Sign in to Wattrail.

Wattrail checks smart meter readings as they arrive, finds and counts the half-hours that never arrived, and keeps a proof for every meter-day. When a reading or a settlement figure is questioned, you can show what really arrived.

What you can do in Wattrail

Check each reading as it arrives

Readings that were changed, sent twice, or came from a meter Wattrail doesn't know are refused, and the refusal is recorded.
Find and count missing half-hours

If a reading never arrives, Wattrail records which half-hour is missing, instead of letting it disappear into an average.
Keep the proof

Each meter-day is sealed. Its proof pack can be checked offline, on your own computer, with meldra's verifier tool.

Looking for a pilot?

We're looking for first pilot partners among network operators, suppliers and settlement teams. Bring one disputed figure and we'll trace it with you.

Email enquiry@meldra.ai

14-Day Free Training Pass Active

meldra.py Data Engineering Course

Master Zero-Copy architectures, Iceberg tables, schema evolution, and serverless data quality rules in 5 interactive modules.

Module 1: Lakehouse Architecture Foundations

Understanding serverless metadata scanning vs. continuous cluster execution.

Module 2: Setting up AWS Glue & GCP BigLake

Configuring secure credentials, regions, and cloud object store links.

Module 3: Ingesting Data Tables (Append & Merge)

Programmatically loading CSV files and running merge-upserts.

Module 4: Zero-Copy Schema Evolution

Adding, renaming, and dropping columns instantly in metadata.

Module 5: Active Data Contract Constraints

Writing and enforcing schema validators before pipeline commits.

Access Locked

Claim Your Training Pass

Course Syllabus

Module 1

Architecture Foundations

Module 2

Setting up Cloud Catalogs

Module 3

Ingesting Data Tables

Module 4

Zero-Copy Schema Evolution

Module 5

Data Contracts & Airflow

Module 1 · Conceptual Foundations

Lakehouse Architecture Foundations

Traditional architectures replicate enterprise telemetry across several stages (Raw landing ➔ Stage ➔ Core warehouse ➔ Data marts). This results in massive cloud compute charges (the Spark Tax) and security vulnerabilities due to scattered copies.

meldra.py Zero-Copy Strategy: Ingest raw CSV, API feeds, or database logs directly to open Apache Iceberg tables stored on your own S3/GCS. Queries run serverless in DuckDB or PyIceberg, keeping the storage completely decentralized and zero-copy.

                # Concept: metadata-only reads
                import pyiceberg.catalog
                catalog = pyiceberg.catalog.load_catalog("local")
                table = catalog.load_table("default.smartmeter_readings")
              

Practice these steps in the interactive workspace sandbox.

meldra

Energy data that shows its work.

Create account

Your data lake, powered by AI

Email address

Password

Confirm password

Already have an account?

Check your email

We've sent a 6-digit code to

your@email.com

Code expires in 10:00

Forgot password?

Enter your email and we'll send a reset code.

Email address

Enter reset code

We sent a 6-digit code to

your@email.com

Code expires in 10:00

Back

Set new password

Choose a strong password for your account.

New password

Confirm password

API Gateway Operational

Ctrl+K

meldra ai · energytrust protocol W3C / IEC STANDARDS ALIGNED

ETP Ontology

Reuse four W3C and IEC standards, extend in exactly one place. The semantic layer that turns "IEC CIM aligned" from a claim in a deck into something an integrator can dereference and check.

Dereference URI: https://meldra.ai/ns/etp# Download Turtle (.ttl)

Borrow almost everything

An ontology's credibility comes from what it reuses, not what it invents. Four established vocabularies already describe most of ETP. One thing they cannot express is precisely your differentiator — and that is the only place a new namespace is justified.

etp: — the only new namespace

SequenceCommitment · OmissionAssertion · VerificationActivity · TimestampAnchor

▲ reuses ▲ reuses ▲ reuses ▲ reuses

IEC CIM

61970 / 61968

UsagePoint · Meter
ACLineSegment

SOSA / SSN

W3C

Observation · Sensor
FeatureOfInterest

PROV-O

W3C

Entity · Activity · Agent
wasDerivedFrom

QUDT · TIME

units & intervals

unit:KiloW-HR
time:Interval

The gap none of them fill

PROV-O states what a thing came from. It cannot state that nothing else existed.

