{"id":333,"date":"2026-05-11T16:10:02","date_gmt":"2026-05-11T16:10:02","guid":{"rendered":"https:\/\/jua.ai\/articles\/best-weather-api-2026\/"},"modified":"2026-07-04T05:05:17","modified_gmt":"2026-07-04T05:05:17","slug":"best-weather-api-2026","status":"publish","type":"post","link":"https:\/\/jua.ai\/articles\/best-weather-api-2026\/","title":{"rendered":"Best Weather API for Energy Trading: A Six-Part Comparison"},"content":{"rendered":"<p><em>Written by: Olivier Lam, Physical AI Team, Jua.ai AG | Last updated: July 2, 2026<\/em><\/p>\n<h2 id=\"key-takeaways\">Key Takeaways for Energy Trading Teams<\/h2>\n<ul>\n<li>Jua for Energy\u2019s EPT-2 model delivers deterministic accuracy that beats ECMWF HRES on every lead time for wind, temperature, and solar radiation.<\/li>\n<li>EPT-2e outperforms the 50-member ECMWF ENS mean on both RMSE and CRPS at virtually every lead time, providing superior probabilistic skill.<\/li>\n<li>Up to 24 daily refreshes and 2.5-hour dissemination lead times remove the structural blind spots created by legacy NWP cycles.<\/li>\n<li>Schema-stable Python SDK, Apache Arrow support, and multi-year hindcast access let quant teams integrate and backtest in days rather than quarters.<\/li>\n<li><a href=\"https:\/\/meetings-eu1.hubspot.com\/guett\/energy-trading?uuid=d780665f-ff71-439c-addf-c80e49af0627\" target=\"_blank\"><strong>Book a demo<\/strong><\/a> with Jua to run live benchmarks on your region and variables and see the accuracy advantage firsthand.<\/li>\n<\/ul>\n<h2>How AI-Native Physics Models Change Energy-Weather Economics<\/h2>\n<p>Numerical weather prediction (NWP) decomposes the atmosphere into three-dimensional grid cells and solves differential equations inside each one. That approach has powered energy-market forecasting for forty years. The compute ceiling is hard. A single NWP simulation consumes approximately 8,400 kWh and costs \u20ac1,000\u2013\u20ac20,000 on high-performance computing infrastructure. The result is two to four global forecasts per 24-hour period. <a href=\"https:\/\/www.bloomberg.com\/news\/articles\/2026-03-26\/energy-traders-turn-to-ai-to-forecast-the-weather-forecast?embedded-checkout=true\" target=\"_blank\">ECMWF&#8217;s two-week outlook is the definitive reference point for traders repricing risk around heating demand, renewable output, and system tightness<\/a>, and it runs twice a day. That refresh ceiling creates systematic blind spots as traders operate on stale forecasts while atmospheric conditions evolve.<\/p>\n<p>AI-native physics models remove that constraint by changing the underlying economics. A single EPT-2 inference runs on a single GPU in minutes at approximately 0.25 kWh and $0.20\u2013$15, which is roughly four orders of magnitude cheaper than an equivalent NWP run. Jua is a foundation model and agent company. Its Earth Physics Transformer (EPT) family is a general spatiotemporal transformer foundation model that learns the governing physics of complex systems, including mass, momentum, and energy conservation, directly from observational data. The architecture is domain-agnostic. Atmospheric prediction is the first physical system EPT has been fine-tuned for. Jua for Energy is the first applied product built on EPT and Athena, Jua&#8217;s AI agent. The relationship mirrors Anthropic and Claude Code: a horizontal AI platform with a flagship vertical product.<\/p>\n<h2>Core Technical Concepts for Evaluating Weather APIs<\/h2>\n<p><strong>RMSE<\/strong> (root mean square error) measures the average magnitude of forecast errors, and lower values indicate higher accuracy. <strong>CRPS<\/strong> (continuous ranked probability score) measures probabilistic forecast skill across the full distribution, and lower values indicate better-calibrated ensembles. <strong>Grib<\/strong> is the binary file format in which NWP centers distribute raw model output. <strong>NWP<\/strong> (numerical weather prediction) solves atmospheric physics equations on a discrete grid. An <strong>ensemble<\/strong> is a set of perturbed model runs that samples forecast uncertainty, and the spread of members quantifies probabilistic risk.