Application health

No application

Concurrent users vs response time

P95 latencyP50 latency

Error rate during stress test

Latest stress test

Slowest endpoints at the breaking point

P95 during the first failing stage

What the sandbox saw

peak values during the run

Next: Diagnose (M4)

Diagnosis will connect these symptoms to the code that causes them and propose a fix.

Applications

Every repository you have connected, with what its most recent discovery found.

Application	Stack	Last discovery	Endpoints	Workflows	DB operations	Status

Test an application

Paste a GitHub link. StressPilot reads the code to find every endpoint and user journey, then generates the virtual users that will test it.

Repository

required

GitHub repository link Accepts a full URL, an SSH remote, or owner/repo. That does not look like a GitHub repository.

Why the source code?

Reading the code lets StressPilot find every endpoint, including ones no crawler reaches, and later trace a slow endpoint to the handler or query behind it.

Discovery is static. No requests are sent to your application.

What happens next

Clone: a shallow copy of the default branch goes into an isolated workspace.
Detect stack: framework, database, ORM, auth library and Docker setup.
Extract endpoints: every route with its handler, auth requirement and database calls.
Infer journeys: endpoints become named user workflows and weighted virtual-user scenarios you can edit.

Supported frameworks

PythonFastAPI · Flask · Django

Node.jsExpress · Fastify · NestJS · Next.js

AnyOpenAPI / Swagger file

Application map

Run discovery on a repository to see its map.

Primary user workflow

EntryNext workflow stepStatic load risk

Discovered resources

Stack

Endpoints

0

Method	Path	Auth	DB ops	Handler

Automatically generated scenarios

0 profiles

Traffic weights total 100%. Every scenario is editable before load is applied.

Scenarios

Virtual-user profiles generated from the application's code. Select a scenario to inspect its step sequence, think times and assertions.

Autonomous user scenarios

click a row to inspect steps

Scenario	Traffic %	Actions	Requests/session	Avg duration	Status

Request mix across all scenarios

Scenario health

from the last completed run

No stress runs yetPer-scenario completion rates are not tracked yet. See Test runs for stage results.

Test setup

Where the virtual users send traffic, how they sign in, and the tests generated from your scenarios.

Download tests (.zip)

Target

where load is sent

Docker sandboxStressPilot builds and runs the repo in isolated containers on this machine. Recommended. A URL you controlAn app you already run, e.g. on localhost or a staging server.

How virtual users sign in

inferred from the code · editable

Login endpoint

Credentials sent as

Username field

Password field

Wrap credentials ine.g. user sends {"user": {...}}

Token header

Sign up fresh accountsEach virtual user registers its own account first. Use test accountsAccounts that already exist in the target.

Username pattern{n} becomes 1, 2, 3… per virtual user

Password

Accounts

Functional checks

every endpoint once, then auth enforcement

Generated load test

Locust

Safety

Sandbox CPU / memory2 CPUs · 2 GB per container

Sandbox networkpublished on 127.0.0.1 only

Remote URLsownership check required

Deletesonly records the test created

Configure stress test

How hard to push, for how long, and what counts as breaking.

Load profile

stepped ramp

Peak users

VU

Stages

Each stage

s

Think time

×

Think time scales the pauses between steps: 1× is realistic human pacing, 0.5× packs twice the requests into each virtual user. A laptop can usually drive a few hundred virtual users.

Traffic distribution

Breaking-point budgets

a stage passes only if it stays inside both

P95 latency budget

ms

Error budget

%

Auto-abort at errors

%

Slow responses count against the P95 budget; errors are HTTP 4xx/5xx and connection failures. Auto-abort stops the run when errors pass the limit, or P95 passes 15 s, for three samples in a row.

Run summary

Baseline for comparison

Stress test

Users0

Requests/sec0

P95 (rolling)—

Errors0%

App CPU—sandbox only

App memory—

DB connections—Postgres only

Live load

virtual users over time

P95 latency

Error rate

Event feed

Stages

Stage	Users	P95	Err	req/s

Slowest endpoints

P95, last completed stage

Top errors

None so far.

