Proto commits in littlehorse-enterprises/littlehorse

These commits are when the Protocol Buffers files have changed: (only the last 100 relevant commits are shown)

2026-07-07

Merge branch 'master' into feat-bulk-job-proposal

feat(server): add LittleHorse Native Maps (#2368)

Merge branch 'master' into feat-bulk-job-proposal

2026-07-06

Merge branch 'master' into feat/add-maps-to-lh

Merge remote-tracking branch 'origin/master' into feat/workflow-migrations

2026-07-05

search and get bulk jobs

2026-07-01

Merge remote-tracking branch 'origin/master' into feat/thread-run-queue

2026-06-30

feat(server): thread run queue

feat(proto): add `InlineStructDef` type to `TypeDefinition` (#2257)

track of the last seen timestamp

2026-06-29

fix ambigous field "dependencies" and remove unnecassary comments

Merge branch 'master' into feat/proto/add-inlinestructdef-to-typedef

2026-06-25

Refactor proposal

feat(proto): add maps

2026-06-22

Workflow Migrations: rpc applyWorkflowMIgrationPlan, putWorkflowMigrationPlan, getWorkflowMIgrationPlan

2026-06-17

Track progress of a bulk job

2026-06-11

refactors

2026-06-10

Switch to node_migrations

2026-06-07

Create models for bulk job wfrun delete

Initial implementation

2026-06-03

Merge remote-tracking branch 'origin/master' into feat/workflow-migrations

rpc putWorkflowMigrationPlan, applyWorkflowMigrationPlan, GetWorkflowMIgrationPlan

2026-06-01

feat(server): Count all the NodeRun in a tenant (#2304) Extends the counted-tag proposal to include the ability to count the number of NodeRun on a single tenant.

2026-05-27

Register workflowmigration Plan with lh server and some implementation for applyworkflowmigration

2026-05-26

feat(server): Add counted tags to the existing Tag architecture (#2265) Implement proposal 021 for counted tags.

2026-05-20

feat(dashboard): overhaul WfSpec detail page layout and metadata Redesign the WfSpec header, thread selector, and thread details sections for clearer navigation and add CountNodeRun storage usage support.

2026-05-19

ci(snapshots): publish snapshots for minor versions (#2278) Adds a new `publish-snapshots` workflow that publishes both Java and Docker SNAPSHOT artifacts on pushes to `master` and release branches (e.g., `1.1`). The snapshot version is computed from `gradle.properties` by extracting only the `major.minor` portion: - `master` (`version=1.2.0-SNAPSHOT`) → publishes `1.2-SNAPSHOT` - `1.1` branch (`version=1.1.1-SNAPSHOT`) → publishes `1.1-SNAPSHOT` Co-authored-by: Eduwer Camacaro <eduwerc@gmail.com>

ci(snapshots): publish snapshots for minor versions (#2278) Adds a new `publish-snapshots` workflow that publishes both Java and Docker SNAPSHOT artifacts on pushes to `master` and release branches (e.g., `1.1`). The snapshot version is computed from `gradle.properties` by extracting only the `major.minor` portion: - `master` (`version=1.2.0-SNAPSHOT`) → publishes `1.2-SNAPSHOT` - `1.1` branch (`version=1.1.1-SNAPSHOT`) → publishes `1.1-SNAPSHOT` Co-authored-by: Eduwer Camacaro <eduwerc@gmail.com>

2026-05-06

feat(proto): add InlineStructDef to TypeDef defined_type

2026-04-29

feat(server): add POW operation to WfSpec DSL (#2243)

2026-04-27

feat(server): implement `Quota` proposal (#2141) - Creates `rpc PutQuota`, `rpc DeleteQuota`, `rpc GetQuota`, and `rpc SearchQuota` - Adds intelligent server-side throttling with `RESOURCE_EXHAUSTED` - Adds client interceptors to gracefully wait and then retry upon `RESOURCE_EXHAUSTED` - Introduces `LHC_GRPC_RESOURCE_EXHAUSTED_RETRY` configuration