IEC CIM – grid & metering

SOSA/SSN – observations

PROV-O – derivation

etp: – completeness

Class mapping

Every ETP entity maps onto existing classes except the last three. Anything an integrator already understands stays in their vocabulary.

ETP Entity	IEC CIM	SOSA / SSN	PROV-O
UsagePoint	cim:UsagePoint	sosa:FeatureOfInterest	prov:Entity
Meter	cim:Meter	sosa:Sensor	prov:Agent
Register	cim:ReadingType	sosa:ObservableProperty	—
IntervalReading	cim:IntervalReading	sosa:Observation	prov:Entity
Feeder	cim:ACLineSegment	—	—
Substation	cim:Substation	—	—
Gateway verify	—	sosa:Procedure	prov:Activity
Timestamp authority	—	—	prov:Agent
TelemetryBlock	—	—	prov:Entity †
SequenceCommitment	—	—	prov:Entity †
OmissionAssertion	—	—	prov:Entity †

† Subclassed in the etp: namespace because no existing vocabulary carries the semantics.

Formal W3C Turtle Specification & Worked Instance

@prefix etp:  <https://meldra.ai/ns/etp#> .
@prefix cim:  <http://iec.ch/TC57/CIM100#> .
@prefix sosa: <http://www.w3.org/ns/sosa/> .
@prefix prov: <http://www.w3.org/ns/prov#> .
@prefix owl:  <http://www.w3.org/2002/07/owl#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix xsd:  <http://www.w3.org/2001/XMLSchema#> .

# ── A signed reading ──────────────────────────────────────────
etp:TelemetryBlock  a owl:Class ;
    rdfs:subClassOf prov:Entity , sosa:Observation ;
    rdfs:label "Telemetry Block" ;
    rdfs:comment "A single interval reading bound to a monotonic sequence number and an ECDSA signature." .

etp:hasNonce       a owl:DatatypeProperty ;
    rdfs:domain etp:TelemetryBlock ; rdfs:range xsd:long .
etp:blockHash      a owl:DatatypeProperty ;
    rdfs:domain etp:TelemetryBlock ; rdfs:range xsd:hexBinary .
etp:previousHash   a owl:DatatypeProperty ;
    rdfs:domain etp:TelemetryBlock ; rdfs:range xsd:hexBinary .
etp:signature      a owl:DatatypeProperty ;
    rdfs:domain etp:TelemetryBlock ; rdfs:range xsd:hexBinary .

# ── The novel assertion: completeness over an interval ────────
etp:SequenceCommitment  a owl:Class ;
    rdfs:subClassOf prov:Entity ;
    rdfs:label "Sequence Commitment" ;
    rdfs:comment "A cryptographic commitment that the observations over a closed nonce interval are exactly those enumerated — asserting completeness, not merely integrity. PROV-O has no equivalent." .

etp:merkleRoot          a owl:DatatypeProperty ; rdfs:range xsd:hexBinary .
etp:anchorDigest        a owl:DatatypeProperty ; rdfs:range xsd:hexBinary .
etp:firstNonce          a owl:DatatypeProperty ; rdfs:range xsd:long .
etp:lastNonce           a owl:DatatypeProperty ; rdfs:range xsd:long .
etp:observedLeafCount   a owl:DatatypeProperty ; rdfs:range xsd:int .
etp:expectedPeriodCount a owl:DatatypeProperty ; rdfs:range xsd:int .
etp:previousCommitment  a owl:ObjectProperty ;
    rdfs:domain etp:SequenceCommitment ;
    rdfs:range  etp:SequenceCommitment ;
    rdfs:comment "Chains day N to day N-1, closing the boundary." .

# ── A positive assertion that data is absent ──────────────────
etp:OmissionAssertion  a owl:Class ;
    rdfs:subClassOf prov:Entity ;
    rdfs:label "Omission Assertion" ;
    rdfs:comment "Asserts that identified members of a committed sequence are absent. The assertion is itself evidence, and is anchored alongside the root." .

etp:missingFrom     a owl:DatatypeProperty ; rdfs:range xsd:int .
etp:missingTo       a owl:DatatypeProperty ; rdfs:range xsd:int .
etp:gapCount        a owl:DatatypeProperty ; rdfs:range xsd:int .
etp:omissionClass   a owl:DatatypeProperty ; rdfs:range xsd:string .

