hive-router 0.2.0 (2026-08-19)

Breaking Changes

Merge the hive-router-query-planner, hive-router-plan-executor, hive-router-config, and hive-router-internal crates into the main hive-router crate. They are no longer published as separate crates on crates.io.

If you just run the router (binary or Docker image), or write plugins against the hive-router crate: nothing changes for you. The plugin API (hive_router::plugins::*, RouterPlugin, hooks, and everything re-exported at the crate root) is unaffected.

If you depend directly on one of the four merged crates, you'll need to switch to depending on hive-router instead and update your imports:

Before	After
hive-router-query-planner (Cargo dep)	hive-router
hive_router_query_planner::...	hive_router::query_planner::...
hive-router-plan-executor (Cargo dep)	hive-router
hive_router_plan_executor::...	hive_router::executor::...
hive-router-config (Cargo dep)	hive-router
hive_router_config::...	hive_router::config::...

hive-router-internal has no public replacement — it was never meant for use outside of the router itself, and is now a private module.

The four crates' existing published versions on crates.io are untouched, but they won't receive any further releases.

hive-router 0.1.2 (2026-08-18)

Features

Track per-subgraph call durations on the request summary

The request summary now tracks subgraph_calls_duration, a map of subgraph name to the durations of every call made to it during the request.

This is available to custom plugins via get_current_summary(), alongside the existing subgraph tracking fields.

Fixes

Fix JWT audience validation rejecting tokens when audiences is not configured.

Previously, any token containing an aud claim was rejected with a 403 error even though audience validation is documented to be skipped when audiences is left empty.

Support { from_env: "VAR" } for any primitive config value

Any primitive field in the router config (strings, numbers, booleans, and the existing "either/or" fields like retry toggles or single-or-multiple lists) can now be set from an environment variable instead of a literal value:

http:
  port:
    from_env: PORT

If the referenced environment variable is not set, the field falls back to its default (or fails validation as usual for required fields), and a warning is logged once the router's logger has started up.

You may also add an inline fallback can also be given with default, which is used instead of the field's own default when the environment variable is unset:

http:
  port:
    from_env: PORT
    default: 4000

hive-router-plan-executor 9.1.0 (2026-08-18)

Features

Track per-subgraph call durations on the request summary

The request summary now tracks subgraph_calls_duration, a map of subgraph name to the durations of every call made to it during the request.

This is available to custom plugins via get_current_summary(), alongside the existing subgraph tracking fields.

Fixes

Support { from_env: "VAR" } for any primitive config value

Any primitive field in the router config (strings, numbers, booleans, and the existing "either/or" fields like retry toggles or single-or-multiple lists) can now be set from an environment variable instead of a literal value:

http:
  port:
    from_env: PORT

If the referenced environment variable is not set, the field falls back to its default (or fails validation as usual for required fields), and a warning is logged once the router's logger has started up.

You may also add an inline fallback can also be given with default, which is used instead of the field's own default when the environment variable is unset:

http:
  port:
    from_env: PORT
    default: 4000

hive-router-internal 0.1.2 (2026-08-18)

Features

Track per-subgraph call durations on the request summary

The request summary now tracks subgraph_calls_duration, a map of subgraph name to the durations of every call made to it during the request.

This is available to custom plugins via get_current_summary(), alongside the existing subgraph tracking fields.

Fixes

Support { from_env: "VAR" } for any primitive config value

Any primitive field in the router config (strings, numbers, booleans, and the existing "either/or" fields like retry toggles or single-or-multiple lists) can now be set from an environment variable instead of a literal value:

http:
  port:
    from_env: PORT

If the referenced environment variable is not set, the field falls back to its default (or fails validation as usual for required fields), and a warning is logged once the router's logger has started up.

You may also add an inline fallback can also be given with default, which is used instead of the field's own default when the environment variable is unset:

http:
  port:
    from_env: PORT
    default: 4000

hive-router-config 0.1.14 (2026-08-18)

Features

Support { from_env: "VAR" } for any primitive config value

Any primitive field in the router config (strings, numbers, booleans, and the existing "either/or" fields like retry toggles or single-or-multiple lists) can now be set from an environment variable instead of a literal value:

http:
  port:
    from_env: PORT

If the referenced environment variable is not set, the field falls back to its default (or fails validation as usual for required fields), and a warning is logged once the router's logger has started up.

