Docs
Every dataset is a Parquet file at a stable URL, rebuilt daily. Query it from anywhere that reads Parquet, and check definitions before drawing conclusions.
Query the files
Each table in the catalog lives at https://data.filecoindataportal.xyz/<dataset>.parquet. Run SQL in the DuckDB CLI or in the browser with SQL Workbench. daily_network_metrics is also published as JSON for dashboards.
select date, onboarded_pibs, raw_power_pibs
from 'https://data.filecoindataportal.xyz/daily_network_metrics.parquet'
order by date desc
limit 7;import polars as pl
df = pl.read_parquet("https://data.filecoindataportal.xyz/miners.parquet")const url = "https://data.filecoindataportal.xyz/daily_network_metrics.json"
const rows = await (await fetch(url)).json()attach 'https://data.filecoindataportal.xyz/fdp.duckdb' as fdp (read_only);
use fdp;
show tables;Ask an agent
Point a coding agent at SKILL.md or llms.txt. Every dataset page also has a Markdown version listed in the Markdown catalog. Or paste this self-contained prompt and add your question.
Answer my question with Filecoin Data Portal datasets.
Each dataset is a Parquet file at https://data.filecoindataportal.xyz/<dataset>.parquet.
Dataset definitions and column descriptions live in https://github.com/davidgasquez/filecoin-data-portal/tree/main/assets/main.
Query with DuckDB. Read the column descriptions before writing SQL.
Dates are UTC. Null means unknown, not zero. Atto-denominated columns are exact
DECIMAL(38,0) integers; keep them exact and divide by 1e18 only for display.
Show the SQL you ran and the latest date each answer covers.
Question: Read before citing
- Daily tables end at UTC yesterday at the latest. Sources update on different schedules, so tables can end on different dates.
- Null means unavailable or undefined, not zero. Gaps in a source leave dependent metrics empty.
- Definitions live in column descriptions and asset SQL. A metric name alone is not a definition.
- Atto-denominated columns are exact integers. Keep them exact and divide by 1e18 only for display. Never add amounts in different tokens.
- Filecoin Pay volume covers all rails, not only storage. Warm storage capacity and network power measure different populations; do not divide one by the other.
- Provider checks are synthetic probes, not client traffic or a promise of availability.
- Ratios use pooled counts. Do not average per-provider percentages.
Sources
Datasets come from Lily chain data, Filecoin JSON-RPC, the RIBA FEVM receipt archive, the Filecoin API, ProbeLab Dealbot checks, and market data. Each dataset page lists its sources and links the asset code that declares them. The Filecoin data sources guide compares these and other APIs, indexers, and RPC endpoints, with access details and limits.
Open source
SQL and Python pipelines, dataset definitions, and this site live in one repository. Each dataset page links its asset code. Missing a source or a dataset? Corrections and requests are welcome as issues or pull requests.