> For the complete documentation index, see [llms.txt](https://blackprince001.gitbook.io/flowbench-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://blackprince001.gitbook.io/flowbench-documentation/readme.md).

# FlowBench

FlowBench is a scripting-first API and flow testing toolkit: one canonical flow, authored in YAML or Python, executed by one Go engine under four execution profiles — integration, system, load/stress, and soak. It is a set of local-first tooling packages, not a hosted platform.

A **flow** chains steps by extracting values from one response and injecting them into later requests: login, take the token, act, assert. Every step, protocol phase, extraction, assertion, and retry attempt emits a span, and one span model powers both the flame graph (where aggregate time goes) and the waterfall (what exactly happened in one iteration). Rate limiting is a first-class signal — `throttled` is its own outcome class, never folded into `failed` — and a lightweight agent on the target host streams CPU and memory into the run so target saturation, generator saturation, and knee points are never confused.

Flows can also observe prompt calls: your code calls whatever LLM SDK it already uses inside a scripted step, and FlowBench wraps the call — capturing every prompt and completion, keeping prompt variants separately comparable, and diffing completions across variants and against baseline runs. Prompt changes get reviewed like code changes, not eyeballed in a playground.

## The deliverables

| Artifact                   | Language | What it is                                                                                |
| -------------------------- | -------- | ----------------------------------------------------------------------------------------- |
| `flowbench` (engine + CLI) | Go       | Parser, planner, goroutine-per-VU executor, collector, run store; results server embedded |
| `flowbench-agent`          | Go       | Streams target-host resource metrics into runs (Linux)                                    |
| `flowbench` Python SDK     | Python   | The authoring surface as an importable package; flows run like normal test files          |
| YAML DSL                   | —        | Declarative authoring surface compiling to the same canonical IR as Python                |

## Where to start

* New here: [Installation](/flowbench-documentation/getting-started/installation.md), then the [Quickstart](/flowbench-documentation/getting-started/quickstart.md) — one YAML flow to a first stress run and flame graph in under ten minutes.
* Writing flows: the [Guide](/flowbench-documentation/guide/flows.md) walks the authoring surface end to end; the [Cookbook](/flowbench-documentation/cookbook/cookbook.md) holds worked patterns, each runnable from `examples/`.
* Looking something up: the [CLI](/flowbench-documentation/reference/cli.md), [YAML DSL](/flowbench-documentation/reference/yaml-dsl.md), and [Python API](/flowbench-documentation/reference/python-api.md) references are complete.
* Why is it built this way: [Architecture](/flowbench-documentation/project/architecture.md).

One vocabulary runs through everything — a flow is never a "test case", a throttle is never an "error". The [glossary](/flowbench-documentation/project/context.md) is the source of truth for the words.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://blackprince001.gitbook.io/flowbench-documentation/readme.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.