You may also add an inline fallback can also be given with default, which is used instead of the field's own default when the environment variable is unset:

http:
  port:
    from_env: PORT
    default: 4000

hive-router 0.1.1 (2026-08-16)

Features

Isolate plugin-selected supergraph configuration

Plugin-selected supergraphs no longer inherit graph-bound settings from the router's configured supergraph. Requests, WebSocket connections, persisted-document resolution, subgraph execution, usage reports, and Hive traces now use the options attached to the selected Supergraph snapshot.

Persisted-document reloaders and Hive usage agents now use separate background-task groups scoped to the selected supergraph runtime. Cancelling a runtime waits for any active Hive flush and explicitly flushes the remaining report buffer before removing its worker. Router shutdown also waits for these graceful background tasks to finish.

Migration

Configured supergraphs require no YAML changes. The router continues deriving their graph-bound options from the existing configuration.

Plugins that construct supergraph variants must import SupergraphOptions, provide the graph-bound settings each variant needs, and retain the owner while it remains selectable:

use std::sync::Arc;

use hive_router::plugins::hooks::on_supergraph_load::{Supergraph, SupergraphOptions};

let mut options = SupergraphOptions::default();
options.traffic_shaping.all.forward_operation_name = true;
options.error_masking.redacted_error_message = "Variant error".to_string();
options.hive_target = Some("organization/project/variant".to_string());

let variant = Arc::new(Supergraph::from_sdl(sdl, options)?);

Move any existing QueryPlannerOptions into SupergraphOptions::query_planner. Also copy every graph-specific setting the variant previously inherited from router configuration, including subgraph traffic shaping, URL overrides, headers, override labels, demand control, subscription transports, error masking, persisted documents, and its Hive target. Omitted fields use SupergraphOptions::default() and no longer inherit values from the configured supergraph.

Fixes

Add persisted documents support to WebSocket operations

WebSocket subscribe payloads can now omit the GraphQL query and provide a persisted document ID through their extensions. The router resolves the document before parsing, validation, and execution.

Persisted-document extraction, ID enforcement, resolution, metrics, and missing-ID logging use the supergraph selected for the WebSocket connection. This prevents an operation from resolving a document from one supergraph's manifest and executing it against another supergraph.

Fix http.route lable value in metrics

The http.route label value in metrics was set to the effective route path, instead of the route template.

In setups with http.graphql_endpoint is used, or when persisted documents are used over HTTP paths, this led to high cardinality in metrics.

Thanks praguevara for contributing.

Fix http.route lable value in traces

The http.route span attribute was set to the effective route path, instead of the route template.

The OTEL specification for HTTP spans describes the http.router field as:

The matched route template for the request. This MUST be low-cardinality and include all static path segments, with dynamic path segments represented with placeholders.

Thanks praguevara for contributing.

Refactor internal HiveRouterConfig

HiveRouterConfig can now be treated as '&static (by calling .into_static()). This makes the work with the config struct easier, as it can be used directly without worrying about lifetimes, and without cloning.

Run GraphQL parameters hooks for WebSocket operations

WebSocket subscribe payloads now run the on_graphql_params start hook and its registered end callbacks before parsing, validation, and execution. The start hook receives the synthetic WebSocket operation request and the already-decoded parameters in OnGraphQLParamsStartHookPayload.graphql_params.

When a hook ends preparation with an early response, the router sends the response's GraphQL body as a WebSocket next message followed by complete. HTTP-only response metadata, including its status and headers, cannot be represented by the GraphQL over WebSocket protocol and is not forwarded.

Stop cancellation from interrupting an active usage flush

The usage agent now observes cancellation while waiting for the next flush interval instead of checking only after the full interval has elapsed. Cancellation remains pending while an active flush completes, preventing a drained report batch from being lost when its send future is interrupted.

Bind graph-specific configuration to Supergraph

Supergraph::from_sdl and Supergraph::from_document now accept SupergraphOptions instead of QueryPlannerOptions. The immutable options snapshot includes planner, executor, subgraph subscription, persisted-document, error-masking, and Hive target settings.