Test runs

Every stress test for this application, newest first.

Run	Started	Duration	Peak users	Max stable	P95 at peak	Errors	Result

Bottlenecks

What limited capacity in the latest stress run, traced to the code.

Performance bottlenecks

	Issue	Component	Severity	Evidence	Status

Security findings

found in the code while tracing the bottlenecks

	Issue	Location	Severity	Status

What the app logged under load

exceptions read from the sandbox container during the run

Root cause

View on GitHub

Repairs

Fixes generated from the diagnosis. Tick the ones to apply, then replay the same workload against the patched code.

Repair verification

P95 latency by stage, before vs after

same ramp replayed

What changed

Stage by stage

P95 and errors at each load level

Stage	Users	P95 before	P95 after	Errors before	Errors after	Req/s before	Req/s after

Stress test report

Observability

Infrastructure signals StressPilot reads while a test runs. These are the series that root-cause analysis correlates against request latency.

Last 30 min

Sample data

This page is built in a later milestone. The numbers below are illustrative, not from your application.

API pods8all ready

CPU72%rising

Memory68%5.4 GB

DB connections91%146 / 160

Event loop lag88msthreshold 100 ms

Queue depth412report-worker

CPU & memory

mean across 8 pods

CPUMemory

Database connections

active of 160

pool limit

Worker heap

report-worker · not released between stages

Leak suspected

Network egress

Feed disconnected

No samples since 10:31 The metrics agent on ingest-2 stopped reporting 4 minutes ago. Latency and error data are unaffected — only egress counters are missing from this window.

Hosts

12 instances

Host	Role	CPU	Memory	Connections	Uptime	Status
api-7f4c-01	API pod	74%	69%	20/20	6d 04h	Pool full
api-7f4c-02	API pod	71%	67%	20/20	6d 04h	Pool full
api-7f4c-03	API pod	69%	64%	19/20	6d 04h	Ready
db-primary-a	Postgres 16	46%	58%	160/500	34d 11h	Ready
db-replica-b	Postgres 16	18%	41%	22/500	34d 11h	Ready
worker-3a9e-01	Report worker	52%	88%	4/20	2d 09h	Heap growth
ingest-2	Metrics agent	—	—	—	—	Unreachable

Settings

Workspace configuration for HealthSync. Changes apply to every application in this workspace unless noted.

Environments

where tests are allowed to run

Environment	Base URL	Max virtual users	Approval required
Production	https://api.healthsync.io	10,000	Two reviewers
Staging	https://staging.healthsync.io	25,000	None
Local	http://localhost:8080	500	None

Safety limits

auto-abort thresholds

Abort when error rate exceeds

% for 60s

Abort when P95 exceeds

s

Hard ceiling, all environments

VU

Production test window

Require approval for production runsTwo reviewers must approve before load is applied Apply repairs automatically in sandboxGenerate, apply and retest without waiting for review Open a pull request when a fix is verifiedNever merges — the PR waits for a human

Integrations

	GitHub company/healthsync · read + pull requests	Connected
	Slack #platform-alerts · run results and critical bottlenecks	Connected
	Datadog metric forwarding · stresspilot.* namespace	Connected
	PagerDuty not configured — critical bottlenecks stay in-app	Not connected

Team & access

4 members

Member	Role	Can run production tests	Last active
priyanshu@healthsync.io you	Owner	Yes	now
sre-oncall@healthsync.io	Admin	Yes	22 min ago
backend@healthsync.io	Engineer	Approval needed	3 hours ago
qa@healthsync.io	Viewer	No	yesterday

Danger zone

Delete HealthSync API

Removes the application, its 1,204 stored runs, every report and all scenario definitions. Reports already shared by link stop resolving. This cannot be undone.

Sign in to StressPilot

Connect a repository

Application health

Latest stress test

Applications

Test an application

Application map

Endpoints

Static risks

Automatically generated scenarios

Scenarios

Test setup

Configure stress test

Stress test

Test runs

Bottlenecks

Root cause

Repairs

Repair verification

Stress test report

Observability

Settings