# PromptFlow — Build and Test LLM Apps

> PromptFlow is a CLI + framework for building and testing LLM flows. Install `promptflow` + `promptflow-tools`, then run `pf flow init` and `pf flow test`.

## Install

Copy the content below into your project:

# PromptFlow — Build and Test LLM Apps

> PromptFlow is a CLI + framework for building and testing LLM flows. Install `promptflow` + `promptflow-tools`, then run `pf flow init` and `pf flow test`.

## Quick Use

1. Install:
   ```bash
   pip install promptflow promptflow-tools
   ```
2. Run:
   ```bash
   pf flow init --flow ./my_chatbot --type chat
   ```
3. Verify:
   - Run `pf flow test --flow ./my_chatbot --interactive` and confirm you can chat with the flow


---

## Intro

PromptFlow is a CLI + framework for building and testing LLM flows. Install `promptflow` + `promptflow-tools`, then run `pf flow init` and `pf flow test`.

- **Best for:** teams shipping LLM apps who want a flow-based dev loop with templates, connections, batch tests, and evaluation paths
- **Works with:** Python (repo recommends 3.9–3.11), `pf` CLI, OpenAI/Azure OpenAI connections (repo quickstart)
- **Setup time:** 14 minutes


### Quantitative Notes

- Python 3.9–3.11 recommended (repo)
- Setup time ~14 minutes
- CLI commands: `pf flow init`, `pf flow test` (repo)


---

## Practical Notes

Adopt PromptFlow as your team’s LLM dev loop: template a flow, store connections as named resources, and run interactive tests locally. Then add batch tests and evaluations so every prompt change is measurable. For agents, model each tool step as a node so you can debug failures with clear inputs/outputs per node.

**Safety note:** Keep secrets in connections, not in YAML committed to git; rotate keys and restrict scopes.

### FAQ

**Q: Do I need Azure to use PromptFlow?**
A: No. The README includes both OpenAI and Azure OpenAI connection examples; you can start locally.

**Q: Where should I use flows vs code?**
A: Use flows for repeatable LLM pipelines (prompt → tool → evaluator). Use code for custom logic and integrations.

**Q: How do I keep quality high?**
A: Use batch tests + evaluation runs; treat prompts as versioned artifacts and gate releases on eval metrics.

---

## Source & Thanks

> GitHub: https://github.com/microsoft/promptflow
> Owner avatar: https://avatars.githubusercontent.com/u/6154722?v=4
> License (SPDX): MIT
> GitHub stars (verified via `api.github.com/repos/microsoft/promptflow`): 11,122


---

<!-- ZH -->

# PromptFlow——用 Flow 构建与测试 LLM 应用

> PromptFlow 是用于构建与测试 LLM 应用的 CLI + flow 框架。安装 `promptflow` 与 `promptflow-tools` 后，可用 `pf flow init` 初始化，再用 `pf flow test` 交互测试。

## 快速使用

1. 安装：
   ```bash
   pip install promptflow promptflow-tools
   ```
2. 运行：
   ```bash
   pf flow init --flow ./my_chatbot --type chat
   ```
3. 验证：
   - Run `pf flow test --flow ./my_chatbot --interactive` and confirm you can chat with the flow


---

## 简介

PromptFlow 是用于构建与测试 LLM 应用的 CLI + flow 框架。安装 `promptflow` 与 `promptflow-tools` 后，可用 `pf flow init` 初始化，再用 `pf flow test` 交互测试。

- **适合谁（Best for）:** 要交付 LLM 应用的团队，希望用 flow 方式做开发闭环：模板、连接管理、批量测试与评估流程
- **兼容工具（Works with）:** Python（仓库推荐 3.9–3.11）、`pf` CLI、OpenAI/Azure OpenAI 连接（仓库 quickstart）
- **安装时间（Setup time）:** 14 分钟


### 量化信息

- 仓库推荐 Python 3.9–3.11
- 装机约 14 分钟
- CLI 命令：`pf flow init`、`pf flow test`（仓库）


---

## 实战要点

把 PromptFlow 当作团队的 LLM 开发闭环：用模板创建 flow，把连接配置沉淀成命名资源，本地先交互测试。再接入批量测试与评估，让每次 prompt 变更都有可量化结果。对 agent 来说，把每个工具步骤建模成节点，能用节点级输入/输出快速定位失败原因。

**安全提示：** 密钥放在 connection 里，不要写进提交到 git 的 YAML；定期轮换并收紧权限范围。

### FAQ

**Q: 必须用 Azure 吗？**
A: 不必。README 同时提供 OpenAI 与 Azure OpenAI 的连接示例；你可以本地先跑起来。

**Q: 什么时候用 flow，什么时候写代码？**
A: 可复用的 LLM 流水线（prompt → 工具 → 评估）用 flow；定制逻辑与集成用代码。

**Q: 如何保证质量？**
A: 用批量测试与评估跑分；把 prompt 当作可版本化产物，并用评估指标作为发布门槛。

---

## 来源与感谢

> GitHub：https://github.com/microsoft/promptflow
> Owner avatar：https://avatars.githubusercontent.com/u/6154722?v=4
> 许可证（SPDX）：MIT
> GitHub stars（已通过 `api.github.com/repos/microsoft/promptflow` 核验）：11,122


---
Source: https://tokrepo.com/en/workflows/promptflow-build-and-test-llm-apps
Author: Agent Toolkit