2026-04-24

feat(server): add SizeOf unary operation for Strs, Arrays, JsonArrs (#2230)

2026-04-23

feat(server): allows building structs inside the wfspec dsl (#2224) - Adds `StructBuilder` option to `VariableAssignment` - Implements it in the Java SDK

2026-04-17

feat(server): add `is_nullable` flag for StructFieldDefs (#2215)

2026-04-15

feat(server): GetMetricWindow and SearchWfMetricWindow RPCs (#2204) Metrics now support tags for easier searching. Added `GetMetricWindow` to fetch a metric window by ID and `SearchWfMetricWindow` to find workflow metric windows by name and optional time range.

2026-04-14

feat(server): Tenant aggregated metrics (#2171) Adds tenant-level aggregated metrics alongside existing workflow/task metrics. Each punctuator now produces both scoped and global aggregations, enabling queries that return totals across all workflows or tasks.

2026-04-01

feat(server): Add native `ARRAY` type (#2179)

2026-03-24

feat(server): Workflow metrics (#2071) Adds workflow and task metrics to the server. Metrics are collected per WfSpec and per TaskDef in 1-minute windows and exposed via a gRPC RPC to query metrics for a specific WfSpec or TaskDef.

2026-03-11

docs(proto): add comments to all RPCs, Messages, Enums (#2140)

docs(proto): add comments to all RPCs, Messages, Enums (#2140)

2026-03-06

fix(server): fix and refactor Struct field masking (#2101)

feat(server): Edge condition refactor (#2080)

2026-03-03

feat(server): Archive Inactive ThreadRuns (#2044)

2026-02-24

fix(server): Fix pagination for List rpc (#2027)

2026-02-13

feat(server): allow waking up parent wfRun when child is rescued (#2074)

2026-02-12

First iteration: -Add new proto structures - Remove dependency for WorkflowCondition

2026-02-10

migrations

2026-02-06

feat(server): Support using WfRunVariable as input for RunChildWfNode (#2045) This PR extends the RunChildWfNode to support using WfRunVariable as the WfSpec name when running a child workflow. This allows the child WfSpec to be determined dynamically at runtime instead of being fully specified at WfSpec compile time. Changes include: Support for resolving the child WfSpec from a VariableAssignment instead of only from a static RunChildWfNode .wfSpecName. New backward compatible field for RunChildWfNode Existing behavior using a static wfSpecName continues to work unchanged. Validation of child workflow inputs/outputs is not executed at WfSpec compile time when the spec is provided dynamically.

2026-02-04

fix(server): save timer iterator progress and isolate timers (#2042) This PR fixes two problems with the Timer Processor in the core topology. - We iterate from the beginning of the timer store in the first punctuation. When creating a new processor (eg. after a rebalance) that involves iterating over tombstones, and that has been observed to take more than 60 seconds, which increases our downtime window. - We were not isolating LHTimer's with the Storeable infrastructure but rather storing them as if they were the only thing in the core store. Thus when we put something inside the cluster-scoped store, we get problems as the timer iterator sees those things as timers. We'll also need to guard the Tenant names to prevent them from starting with a digit, as that could result stuff in the tenant-scoped stores being mistaken for a cluster-scoped timer. This PR also disables the timer topology by default.

2026-01-26

fix(server): save timer iterator progress and isolate timers (#2042) This PR fixes two problems with the Timer Processor in the core topology. - We iterate from the beginning of the timer store in the first punctuation. When creating a new processor (eg. after a rebalance) that involves iterating over tombstones, and that has been observed to take more than 60 seconds, which increases our downtime window. - We were not isolating LHTimer's with the Storeable infrastructure but rather storing them as if they were the only thing in the core store. Thus when we put something inside the cluster-scoped store, we get problems as the timer iterator sees those things as timers. We'll also need to guard the Tenant names to prevent them from starting with a digit, as that could result stuff in the tenant-scoped stores being mistaken for a cluster-scoped timer. This PR also disables the timer topology by default.