<\/p>\n<p><strong>Lead time<\/strong> is the number of hours between forecast initialization and the valid time being predicted. <strong>Dissemination<\/strong> is the time at which a completed model run becomes available to end users. A <strong>hindcast<\/strong> (or re-forecast) is a model run initialized from historical conditions, used to backtest systematic strategies against years of past data. <strong>Any-\u0394t forecasting<\/strong> means EPT-2 is trained to predict at arbitrary time steps rather than rolling forward in fixed increments. Aurora and most peers use a fixed 6-hour grid and compound error with each roll. EPT-2 does not roll.<\/p>\n<p>With these technical concepts established, the next step is to evaluate weather APIs across the six dimensions that determine production fitness for energy trading.<\/p>\n<h2>Six-Dimension Evaluation Framework for Energy Trading APIs<\/h2>\n<p><strong>1. Deterministic accuracy versus ECMWF HRES.<\/strong> ECMWF HRES is the universal benchmark, with forty years of NWP leadership at 9 km resolution. <a href=\"https:\/\/arxiv.org\/abs\/2507.09703\" target=\"_blank\" rel=\"noindex nofollow\">EPT-2 outperforms ECMWF HRES on every lead time and on 10 m wind, 100 m wind, 2 m temperature, and surface solar radiation across the full 0\u2013240 hour range<\/a>, evaluated against more than 10,000 real ground stations on open-source StationBench with no post-processing or station fine-tuning.<\/p>\n<p><strong>2. Ensemble skill.<\/strong> Probabilistic forecasts support P&amp;L-critical decisions such as imbalance penalties, options pricing, and ramp-risk hedging, which all depend on the full distribution of outcomes. <a href=\"https:\/\/arxiv.org\/abs\/2507.09703\" target=\"_blank\" rel=\"noindex nofollow\">EPT-2e beats the 50-member ECMWF ENS mean on both RMSE and CRPS at virtually every lead time<\/a>. No AI peer ships a productized ensemble equivalent.<\/p>\n<p><strong>3. Update frequency and dissemination time.<\/strong> Stale forecasts between runs impose a structural cost on intraday strategies. EPT-2 RR refreshes up to 24 times per day, and a typical Jua run completes approximately 2.5 hours ahead of competing operational runs at the same cycle. Traditional NWP delivers two to four global forecasts per 24 hours.<\/p>\n<p><strong>4. Hindcast availability.<\/strong> Multi-year backtesting requires historical forecast data, not reanalysis, but actual model runs initialized from past conditions. Jua for Energy provides hindcast data across multiple Jua and third-party models via the Python SDK. Quant teams can validate strategies against years of history before committing capital.<\/p>\n<p><strong>5. Developer ergonomics.<\/strong> <code>pip install jua<\/code> installs the Python SDK. The REST API exposes more than 25 models through a single schema with Apache Arrow support for large payloads. Schema stability, OpenAPI documentation at <code>query.jua.ai\/docs<\/code>, and a developer dashboard at <code>developer.jua.ai<\/code> reduce integration timelines from quarters to days for typical quant teams.<\/p>\n<p><strong>6. Energy-specific variables.<\/strong> Hub-height wind at multiple vertical levels supports a range of commercial turbine classes. Surface solar radiation (SSRD) is a native EPT-2 output, while Microsoft Aurora produces no SSRD output. The platform exposes twenty-five variables in total, including precipitation, cloud cover, temperature at multiple levels, and pressure.<\/p>\n<h2>Head-to-Head Comparison Across Six Dimensions<\/h2>\n<p>The table below compares providers across the six evaluation dimensions. Data points for Jua for Energy are anchored to <a href=\"https:\/\/arxiv.org\/abs\/2507.09703\" target=\"_blank\" rel=\"noindex nofollow\">arXiv:2507.09703<\/a> and <a href=\"https:\/\/arxiv.org\/abs\/2410.15076\" target=\"_blank\" rel=\"noindex nofollow\">arXiv:2410.15076<\/a> using StationBench methodology. Hobbyist APIs (Open-Meteo, OpenWeatherMap) and general-purpose commercial APIs (Tomorrow.io, Visual Crossing) appear here because they dominate the current SERP landscape, but they are not production-grade alternatives for systematic energy trading.