Generate an SEO summary of an article using HTTP REST API and OpenAI plugins

Turn long-form content into concise, search-friendly summaries automatically. This blueprint downloads an article from any public URL, sends its contents to an OpenAI chat model, and produces a short SEO-optimized summary you can drop into meta descriptions, social previews, or content indexes. It removes the manual copy, paste, and prompt loop that slows down content and marketing teams, and makes summary generation a repeatable, versioned pipeline.

How it works

1. fetch_article uses io.kestra.plugin.core.http.Download to retrieve the article from the article_url input and stores the response in Kestra internal storage.
2. generate_seo_summary uses io.kestra.plugin.openai.ChatCompletion with the gpt-4o-mini model to summarize the downloaded file. The prompt reads the file with {{ read(outputs.fetch_article.uri) }} and is capped at maxTokens: 500 for a roughly 200 word result.
3. log_output uses io.kestra.plugin.core.log.Log to print the summary, extracting the model text with {{ outputs.generate_seo_summary.choices | jq('.[].message.content') | first }}.

What you get

• A concise, SEO-oriented summary generated from any article URL.
• A clean three-step pipeline you can run on demand or schedule.
• Downloaded content captured in Kestra internal storage for reuse and debugging.
• Model output parsed and surfaced in execution logs, ready to forward downstream.

Who it's for

• Content and SEO teams automating meta descriptions and previews.
• Marketing engineers building content enrichment pipelines.
• Developers prototyping LLM-backed text processing in Kestra.

Why orchestrate this with Kestra

The OpenAI API has no scheduler, retry policy, or run history of its own. Kestra adds event and schedule triggers, automatic retries on transient HTTP or API failures, full execution lineage from download to summary, and declarative YAML you can version in Git. Inputs make the flow reusable across any article URL, and outputs flow cleanly between tasks without glue code.

Prerequisites

• A Kestra instance (io.kestra.plugin.openai and core plugins are bundled).
• An OpenAI API key with access to gpt-4o-mini.

Secrets

This sample hardcodes a placeholder apiKey for demonstration. In any real deployment, replace it with {{ secret('OPENAI_API_KEY') }} and store the value in your Kestra secret backend. Never commit a live key.

Quick start

1. Add the blueprint to your namespace.
2. Set the apiKey on generate_seo_summary to {{ secret('OPENAI_API_KEY') }} and configure that secret.
3. Execute the flow, keeping the default article_url or supplying your own.
4. Open the execution and read the summary in the log_output task logs.

How to extend

• Swap gpt-4o-mini for another OpenAI model, or tune maxTokens and the prompt for length and tone.
• Add a Schedule trigger to summarize newly published articles automatically.
• Replace log_output with a write to a database, object store, or a Slack notification.
• Loop over a list of URLs to batch-summarize an entire content backlog.

Links

• Plugin documentation: OpenAI