Build a bounded cross-platform dataset of posts for keywords or hashtags, with count, ordering, transcripts, and safe-search controls.
Where: Flowgen → Add → Research → Content Harvest
Key ideas
- Bounded collection: Choose 25, 50, 100, or 250 items per platform and sort by Top or Recent instead of requesting an undefined scrape.
- Transcript-aware video research: Include transcripts appears only when a selected video platform supports transcript collection and is marked as an add-on.
- Dataset output: Collected rows retain post metadata and can include captions, transcripts, publish time, duration, and engagement fields when available.
- Content Harvest card: ContentHarvestNode uses the generic shared-card path. Collect a bounded dataset of posts by keyword/hashtag across selected platforms.
- Properties sidebar: researchNodes.tsx
- Opened surfaces: Structured result preview
Steps
- Enter keywords or hashtags and choose only the platforms relevant to the brief.
- Set the per-platform count and choose Top for proven examples or Recent for current material.
- Keep Safe search on unless the research explicitly requires a broader set; enable transcripts only when spoken content matters.
- Run, inspect several raw rows for relevance, then connect the dataset to Hook Miner, Performance Insights, Ask the Data, or Brain.
Content Harvest controls and presets
- Add path: Flowgen → Add → