Prompt flow is an Azure AI toolset for treating prompt engineering as an operational workflow: teams design, evaluate, test, and deploy flows that combine prompts, language models, and code. section 6.9
A practical prompt-flow workflow supports:
- Iterating on prompts against a specific task and data, while accounting for inconsistent responses, model limitations, and difficulty quantifying quality. section 6.9
- Connecting prompts with language-model calls and application code into a deployable workflow. section 6.9
- Relating prompt operations to production observability practices such as metrics, logging, and tracing. section 11.4.7
From prompt iteration to operations
A team can iterate on prompt techniques for a business task, evaluate response quality, test the resulting flow, and prepare it for deployment while recognizing prompt injection as a security concern. section 6.9
Figure 11.11 Prompt flow lifecycle
Use the prompt-flow lifecycle figure to orient the progression from prompt work toward operational use. section 11.4.7