# ── External time attestation ─────────────────────────────────
etp:TimestampAnchor  a owl:Class ;
    rdfs:subClassOf prov:Entity ;
    rdfs:comment "An RFC 3161 token attributed to a timestamp authority acting as a prov:Agent." .

etp:isSimulated  a owl:DatatypeProperty ; rdfs:range xsd:boolean ;
    rdfs:comment "True when locally generated rather than issued by an external authority. Always asserted, never omitted." .
etp:pkiStatus    a owl:DatatypeProperty ; rdfs:range xsd:int .

# ── The act of verifying ──────────────────────────────────────
etp:VerificationActivity  a owl:Class ;
    rdfs:subClassOf prov:Activity ;
    rdfs:comment "Route validation, replay check, canonical hash recomputation and signature verification, performed at the ingestion boundary." .

etp:verificationStatus a owl:DatatypeProperty ; rdfs:range xsd:string .

Estate overview

Your meters, the readings received, and the cases that need someone to look at them. Figures come from your organization's data.

—

Meters

—

Readings received

—

Open cases

—

Last day sealed

Estate Telemetry Activity

Verified readings & gaps detected (last 14 days)

Loading…

Getting set up

Loading…

Recent activity

From the real audit trail

Loading…

meldra Core Domain

meldra's own Apache Iceberg catalog, RBAC-aware SQL engine, and graph store — the same live, tenant-scoped API the rest of the app uses.

Select Tool

Tool Description

List the Iceberg namespaces visible to your tenant

Arguments (JSON)

Execution Console Output

Ready. Select a tool and execute to see trace logs.

Returned Value (JSON)

{}

Connect an external MCP client

The tools above are also exposed as a real, spec-compliant MCP (Model Context Protocol) stdio server — backend/mcp_server.py. Any MCP client (Claude Desktop, Claude Code, another agent) can connect to it directly. Generate an API token in Settings › API Access, then point your client at this server:

{
  "mcpServers": {
    "meldra": {
      "command": "python",
      "args": ["/absolute/path/to/backend/mcp_server.py"],
      "env": {
        "MELDRA_API_URL": "https://api.meldra.ai",
        "MELDRA_API_TOKEN": "<your generated token>"
      }
    }
  }
}

Ingest Data

Load data into your Iceberg lakehouse from local files, SaaS applications, or sample data sets.

Catalog not checked

1. Upload CSV Dataset

Up to 50 MB · UTF-8 · comma or semicolon delimited. Schema is detected automatically.

Drag & drop your CSV file here or

2. Configuration & Schema Mapping

Namespace

Table Name

Write Mode

Merge Key (Optional)

Schema Mapping Specs

Column	Iceberg Type
Upload a CSV to view schema

Writing files to S3...

Creating Iceberg metadata snapshots

Choose a source system

Authentication fields and schemas adapt dynamically to the system you select.

AMI Smart Meter HES Authentication

Credentials are stored encrypted in the lakehouse key vault and never shown again.

Configure and Ingest

Unlocks after a successful connection test.

AMI

Smart Meter Readings

Synthetic · 10 invented rows · 6 columns

An invented example file in the shape of AMI interval telemetry: meter ID, active power (kW), reactive power (kVAR), voltage (V), timestamp and status. Downloads as CSV for trying Upload File. Not real data.

Fields: meter_id, kw_active, kvar_reactive, voltage, timestamp, status

LDN

Real London households

Real data · UK Power Networks, 2011–2014

Half-hourly electricity use from the Low Carbon London smart-meter trial, published on the London Datastore. The server checks the published file's SHA-256 before loading. Open data, not meter-signed.

Fields: lcl_id, tariff, reading_ts, kwh, source_file, source_line

Households (1–200)

SCD

Feeder & Substation Telemetry

Synthetic · 10 invented rows · 6 columns

An invented example file in the shape of DNO feeder telemetry: feeder ID, substation name, active power (MW), power factor and transformer temperature. Downloads as CSV for trying Upload File. Not real data.

Fields: feeder_id, substation_name, active_power_mw, power_factor, transformer_temp_c, timestamp

ETP

ETP Mutation Audit Logs

Synthetic · 10 invented rows · 6 columns

An invented example file in the shape of metering-boundary security events: route HMACs, sequence nonces, drift and decoy diversions. Downloads as CSV for trying Upload File. Not real data.