Merge branch 'master' into fix/threadruns-archive

Merge branch 'feature/archive-threadruns' into fix/threadruns-archive

2026-01-25

feat(server): allow different strategies in `WaitForThreadsNode` (#2040) Implements #2037 according to Proposal 13.

2026-01-23

feat(proto): add inactivethreadrun proto

2026-01-09

remove unused proto / mock rpc getarchivedthreadrun

2026-01-07

Merge branch 'master' into fix/threadruns-archive

2026-01-05

Merge branch 'master' into metrics

refactor(server): Remove timer topology (#1979) Introduces ServerTopologyV2, unifying core logic and timer processing into a single Kafka Streams topology behind the LHS_X_ENABLE_TIMER_STREAMS feature flag. This removes the separate timer topology, reducing operational complexity, resource usage, and inter-topic communication. Timers are stored and executed within the core pipeline using a shared state store with smart routing and repartitioning only when required.

2025-12-22

feat(sdk-java): add description to LHTaskMethod annotation (#1989)

fix(server): Improve how node outputs are retrieved (#1975) This PR adds a new NODE_OUTPUT storeable that saves node outputs by node name, enabling direct and faster retrieval. Only outputs from nodes referenced in the thread spec are stored to optimize storage. Backward compatibility is preserved by falling back to the legacy backward lookup for existing workflows.

feat(server): add descriptions to TaskDefs (#1990)

2025-12-19

add migration variables to proto, allow migration request to provide variableValues

2025-12-17

Init rough draft, base case migrates, still need variables and threading

WIP add archivedthreadrunid model

2025-12-09

feat(proto): add archivedthreadrun

2025-12-01

wip

2025-11-26

Merge branch 'master' into metrics

2025-11-13

perf(server): put getables for a single wfrun under same prefix (#1842) Implement index locality optimization for workflow runs - Grouping Getables and also and WfRunStoredInventory under a shared prefix - Introducing new key formats with backward compatibility - Old running workflows can be completed in the new format

2025-11-11

perf(server): allow safe deletion of massive WfRun's (#1924) - Allows deleting large WfRun's by breaking the deletion into multiple Commands to prevent blocking the StreamThread. - Introduces LHS_X_MAX_DELETES_PER_COMMAND which controls the size of chunks.

perf(server): allow safe deletion of massive WfRun's (#1924) - Allows deleting large WfRun's by breaking the deletion into multiple Commands to prevent blocking the StreamThread. - Introduces LHS_X_MAX_DELETES_PER_COMMAND which controls the size of chunks.

2025-10-29

Merge branch 'master' into perf/index-locality

2025-10-22

fix(server): default to JSON_OBJ when typeDef defined case is not set (#1905) This fixes an incompatibility issue with old clients (< 0.14.1) that expect an undefined `TypeDefinition.primitive_type` to serialize to `JSON_OBJ`.

The documentation is generated from this commit.

fix(server): dodge tombstones on rehydration (#1902) See this documentation in RocksDB: https://github.com/facebook/rocksdb/wiki/Implement-Queue-Service-Using-RocksDB When doing a range scan in rocksdb, if there were keys recently deleted from that range, then you end up iterating over tombstones for all of those keys. In the case of a queue-like service, you may have many (potentially 100k to 1M+) tombstones. Assuming it takes 2 hours for a tombstone to be fully compacted down to the bottom level (which is totally reasonable in the case of significant state > 20GB), if you have 50 tasks per second per partition then you can end up with 50 * 3600 * 2 == 360,000 tombstones in the range. This is because for every `TaskRun` we end up with a new `Schedule In our Kafka Streams Processor `init()` method, we do a full prefix scan over `ScheduledTaskModel` store. In case of high per-partition load, this can take a very long time, causing the producer to get fenced which is a big orzdash. This PR solves the problem with two changes: * Storing `ScheduledTask`s in chronological order and processing them in chronological order. * Remembering the timestamp of the last-processed ScheduledTask * In the `init()` method, starting the iteration from the timestamp of the last processed `ScheduledTask` so as to skip over the tombstones in the range.

fix(server): default to JSON_OBJ when typeDef defined case is not set (#1905) This fixes an incompatibility issue with old clients (< 0.14.1) that expect an undefined `TypeDefinition.primitive_type` to serialize to `JSON_OBJ`.