<\/p>\n<table>\n<thead>\n<tr>\n<th>Dimension<\/th>\n<th>Open-Meteo \/ OpenWeatherMap \/ Tomorrow.io \/ Visual Crossing<\/th>\n<th>ECMWF HRES &amp; Microsoft Aurora<\/th>\n<th>Jua for Energy (EPT-2 \/ EPT-2e)<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>Deterministic accuracy vs. HRES<\/strong><\/td>\n<td>Resell or wrap NWP outputs, with no published head-to-head benchmark on hub-height wind or SSRD<\/td>\n<td>HRES: the 40-year benchmark. <a href=\"https:\/\/arxiv.org\/abs\/2507.09703\" target=\"_blank\" rel=\"noindex nofollow\">Aurora loses to EPT-2 on 10 m wind, 100 m wind, and 2 m temperature across 0\u2013240 h<\/a><\/td>\n<td><a href=\"https:\/\/arxiv.org\/abs\/2507.09703\" target=\"_blank\" rel=\"noindex nofollow\">EPT-2 beats HRES on every lead time across 10 m wind, 100 m wind, 2 m temperature, and SSRD (0\u2013240 h, StationBench, 10,000+ stations)<\/a><\/td>\n<\/tr>\n<tr>\n<td><strong>Ensemble \/ probabilistic skill<\/strong><\/td>\n<td>No productized ensemble; Open-Meteo exposes GFS ensemble mean only<\/td>\n<td>ECMWF ENS: 50-member gold standard. Aurora: no productized ensemble<\/td>\n<td><a href=\"https:\/\/arxiv.org\/abs\/2507.09703\" target=\"_blank\" rel=\"noindex nofollow\">EPT-2e beats 50-member ECMWF ENS mean on RMSE and CRPS at virtually every lead time<\/a><\/td>\n<\/tr>\n<tr>\n<td><strong>Update frequency<\/strong><\/td>\n<td>Typically 1\u20134\u00d7 per day depending on underlying NWP source<\/td>\n<td>HRES: 2\u20134\u00d7 per day. Aurora: typically 4\u00d7 per day research cadence, with no operational schedule<\/td>\n<td><a href=\"https:\/\/nebius.com\/customer-stories\/jua\" target=\"_blank\">EPT-2 RR, EPT-2e, and actual-generation power forecasts follow the refresh and lead-time profile described in dimension 3<\/a><\/td>\n<\/tr>\n<tr>\n<td><strong>Hindcast access<\/strong><\/td>\n<td>Limited or absent, with no multi-year re-forecast datasets for backtesting<\/td>\n<td>HRES hindcasts available via MARS (member access). Aurora: no productized hindcast<\/td>\n<td>Hindcast data available across multiple Jua and third-party models via Python SDK for multi-year backtesting<\/td>\n<\/tr>\n<tr>\n<td><strong>Developer ergonomics<\/strong><\/td>\n<td>REST APIs with JSON, no Apache Arrow, schema varies by provider, and no energy-specific SDK<\/td>\n<td>HRES: grib via MARS, member access required. Aurora: research code with limited API<\/td>\n<td><code>pip install jua<\/code>, REST plus Apache Arrow, single schema across 25+ models, OpenAPI docs, schema-stable, and <a href=\"https:\/\/arxiv.org\/abs\/2507.09703\" target=\"_blank\" rel=\"noindex nofollow\">EPT-2 trained on 8 \u00d7 H100 GPUs in 10 days vs. Aurora&#8217;s 32 \u00d7 A100 GPUs over 18 days<\/a><\/td>\n<\/tr>\n<tr>\n<td><strong>Energy-specific variables<\/strong><\/td>\n<td>Standard surface variables only, with no hub-height wind levels and no SSRD<\/td>\n<td>HRES: full variable set including hub-height wind. Aurora: no SSRD output<\/td>\n<td>Wind at multiple vertical levels, SSRD, 25 variables total, native any-\u0394t, up to 1 km resolution in product, and EPT2-HRRR, a high-resolution variant of the EPT-2 family, natively forecasts at approximately 5 km resolution over Europe<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p> <a href=\"https:\/\/meetings-eu1.hubspot.com\/guett\/energy-trading?uuid=d780665f-ff71-439c-addf-c80e49af0627\" target=\"_blank\"><strong>See EPT-2 head-to-head against your current forecast provider. Book a demo.<\/strong><\/a> <\/p>\n<h2>Strategic Trade-offs for Energy-Trading Weather Data<\/h2>\n<p><strong>Accuracy versus cost.<\/strong> Free APIs such as Open-Meteo and OpenWeatherMap carry no direct data cost but deliver no hub-height wind, no SSRD, no ensemble, and no hindcast. <a href=\"https:\/\/nebius.com\/customer-stories\/jua\" target=\"_blank\">A 1 GW wind portfolio that gains four percentage points of forecast accuracy saves approximately \u20ac1.5 M per year, and a 1 GW solar portfolio at the same accuracy gain saves approximately \u20ac3 M per year<\/a>. The cost of a free API is the alpha it cannot capture.