Fields: route_id, meter_mac, sequence_nonce, hmac_signature, route_drift_ms, phantom_diverted

Nodes in Graph

Edges in Graph

Active Database

openCypher Query Console

Example: MATCH (a:Entity)-[r:FED_BY]->(b:Entity) RETURN a.node_id, b.node_id LIMIT 10

Iceberg-to-Graph Projection

Project columns of an Iceberg catalog table into node vertices and directional relationship edges.

Iceberg Table

Edge Label

Source Column (From)

Target Column (To)

Query Output

Run a Cypher query above to see results

📖 REST API Reference

Complete documentation for the meldra backend API. All endpoints are available at your configured backend URL.

API Gateway Production Endpoint (Secured)

FastAPI /docs

🌐 meldra.py Developer Workspace

Configure multi-cloud connections, execute Python manipulation code, and export Airflow DAG pipelines.

Cloud Catalog Connection

Target S3 Bucket URI

Environment region

Glue Catalog Database

AWS Authentication Method

Target IAM Role ARN (STS AssumeRole / OIDC)

External ID (Optional)

CISO Approved · Short-Lived Tokens Only · Zero Long-Lived Keys Stored

meldra.py Python Notebook

Console logs output

System idle. Click 'Execute Sandbox' to run python manipulation...

Airflow / Cloud Composer DAG Exporter

Deploy scheduled data validation & compaction pipelines matching this environment configuration.

from datetime import datetime, timedelta
from airflow import DAG
from airflow.operators.python import PythonOperator

default_args = {
    'owner': 'meldra',
    'start_date': datetime(2025, 1, 1),
    'retries': 2,
    'retry_delay': timedelta(minutes=5),
}

def run_meldra_compaction():
    from meldra import MeldraCatalog
    catalog = MeldraCatalog()
    # Optimize small files into manifest clusters on S3
    catalog.optimize_table("default", "smartmeter_readings_sample")

with DAG(
    'meldra_lakehouse_compaction_dag',
    default_args=default_args,
    description='Compacts manifest Parquet files on AWS S3 / Google Cloud GCS',
    schedule_interval='@daily',
    catchup=False,
) as dag:
    
    compact_task = PythonOperator(
        task_id='trigger_compaction_task',
        python_callable=run_meldra_compaction,
    )

Audit Ledger

Actions recorded by the server: who did what, and when. Admins see every account in their deployment; everyone else sees their own actions.

Filter:

User Directory & Active Role Scope Management

Manage enterprise user accounts, assigned RBAC role scopes, column-level masking rules, and employee profile legal document vaults.

Provisioned Users & Active Role Scopes

User & Email	Role	Approver	Verified	Last Login	Profile & Actions
Loading…

Persona Role Scopes & Column Masking Rules

Business Analyst

Read-only access to conformed silver and gold tables. Customer PII and raw addresses are Level 3 Masked.

Data Engineer

Full DDL/DML access to schemas, pipeline DAG orchestration, and Iceberg catalog schema evolution.

Compliance Auditor

Full read access to Audit Ledger, dispute evidence packs, Omission Register, and Merkle root anchors.

Admin

Full system administration, user role assignments, approver designation, and row-level filter policies.

Environment Namespace

Iceberg Tables

No tables found.

Interactive SQL Console (DuckDB Engine)

0 rows returned in 0ms

No active query
Run a query to populate results.

Production Approvers

Who can approve a production promotion — independent of role, and never the person who requested it (enforced server-side in /v1/promotions/{id}/decide).

Loading…

API Traffic Monitor

CONNECTED

# Method Endpoint / Tool Name Status Latency

Waiting for API traffic events...

🔍 Request & Response Payload Inspector

Select an event row from the timeline to inspect payloads.

Total Events: 0 HTTP: 0 AI Tools: 0 Errors: 0

Distributed Query Engine — Query Lab

Run SQL, Graph algorithms, or Python extractions across your Iceberg tables. All results are cached 1 hour.

DuckDB + NetworkX

Query Mode

Namespace

Table Name

SQL Query (use 'iceberg_table' as source)

Limit:

Graph Name

Graph mode runs on a network of nodes and links, not directly on a table. Build one first: Network model → Project table to graph (graph name grid_network). Example on the London data: source lcl_id, target tariff, then run Degree Centrality.

Algorithm (or leave blank and write Cypher)

Cypher Pattern (optional)

Namespace

Table Name

Python Script (must set result_df)

Security sandbox: No OS/file/network access. 30s timeout enforced.

