langgraph-api 0.12.0.dev11__tar.gz → 0.12.0.dev13__tar.gz

{langgraph_api-0.12.0.dev11 → langgraph_api-0.12.0.dev13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langgraph-api
-Version: 0.12.0.dev11
+Version: 0.12.0.dev13
 Author-email: Will Fu-Hinthorn <will@langchain.dev>, Josh Rogers <josh@langchain.dev>, Parker Rule <parker@langchain.dev>
 License: Elastic-2.0
 License-File: LICENSE

{langgraph_api-0.12.0.dev11 → langgraph_api-0.12.0.dev13}/benchmark/.gitignore

@@ -9,5 +9,3 @@ staircase_*.json
 
 # Node.js dependencies
 node_modules/
-capacity_summary.json
-cap-bench-results/

langgraph_api-0.12.0.dev13/benchmark/Makefile

@@ -0,0 +1,30 @@
+# Benchmark commands
+
+benchmark-ramp:
+	make benchmark-reset
+	npm install
+	npm run build
+	k6 run --out json=raw_data_$(shell date +%Y-%m-%dT%H-%M-%S).json --system-tags=[] ramp.js
+
+benchmark-charts:
+	npm install
+	node graphs.js $(shell ls -t raw_data_*.json | head -1) true
+
+benchmark-reset:
+	node clean-cli.js
+
+benchmark-new-revision:
+	node update-revision.js
+
+benchmark-staircase:
+	python staircase.py
+
+benchmark-staircase-reset:
+	python ../../scripts/reset_deployment.py -a $(BENCHMARK_APP_NAME) -n $(BENCHMARK_NAMESPACE) -db $(BENCHMARK_DB_INSTANCE)
+	python staircase.py
+
+benchmark-clean:
+	rm -f results_*.json summary_*.json raw_data_*.json staircase_*.json *_chart_*.png
+
+benchmark-clean-charts:
+	rm -f *_chart_*.png

{langgraph_api-0.12.0.dev11 → langgraph_api-0.12.0.dev13}/benchmark/README.md

@@ -20,23 +20,6 @@ This directory contains k6 + Python/Node benchmark tooling for LangGraph API.
   - **Example:**
     - `BENCHMARK_PROFILE=parallel-small START_LOAD=30 STEP_SIZE=15 NUM_STEPS=10 make benchmark-staircase`
 
-- `make benchmark-capacity` - Runs binary-search capacity tests and writes `capacity_summary.json`.
-  - **Environment variables:**
-    - `CLUSTER_NAME`
-    - `WORKLOAD_NAMES` (comma-separated) or `WORKLOAD_NAME`
-  - **Support note:** `BENCHMARK_TYPE` is supported but usually determined by profile/workload defaults.
-  - **Example:**
-    - `CLUSTER_NAME=dr-1-node WORKLOAD_NAMES=parallel-small,parallel-medium make benchmark-capacity`
-
-- `make benchmark-mixed` - Runs quick and long workloads concurrently to measure contention/fairness.
-  - **Environment variables:**
-    - `SLOWEST_N`
-    - Quick workload knobs: `QUICK_VUS`, `QUICK_ITERATIONS`, `QUICK_STEPS`, `QUICK_DATA_SIZE`, `QUICK_MAX_WAIT_SECONDS`, `QUICK_POLL_INTERVAL`
-    - Long workload knobs: `LONG_VUS`, `LONG_ITERATIONS`, `LONG_STEPS`, `LONG_DELAY`, `LONG_DATA_SIZE`, `LONG_MAX_WAIT_SECONDS`, `LONG_POLL_INTERVAL`
-  - **Support note:** Does **not** support benchmark universal/profile/context knobs (`BENCHMARK_TYPE`, `BENCHMARK_PROFILE`, `RUN_MODE`, `STREAM_RESUMABLE`, or context overrides). Only mixed-workload vars + generic `BASE_URL`/`LANGSMITH_API_KEY` apply.
-  - **Example:**
-    - `BASE_URL=http://localhost:9123 QUICK_VUS=300 LONG_VUS=8 LONG_STEPS=60 make benchmark-mixed`
-
 - `make benchmark-charts` - Builds charts from the latest `raw_data_*.json` ramp output. Only works for ramp.
 - `make benchmark-clean` - Deletes local benchmark result/chart artifacts.
 - `make benchmark-reset` - Cleans benchmark server state via `clean-cli.js`.
@@ -70,7 +53,7 @@ Profiles are quick ways to configure the benchmark graph for any run depending o
 Each profile defines:
 