<\/p>\n<p><strong>Generality versus specialization.<\/strong> General-purpose commercial APIs such as Tomorrow.io and Visual Crossing serve consumer and IoT use cases. They are not designed for hub-height wind at multiple vertical levels, ensemble CRPS, or Apache Arrow payloads. Production energy-trading pipelines require variables and schemas that general APIs do not expose.<\/p>\n<p><strong>Research outputs versus productized platforms.<\/strong> AI weather peers such as Aurora and GraphCast are research outputs from large-company AI labs. They deliver raw model files without ensembles, hindcasts, or operational refresh schedules. Quant teams that subscribe to them must build the ingestion pipeline, ensemble logic, benchmarking harness, and hindcast access themselves. That engineering capacity is better spent on alpha research. Jua for Energy is a productized platform, and the pipeline is already built.<\/p>\n<h2>Implementation Best Practices for Quant and Trading Teams<\/h2>\n<p><strong>Live benchmarking before procurement.<\/strong> Run a head-to-head accuracy comparison on the region and variable that matters most to the book. Use the Jua platform&#8217;s benchmarking surface, which covers more than 25 models on any region and variable in seconds, instead of relying on vendor-provided graphics. The live benchmark often becomes the internal deal trigger, because meteorologists who were sceptical of vendor accuracy claims gain direct evidence when they run it themselves.<\/p>\n<p><strong>Backtesting workflows.<\/strong> Hindcast data is available across multiple Jua and third-party models via the Python SDK. A typical backtest runs in approximately 5 minutes via Athena, Jua&#8217;s AI agent, or directly through the SDK for programmatic access. Multi-year re-forecast datasets form the prerequisite for validating any systematic strategy, so teams should confirm hindcast availability and variable coverage before committing to a provider.<\/p>\n<p><strong>Integration planning.<\/strong> The REST API at <code>query.jua.ai\/docs<\/code> and the Python SDK (<code>pip install jua<\/code>) expose all models through a single schema, which removes the need to write separate parsers for each model. That unified schema enables Apache Arrow support for continental, multi-variable, multi-model payloads, so large requests do not require custom handling. For European power markets, ENTSO-E grid data integrates directly into the same schema. Because the schema remains stable across model updates, teams avoid pipeline re-engineering when Jua ships new model versions.<\/p>\n<p>Teams can run benchmarks on their own region and variables on the Jua platform. Forecasts appear in less than 5 minutes, head-to-head against more than 25 models, at <a href=\"https:\/\/athena.jua.ai\" target=\"_blank\">athena.jua.ai<\/a>.<\/p>\n<h2>Readiness Checklist and Common Pitfalls<\/h2>\n<p>Before selecting a production weather API for an energy-trading pipeline, confirm the following:<\/p>\n<ul>\n<li>Hub-height wind is available at levels relevant to the portfolio.<\/li>\n<li>Surface solar radiation is a native model output, not a derived or post-processed variable.<\/li>\n<li>Ensemble forecasts are available with published CRPS skill scores versus ECMWF ENS.<\/li>\n<li>Hindcast data covers at least two full years and is accessible programmatically via SDK or API.<\/li>\n<li>Update frequency matches the trade horizon, because intraday strategies require more than four daily refreshes.<\/li>\n<li>The API schema is stable across model updates, because schema breaks invalidate backtests and require pipeline re-engineering.<\/li>\n<li>Accuracy claims are benchmarked against ground-truth observations (station data), not against other model outputs.<\/li>\n<\/ul>\n<p><strong>Common pitfalls.