Dry run — EXPLAIN only (no data)

Results

Configure a query and press Run Query to see results here.

Recent Jobs

Last 10 · auto-expires 1 hour

Mode	Status	Rows	Time
No queries yet

Help & Documentation

Esc to close ×

ON THIS SCREEN

What "reading completeness" means ▾

Reading completeness is the share of expected half-hourly meter intervals that were received and passed validation for the selected period. Intervals that were expected but not received are listed in the Omission Register.

When to raise a case vs. wait ▾

meldra does not set a dispute threshold. Follow your organisation's settlement process; the Omission Register shows which intervals are missing and the evidence pack records what was received, so you can decide whether a case is warranted.

Who can sign off settlement ▾

Who can approve and export evidence is set by your workspace administrator in Roles & Access. Ask your administrator if an action is unavailable to you.

How zero-copy replication works ▾

Zero-copy means meldra queries Apache Iceberg tables in place: it reads the table metadata and Parquet files from your storage rather than copying them into a separate store.

GUIDED WALKTHROUGHS

Close your first dispute 4 min

Set up a new data source 6 min

Configure RBAC Governance 5 min

Still stuck?

Email the meldra team and include what you were doing and the screen you were on.

Priya Raman

Admin Verified

priyaraman@utility.co.uk · DNO Utility Licensee

Personal Information

First Name

Last Name

Email Address

Business Phone

Organization & Employee Master Information (TE Profile)

Employee ID

Tally Employee Code (TE ID)

Organization / Company

Business Unit / Dept

UK Org-to-Org Legal & Regulatory Approvals

NDA Execution Status

UK GDPR Data Sharing Clearance

Settlement data role

Smart Energy Code (SEC) Boundary

UK Security Vetting Level

Work Approval Manager Sign-off

User-Configurable Platform Data Access

Enable PII Masking on Telemetry Feeds Require ETP Cryptographic Proof for Sign-off Allow Direct Zero-Copy S3 Parquet Exports Receive High-Priority Settlement Exception Alerts

Sign in to Wattrail.

What you can do in Wattrail

Check each reading as it arrives

Find and count missing half-hours

Keep the proof

Looking for a pilot?

meldra.py Data Engineering Course

Module 1: Lakehouse Architecture Foundations

Module 2: Setting up AWS Glue & GCP BigLake

Module 3: Ingesting Data Tables (Append & Merge)

Module 4: Zero-Copy Schema Evolution

Module 5: Active Data Contract Constraints

Claim Your Training Pass

Course Syllabus

Architecture Foundations

Setting up Cloud Catalogs

Ingesting Data Tables

Zero-Copy Schema Evolution

Data Contracts & Airflow

Lakehouse Architecture Foundations

Setting up AWS Glue & GCP BigLake Catalogs

Ingesting Data Tables (Append vs. Merge)

Zero-Copy Schema Evolution

Data Contracts & Airflow Automation

Welcome back

Create account

Check your email

Forgot password?

Enter reset code

Set new password

ETP Ontology

Borrow almost everything

Class mapping

Formal W3C Turtle Specification & Worked Instance

Estate overview

Estate Telemetry Activity

Getting set up

Ask the Assistant

Ingest

Data Studio

Query Lab

Governance

Observability

Recent activity

MCP Registry

meldra Core Domain

Connect an external MCP client

Ingest Data

1. Upload CSV Dataset

2. Configuration & Schema Mapping

Writing files to S3...

Choose a source system

AMI Smart Meter HES Authentication

Configure and Ingest

Smart Meter Readings

Real London households

Feeder & Substation Telemetry

ETP Mutation Audit Logs

openCypher Query Console

Iceberg-to-Graph Projection

Query Output

📖 REST API Reference

🌐 meldra.py Developer Workspace

Cloud Catalog Connection

meldra.py Python Notebook

Airflow / Cloud Composer DAG Exporter

Audit Ledger

User Directory & Active Role Scope Management

Provisioned Users & Active Role Scopes

Persona Role Scopes & Column Masking Rules

Environment Namespace

Iceberg Tables

Interactive SQL Console (DuckDB Engine)

Interactive Python Workspace (Notebook)

SaaS Role-Based Access Control & Masking

Row-Level Filters

Production Approvers

Data Catalog

No table selected

Schema Evolution (Zero-Downtime Metadata Updates)