

What is the NVIDIA connection?

A server-side NVIDIA NIM adapter is prepared for inference. It is not activated in this release: API credentials and a model must be configured and tested. RASTER is independent and does not claim an NVIDIA partnership or endorsement.

Are benchmark returns live?

No verified agent-performance results are published. Current public-address observations do not establish returns, drawdown or autonomous control. Those measurements require complete, attributable histories.

Does a wallet connection authorize trading?

No. The workspace reads public Solana data. Connecting a supported wallet supplies its public address; it does not grant custody or authorize a transaction.

Can I use a public address without a wallet?

Yes. Paste any valid Solana public address. No wallet connection, private key or seed phrase is needed for inspection.

AI AGENTS.A CLEARERPICTURE.