<\/strong> Missing hindcast data is the most common blocker for quant teams, because a provider that cannot deliver multi-year re-forecasts cannot support backtesting, and the strategy never gets validated before capital deployment. Even when hindcast data exists, schema instability, including variable names, units, or endpoint structures that change between model versions, silently breaks the backtests themselves and invalidates months of validation work. For strategies that pass backtesting and enter production, stale refresh rates of two to four daily runs create systematic blind spots in intraday execution, because the next forecast is always hours away when the market moves. Finally, accuracy claims anchored to grid-point verification rather than station-based evaluation (StationBench methodology) overstate skill at the locations where energy assets actually sit, which means forecasts underperform in production even when the refresh rate and schema are sound.<\/p>\n<h2>Frequently Asked Questions<\/h2>\n<h3>Does Jua for Energy offer a free tier for evaluation?<\/h3>\n<p>Jua for Energy is a production-grade platform designed for utilities, physical trading houses, and quantitative funds, and it does not offer a consumer free tier. Evaluation follows a structured proof-of-value process. A live benchmark on the prospect&#8217;s own region and variable runs in seconds on the Jua platform, and a full backtest via Athena completes in approximately 5 minutes. This approach gives technical evaluators, including meteorologists and quant developers, direct, auditable evidence of forecast accuracy before any commercial commitment. The Python SDK can be installed with <code>pip install jua<\/code>, and the API is documented at <code>docs.jua.ai<\/code>. Access credentials are provisioned during the evaluation process.<\/p>\n<h3>How does Jua for Energy handle precipitation forecasts?<\/h3>\n<p>Precipitation is one of the 25 variables covered by Jua for Energy&#8217;s weather forecast surface. EPT-2, the deterministic flagship, produces precipitation forecasts at native any-\u0394t, which means forecasts are generated at arbitrary lead times rather than interpolated from fixed 6-hour model steps. For energy-trading use cases, precipitation is most directly relevant to hydro dispatch and gas demand modeling. The primary variables driving renewable-generation P&amp;L, including hub-height wind at multiple levels and surface solar radiation, are where EPT-2&#8217;s accuracy advantage over ECMWF HRES is most directly documented in the peer-reviewed technical report at arXiv:2507.09703. Precipitation skill is evaluated using the same StationBench methodology against more than 10,000 ground stations.<\/p>\n<h3>What is StationBench and why does the evaluation methodology matter?<\/h3>\n<p>StationBench is Jua&#8217;s open-source benchmarking methodology that evaluates forecast accuracy against more than 10,000 real ground-truth weather stations, with no post-processing or station fine-tuning applied to the model outputs. The methodology matters because grid-point verification, which compares model output to the nearest grid cell rather than to an actual observation, systematically overstates skill at the locations where energy assets sit. Turbines and solar farms are point assets, and their generation is determined by the atmospheric state at that specific location, not at the nearest 9 km or 25 km grid centroid. StationBench removes that gap. EPT-2&#8217;s benchmark results, which show an advantage over ECMWF HRES on 10 m wind, 100 m wind, 2 m temperature, and surface solar radiation across 0\u2013240 hour lead times, are all produced under StationBench conditions and published in the technical report at arXiv:2507.09703.<\/p>\n<h3>Can Jua for Energy integrate with existing internal trading pipelines?<\/h3>\n<p>Jua for Energy integrates with existing pipelines through a REST API with Apache Arrow payload support and a Python SDK installable via <code>pip install jua<\/code>. The single unified schema means that switching between EPT-2, ECMWF HRES, Microsoft Aurora, GFS, GraphCast, and the other models on the platform does not require re-engineering the downstream pipeline. ENTSO-E grid data integrates directly for European power-market workflows. Hindcast data is available programmatically for backtesting. Quant teams at capital-markets funds pipe Jua forecasts directly into their own systematic models, and utilities and trading houses connect to existing dispatch, risk, and trading tools. The integration that takes a quarter to build against raw research outputs stands up in days on the Jua platform.<\/p>\n<h3>How does EPT-2e compare to ECMWF ENS for probabilistic energy trading?<\/h3>\n<p>EPT-2e is Jua for Energy&#8217;s ensemble variant and delivers probabilistic forecasts that quantify the full distribution of atmospheric outcomes. It beats the 50-member ECMWF ENS mean on both RMSE and CRPS at virtually every lead time, as documented in the peer-reviewed technical report at arXiv:2507.09703. For energy-trading applications, ensemble skill translates directly into better-calibrated imbalance-cost estimates, more accurate options pricing on weather derivatives, and earlier detection of ramp events before they reprice the market. No AI weather peer, including Aurora, GraphCast, or ECMWF AIFS, ships a productized ensemble equivalent.<\/p>\n<h2>Conclusion and Next Steps for Energy-Trading Teams<\/h2>\n<p>The six dimensions that determine production-grade weather API fit, including deterministic accuracy, ensemble skill, update frequency, hindcast availability, developer ergonomics, and energy-specific variables, resolve to a single conclusion when evaluated against published benchmarks. EPT-2 beats ECMWF HRES on every lead time and on every variable that drives an energy P&amp;L. EPT-2e beats the 50-member ENS mean on RMSE and CRPS at virtually every lead time, as established in the evaluation above. EPT-2 RR refreshes up to 24 times per day. Hindcast data is available programmatically. <code>pip install jua<\/code> stands up the integration in days. Hub-height wind at multiple levels and native SSRD are production outputs, not derived approximations.<\/p>\n<p>Jua is a foundation model and agent company, and Jua for Energy is the first applied product. <a href=\"https:\/\/nebius.com\/customer-stories\/jua\" target=\"_blank\">Customers including Axpo, TotalEnergies, Statkraft, EnBW, EDF, and Hydro-Qu\u00e9bec execute daily trading decisions on the platform across five continents<\/a>. The customer base described there, which spans utilities, physical trading houses, and quantitative funds, relies on the same benchmarks and workflows outlined in this guide. The live benchmark is the deal trigger: pick a region, pick a variable, and the numbers speak.<\/p>\n<p> <a href=\"https:\/\/meetings-eu1.hubspot.com\/guett\/energy-trading?uuid=d780665f-ff71-439c-addf-c80e49af0627\" target=\"_blank\"><strong>Book a demo and see EPT-2 head-to-head against your current forecast provider.<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Jua&#8217;s EPT-2 beats ECMWF HRES on wind, solar &amp; temperature. See how top weather APIs compare for energy trading. Book a live benchmark demo today.<\/p>\n","protected":false},"author":103,"featured_media":332,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"footnotes":""},"categories":[11],"tags":[],"class_list":["post-333","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-weather-forecasting"],"_links":{"self":[{"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/posts\/333","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/comments?post=333"}],"version-history":[{"count":2,"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/posts\/333\/revisions"}],"predecessor-version":[{"id":741,"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/posts\/333\/revisions\/741"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/media\/332"}],"wp:attachment":[{"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/media?parent=333"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/categories?post=333"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/jua.ai\/articles\/wp-json\/wp\/v2\/tags?post=333"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}