Parallel Multilingual Markdown Translation with OpenAI

id: ai-multilingual-markdown-translator
namespace: company.team
description: |
  Concurrent multi-language documentation and content localization pipeline. Translates raw
  Markdown documents into Spanish, German, and French in parallel using OpenAI GPT-4o, validates
  output integrity, and delivers localized publication bundles with Slack notifications.

concurrency:
  limit: 1

inputs:
  - id: source_markdown
    type: STRING
    displayName: Source Markdown Content
    description: Raw markdown text to translate into multiple languages simultaneously.
    defaults: |
      # Product Launch: Autonomous Workflow Engine v2.0

      We are excited to announce the general availability of our next-generation orchestration platform.

      ### Key Features
      - **Sub-millisecond scheduling latency**: Execute mission-critical pipelines with deterministic precision.
      - **Zero-warehouse analytics**: Ingest and curate Parquet files in memory with embedded DuckDB.
      - **Declarative AI integrations**: Connect OpenAI, Anthropic, and Gemini models with first-class tasks.

      ### Getting Started
      Deploy our official Docker container or connect your Kubernetes cluster using our Helm chart. Visit our documentation portal for quickstart guides and configuration templates.

  - id: model
    type: STRING
    displayName: OpenAI Model
    description: OpenAI LLM identifier used for high-fidelity translation and
      technical tone matching.
    defaults: "gpt-4o-mini"

  - id: slack_channel
    type: STRING
    displayName: Slack Notification Channel
    description: Destination channel for localization publication digests.
    defaults: "#localization-updates"

triggers:
  - id: daily_localization_schedule
    type: io.kestra.plugin.core.trigger.Schedule
    description: Daily scheduled batch localization run at 03:00 UTC. Can also run
      on-demand.
    cron: "0 3 * * *"

tasks:
  - id: inspect_source
    type: io.kestra.plugin.core.debug.Return
    description: Extract document statistics and initialize localization metadata.
    format: |
      {
        "source_length_chars": {{ inputs.source_markdown | length }},
        "requested_model": "{{ inputs.model }}",
        "timestamp": "{{ execution.startDate }}"
      }

  - id: parallel_translations
    type: io.kestra.plugin.core.flow.Parallel
    description: Fan out concurrent translation requests across target languages to
      minimize latency.
    tasks:
      - id: translate_spanish
        type: io.kestra.plugin.openai.ChatCompletion
        description: Translate markdown source into fluent, technically accurate Spanish.
        apiKey: "{{ secret('OPENAI_API_KEY') }}"
        model: "{{ inputs.model }}"
        messages:
          - role: system
            content: |
              You are an expert technical translator specializing in software documentation.
              Translate the user's markdown into Spanish (Español).
              Preserve all markdown syntax, headers, code blocks, bullet points, and technical terms intact.
              Return ONLY the translated markdown with no conversational preamble or outro.
          - role: user
            content: "{{ inputs.source_markdown }}"

      - id: translate_german
        type: io.kestra.plugin.openai.ChatCompletion
        description: Translate markdown source into precise, professional German.
        apiKey: "{{ secret('OPENAI_API_KEY') }}"
        model: "{{ inputs.model }}"
        messages:
          - role: system
            content: |
              You are an expert technical translator specializing in software documentation.
              Translate the user's markdown into German (Deutsch).
              Preserve all markdown syntax, headers, code blocks, bullet points, and technical terms intact.
              Return ONLY the translated markdown with no conversational preamble or outro.
          - role: user
            content: "{{ inputs.source_markdown }}"

      - id: translate_french
        type: io.kestra.plugin.openai.ChatCompletion
        description: Translate markdown source into natural, idiomatic French.
        apiKey: "{{ secret('OPENAI_API_KEY') }}"
        model: "{{ inputs.model }}"
        messages:
          - role: system
            content: |
              You are an expert technical translator specializing in software documentation.
              Translate the user's markdown into French (Français).
              Preserve all markdown syntax, headers, code blocks, bullet points, and technical terms intact.
              Return ONLY the translated markdown with no conversational preamble or outro.
          - role: user
            content: "{{ inputs.source_markdown }}"

  - id: verify_translations
    type: io.kestra.plugin.core.flow.If
    description: Validate that all concurrent translation tasks generated complete
      markdown outputs.
    condition: "{{ outputs.translate_spanish.choices[0].message.content != null and
      outputs.translate_german.choices[0].message.content != null and
      outputs.translate_french.choices[0].message.content != null }}"
    then:
      - id: bundle_localized_content
        type: io.kestra.plugin.core.debug.Return
        description: Aggregate all localized versions into a unified multi-language payload.
        format: |
          {
            "languages": ["es", "de", "fr"],
            "translations": {
              "es": {{ outputs.translate_spanish.choices[0].message.content | json }},
              "de": {{ outputs.translate_german.choices[0].message.content | json }},
              "fr": {{ outputs.translate_french.choices[0].message.content | json }}
            }
          }

      - id: notify_slack_localization
        type: io.kestra.plugin.slack.notifications.SlackIncomingWebhook
        description: Dispatch localization completion summary and preview card to Slack.
        url: "{{ secret('SLACK_WEBHOOK_URL') }}"
        messageText: |
          *Multilingual Content Localization Complete*
          • *Source Document Size*: `{{ outputs.inspect_source.value.source_length_chars }}` characters
          • *Model Used*: `{{ inputs.model }}`
          • *Generated Locales*: Spanish (`es`), German (`de`), French (`fr`)
          • *Status*: All 3 parallel translations verified and bundled successfully.

    else:
      - id: log_translation_failure
        type: io.kestra.plugin.core.log.Log
        description: Alert on missing or empty translation responses.
        message: "Localization failure: One or more language translations returned empty
          payloads. Check OpenAI API quotas."

  - id: export_localization_manifest
    type: io.kestra.plugin.core.debug.Return
    description: Emit final publication manifest for downstream static site
      generator or CMS sync.
    format: |
      {
        "pipeline": "ai-multilingual-markdown-translator",
        "processed_at": "{{ execution.startDate }}",
        "locales_generated": 3,
        "verification_status": "{{ outputs.verify_translations.outputs != null ? 'SUCCESS' : 'FAILED' }}"
      }