 - `context` (delay/steps/checkpoint/streaming/burst knobs)
-- `capacity` defaults (`runMode`, `rampEndBase`, `runExecutionTimeoutSeconds`)
+- `runMode` default (`stateless` or `stateful`, can be overridden by `RUN_MODE`)
 - `resumable` default (can be overridden by `STREAM_RESUMABLE`)
 
 ## Core environment knobs
@@ -110,15 +93,6 @@ BENCHMARK_PROFILE=streaming-long BENCHMARK_TYPE=stream_write make benchmark-ramp
 # Staircase benchmark
 make benchmark-staircase
 
-# Capacity benchmark (single workload)
-CLUSTER_NAME=dr-1-node WORKLOAD_NAME=parallel-small make benchmark-capacity
-
-# Capacity benchmark (multiple workloads)
-CLUSTER_NAME=dr-1-node WORKLOAD_NAMES=parallel-small,parallel-medium make benchmark-capacity
-
-# Mixed workload benchmark
-make benchmark-mixed
-
 # Build charts from latest ramp raw_data file
 make benchmark-charts
 
@@ -135,16 +109,10 @@ make benchmark-clean
   - `*_chart_*.png` (when chart generation is run)
 - Staircase:
   - `staircase_summary.json`
-- Capacity:
-  - `capacity_summary.json`
-- Mixed:
-  - `mixed_workload_*.json`
 
 ## CI workflows
 
 - Daily ramp benchmark: `.github/workflows/benchmark.yml`
-- Capacity benchmark matrix: `.github/workflows/benchmark-capacity.yml`
-- Mixed benchmark matrix: `.github/workflows/benchmark-mixed.yml`
 
 ## References
 

{langgraph_api-0.12.0.dev11 → langgraph_api-0.12.0.dev13}/benchmark/benchmark-runners/benchmark_profiles.ts

@@ -32,16 +32,10 @@ export interface BenchmarkProfileEnv {
   BURST_SIZE?: string;
 }
 
-export interface CapacityProfileSettings {
-  runMode: 'stateless' | 'stateful';
-  rampEndBase: number;
-  runExecutionTimeoutSeconds: number;
-}
-
 export interface BenchmarkProfile {
   context: BenchmarkContext;
   resumable: boolean;
-  capacity: CapacityProfileSettings;
+  runMode: 'stateless' | 'stateful';
 }
 
 export interface ParsedBenchmarkProfile {
@@ -50,7 +44,6 @@ export interface ParsedBenchmarkProfile {
   runMode: 'stateless' | 'stateful';
   resumable: boolean;
   context: BenchmarkContext;
-  capacity: CapacityProfileSettings;
 }
 