The documentation is generated from this commit.

2025-10-21

fix(server): dodge tombstones on rehydration (#1902) See this documentation in RocksDB: https://github.com/facebook/rocksdb/wiki/Implement-Queue-Service-Using-RocksDB When doing a range scan in rocksdb, if there were keys recently deleted from that range, then you end up iterating over tombstones for all of those keys. In the case of a queue-like service, you may have many (potentially 100k to 1M+) tombstones. Assuming it takes 2 hours for a tombstone to be fully compacted down to the bottom level (which is totally reasonable in the case of significant state > 20GB), if you have 50 tasks per second per partition then you can end up with 50 * 3600 * 2 == 360,000 tombstones in the range. This is because for every `TaskRun` we end up with a new `Schedule In our Kafka Streams Processor `init()` method, we do a full prefix scan over `ScheduledTaskModel` store. In case of high per-partition load, this can take a very long time, causing the producer to get fenced which is a big orzdash. This PR solves the problem with two changes: * Storing `ScheduledTask`s in chronological order and processing them in chronological order. * Remembering the timestamp of the last-processed ScheduledTask * In the `init()` method, starting the iteration from the timestamp of the last processed `ScheduledTask` so as to skip over the tombstones in the range.

2025-10-19

fix(server): dodge tombstones on rehydration See PR description for longer explanation.

2025-10-15

feat(server): Add `LHPath` for accessing Struct fields at runtime (#1882)

feat(server): introduce durable execution through `Checkpoint` in `TaskRun` (#1881) Implements #1765

Merge branch 'master' into feat/checkpointed-tasks

feat(server): run child wf from inside `WfSpec` (#1808) Implements #1886

docs(server): add comments deprecating jsonpath

2025-10-14

feat(server): Add LHPath indexing and type safety

2025-10-13

feat(server): Add `LHPath` to `VariableMutation`s

feat(server): Add LHPath for VariableAssignments

2025-10-10

feat(proto): add LHPath

2025-10-09

fix(server): forgot to add total checkpoints to scheduledTask

fix(server): forgot to add PutCheckpointRequest to Command proto.

feat(server): adds Checkpoint as a Getable

feat(server): adds checkpointed task proposal and protobuf

Merge branch 'master' into child-workflows

2025-10-08

fix(server): Remove partial InlineArrayDef support (#1878)

feat(server): Add supporf for casting on primitive types (#1806) Introduces full support for automatic and manual type casting for Primitive Types in LittleHorse workflows , including server-side casting for variable assignments and task inputs, explicit casting methods in the Java SDK (.castToInt(), .castToDouble(), etc.).

2025-10-03

Merge branch 'master' into child-workflows

2025-10-02

some progress

2025-10-01

Merge branch 'master' into casting-implementation

2025-09-30

feat(server): add structs to `server` and `sdk-java` (#1796)

Merge branch 'master' into casting-implementation

2025-09-22

Merge branch 'master' into feat/server/structs

2025-09-18

fix(server): uuid in timer key (#1838) Previously the key for a timer in the timer store was {maturation-time}/{topic}/{partitionkey}/{tenant}. The partitionKey is always the WfRunId. Therefore if a single WfRun had two timers maturing at the same time, the second one would override the first (same key). This backwards-compatible PR changes the timer keys to be {maturation-time}/{uuid}. This will also have the benefit of shortening the keys which improves rocksdb performance.

2025-09-16

feat(sdk-go): Add StructDef get, search, delete (#1833)