PP-Data API checking
A transparent data pipeline connecting market, document, table, JSON, and sensor sources

Open data infrastructure · V0.1

P-Data

Connect. Analyze. Clean. Provide.

One provenance-first data layer for research, quant, AI, agents, and applications.

Open the data lab View source
01 Connect02 Analyze03 Clean04 Provide

Browser-local workspace

Inspect before you mutate

Load CSV or JSON, inspect its shape, then apply an explicit cleaning policy. Data stays in this browser.

Analysis report

No dataset loaded

RAW
Rows0
Columns0
Missing0%
Duplicates0
Quality—
Changes0
SourceAwaiting analysis
Load a dataset to inspect the first eight rows.

Four independent agents

A small system with explicit boundaries

Each stage is independently callable. The complete path is available through PDataPipeline.

01

Data Connector

Translates files and remote responses into a common, provenance-carrying dataset.

  • CSV and JSON
  • REST endpoints
  • Kraken OHLC
02

Data Analyst

Profiles schema, missingness, duplicates, timestamps, sampling, and suspicious jumps before cleaning.

  • Non-destructive
  • Machine-readable
  • Policy input
03

Data Cleaner

Runs explicit, composable transformations while reporting every row and cell change.

  • Missing values
  • Outliers and duplicates
  • Schema normalization
04

API Provider

Exposes catalog-shaped access and exports clean data with a sibling lineage manifest.

  • Six formats
  • Catalog methods
  • Traceable outputs

Stable surfaces

Use it from code, CLI, or an agent

The same reports are readable by people and serializable for downstream systems.

from pdata import CSVConnector, DataAnalyst

raw = CSVConnector("market.csv").fetch()
report = DataAnalyst().analyze(
    raw,
    time_column="timestamp",
)

# Raw data is still unchanged.
print(report.missing_ratio)

Live metadata endpoint

GET /api/catalog

Loading API catalog…

Operating principles

Trust comes from visible decisions.

Raw is immutable. Derive a version; keep the source intact.

Analysis precedes mutation. Understand what a cleaning policy will touch.

Lineage is first-class. Every export can explain where it came from.

Simple before distributed. Scale only when measurement requires it.