 const DEFAULT_PROFILE_NAME = 'default';
@@ -58,6 +51,7 @@ const DEFAULT_PROFILE_NAME = 'default';
 export const BENCHMARK_PROFILES: Record<string, BenchmarkProfile> = {
   default: {
     resumable: false,
+    runMode: 'stateless',
     context: {
       delay: 0,
       delay_jitter_ratio: 0.2,
@@ -71,14 +65,10 @@ export const BENCHMARK_PROFILES: Record<string, BenchmarkProfile> = {
       burst_probability: 0,
       burst_size: 0,
     },
-    capacity: {
-      runMode: 'stateless',
-      rampEndBase: 200,
-      runExecutionTimeoutSeconds: 60,
-    },
   },
   'streaming-long': {
     resumable: true,
+    runMode: 'stateful',
     context: {
       delay: 0.5,
       delay_jitter_ratio: 0.2,
@@ -92,14 +82,10 @@ export const BENCHMARK_PROFILES: Record<string, BenchmarkProfile> = {
       burst_probability: 0.05,
       burst_size: 2 * 1024 * 1024,
     },
-    capacity: {
-      runMode: 'stateful',
-      rampEndBase: 20,
-      runExecutionTimeoutSeconds: 1000,
-    },
   },
   'stateless-parallel-small': {
     resumable: false,
+    runMode: 'stateless',
     context: {
       delay: 0,
       delay_jitter_ratio: 0.2,
@@ -113,14 +99,10 @@ export const BENCHMARK_PROFILES: Record<string, BenchmarkProfile> = {
       burst_probability: 0,
       burst_size: 0,
     },
-    capacity: {
-      runMode: 'stateless',
-      rampEndBase: 200,
-      runExecutionTimeoutSeconds: 60,
-    },
   },
   'stateless-parallel-medium': {
     resumable: false,
+    runMode: 'stateless',
     context: {
       delay: 0,
       delay_jitter_ratio: 0.2,
@@ -134,14 +116,10 @@ export const BENCHMARK_PROFILES: Record<string, BenchmarkProfile> = {
       burst_probability: 0,
       burst_size: 0,
     },
-    capacity: {
-      runMode: 'stateless',
-      rampEndBase: 200,
-      runExecutionTimeoutSeconds: 60,
-    },
   },
   'parallel-small': {
     resumable: false,
+    runMode: 'stateful',
     context: {
       delay: 0,
       delay_jitter_ratio: 0.2,
@@ -155,14 +133,10 @@ export const BENCHMARK_PROFILES: Record<string, BenchmarkProfile> = {
       burst_probability: 0,
       burst_size: 0,
     },
-    capacity: {
-      runMode: 'stateful',
-      rampEndBase: 200,
-      runExecutionTimeoutSeconds: 60,
-    },
   },
   'parallel-medium': {
     resumable: false,
+    runMode: 'stateful',
     context: {
       delay: 0,
       delay_jitter_ratio: 0.2,
@@ -176,11 +150,6 @@ export const BENCHMARK_PROFILES: Record<string, BenchmarkProfile> = {
       burst_probability: 0,
       burst_size: 0,
     },
-    capacity: {
-      runMode: 'stateful',
-      rampEndBase: 200,
-      runExecutionTimeoutSeconds: 60,
-    },
   },
 };
 
@@ -254,13 +223,10 @@ export function get_profile(env: BenchmarkProfileEnv): ParsedBenchmarkProfile {
   return {
     name,
     benchmarkType: env.BENCHMARK_TYPE || 'wait_write',
-    runMode: env.RUN_MODE ? (env.RUN_MODE === 'stateful' ? 'stateful' : 'stateless') : profile.capacity.runMode,
+    runMode: env.RUN_MODE ? (env.RUN_MODE === 'stateful' ? 'stateful' : 'stateless') : profile.runMode,
     resumable: env.STREAM_RESUMABLE != null && env.STREAM_RESUMABLE !== ''
       ? parseOptionalBool(env.STREAM_RESUMABLE) ?? profile.resumable
       : profile.resumable,
     context,
-    capacity: profile.capacity,
   };
 }
-
-

{langgraph_api-0.12.0.dev11 → langgraph_api-0.12.0.dev13}/benchmark/continuous/pyproject.toml

@@ -4,10 +4,10 @@ version = "0.1.0"
 description = "Continuous stress test for Agent Server"
 requires-python = ">=3.13"
 dependencies = [
-    "langgraph-sdk>=0.3.15",
+    "langgraph-sdk>=0.4.2",
     "click>=8.4.1",
-    "structlog>=24.1.0",
-    "datadog-api-client>=2.55.0",
+    "structlog>=26.1.0",
+    "datadog-api-client>=2.56.0",
     "python-dateutil>=2.8.0",
     "httpx>=0.27.0",
 ]