Migration

Import SupergraphOptions with Supergraph:

use hive_router::plugins::hooks::on_supergraph_load::{Supergraph, SupergraphOptions};

Replace the query-planner options argument with a complete supergraph options value:

-use hive_router::plugins::hooks::on_supergraph_load::Supergraph;
+use hive_router::plugins::hooks::on_supergraph_load::{Supergraph, SupergraphOptions};
 use hive_router::query_planner::planner::QueryPlannerOptions;

 let query_planner = QueryPlannerOptions {
     experimental_abstract_type_folding: true,
 };
-let supergraph = Supergraph::from_sdl(sdl, query_planner)?;
+let supergraph = Supergraph::from_sdl(
+    sdl,
+    SupergraphOptions {
+        query_planner,
+        ..SupergraphOptions::default()
+    },
+)?;

Callers that used Default::default() can migrate directly:

use hive_router::plugins::hooks::on_supergraph_load::{Supergraph, SupergraphOptions};

let supergraph = Supergraph::from_sdl(sdl, SupergraphOptions::default())?;

hive-router-plan-executor 9.0.0 (2026-08-16)

Breaking Changes

Bind graph-specific configuration to Supergraph

Supergraph::from_sdl and Supergraph::from_document now accept SupergraphOptions instead of QueryPlannerOptions. The immutable options snapshot includes planner, executor, subgraph subscription, persisted-document, error-masking, and Hive target settings.

Migration

Import SupergraphOptions with Supergraph:

use hive_router::plugins::hooks::on_supergraph_load::{Supergraph, SupergraphOptions};

Replace the query-planner options argument with a complete supergraph options value:

-use hive_router::plugins::hooks::on_supergraph_load::Supergraph;
+use hive_router::plugins::hooks::on_supergraph_load::{Supergraph, SupergraphOptions};
 use hive_router::query_planner::planner::QueryPlannerOptions;

 let query_planner = QueryPlannerOptions {
     experimental_abstract_type_folding: true,
 };
-let supergraph = Supergraph::from_sdl(sdl, query_planner)?;
+let supergraph = Supergraph::from_sdl(
+    sdl,
+    SupergraphOptions {
+        query_planner,
+        ..SupergraphOptions::default()
+    },
+)?;

Callers that used Default::default() can migrate directly:

use hive_router::plugins::hooks::on_supergraph_load::{Supergraph, SupergraphOptions};

let supergraph = Supergraph::from_sdl(sdl, SupergraphOptions::default())?;

Fixes

Fix http.route lable value in metrics

The http.route label value in metrics was set to the effective route path, instead of the route template.

In setups with http.graphql_endpoint is used, or when persisted documents are used over HTTP paths, this led to high cardinality in metrics.

Thanks praguevara for contributing.

Fix http.route lable value in traces

The http.route span attribute was set to the effective route path, instead of the route template.

The OTEL specification for HTTP spans describes the http.router field as:

The matched route template for the request. This MUST be low-cardinality and include all static path segments, with dynamic path segments represented with placeholders.

Thanks praguevara for contributing.

Refactor internal HiveRouterConfig

HiveRouterConfig can now be treated as '&static (by calling .into_static()). This makes the work with the config struct easier, as it can be used directly without worrying about lifetimes, and without cloning.

Stop cancellation from interrupting an active usage flush

The usage agent now observes cancellation while waiting for the next flush interval instead of checking only after the full interval has elapsed. Cancellation remains pending while an active flush completes, preventing a drained report batch from being lost when its send future is interrupted.

hive-router-internal 0.1.1 (2026-08-16)

Fixes

Fix http.route lable value in metrics

The http.route label value in metrics was set to the effective route path, instead of the route template.

In setups with http.graphql_endpoint is used, or when persisted documents are used over HTTP paths, this led to high cardinality in metrics.

Thanks praguevara for contributing.

Fix http.route lable value in traces

The http.route span attribute was set to the effective route path, instead of the route template.

The OTEL specification for HTTP spans describes the http.router field as:

The matched route template for the request. This MUST be low-cardinality and include all static path segments, with dynamic path segments represented with placeholders.

Thanks praguevara for contributing.

Refactor internal HiveRouterConfig

HiveRouterConfig can now be treated as '&static (by calling .into_static()). This makes the work with the config struct easier, as it can be used directly without worrying about lifetimes, and without cloning.

Stop cancellation from interrupting an active usage flush

The usage agent now observes cancellation while waiting for the next flush interval instead of checking only after the full interval has elapsed. Cancellation remains pending while an active flush completes, preventing a drained report batch from being lost when its send future is interrupted.

hive-router-config 0.1.13 (2026-08-16)

Fixes

Refactor internal HiveRouterConfig

HiveRouterConfig can now be treated as '&static (by calling .into_static()). This makes the work with the config struct easier, as it can be used directly without worrying about lifetimes, and without cloning.

hive-console-sdk 0.3.20 (2026-08-16)

Fixes

Stop cancellation from interrupting an active usage flush

The usage agent now observes cancellation while waiting for the next flush interval instead of checking only after the full interval has elapsed. Cancellation remains pending while an active flush completes, preventing a drained report batch from being lost when its send future is interrupted.