batch-job-poller

# Batch Job Poller (Supplemental — Recovery Only) ## Status (as of 2026-05-21) This skill is **no longer part of the standard image flow**. The new `openai-image-generator` skill consolidates submit + poll + download + realtime fallback into a single synchronous call via `openai_image_pipeline.py`. This skill remains for two recovery scenarios only: ### Scenario 1: Orphan batch recovery If a prior pipeline run gave up on a batch (timeout, crash, user cancellation), the batch may still complete in OpenAI's queue. Use this skill to: - Poll the batch status - Download images if `completed` - Read error file if `failed` for root-cause diagnosis - Mark the batch as abandoned if `expired` (24h) Example: ```bash # Find orphan batches across all workspaces for status_file in memory/workspace/*/batch_status.json; do batch_id=$(python -c "import json,sys; print(json.load(open('$status_file'))['batch_id'])") python -m scripts.openai.openai_batch_image_api --json poll $batch_id done # Download a specific orphan batch's images python -m scripts.openai.openai_batch_image_api download batch_xxx \ --output-dir memory/workspace/{task_id}/images ``` ### Scenario 2: Diagnose stuck or failed batches If a current batch is reporting `failed`, get the error_file_id and read it: ```bash python -c " import sys sys.path.insert(0, '.') from scripts._core import credential_hub import openai client = openai.OpenAI(api_key=credential_hub.get_credential('openai')) batch = client.batche...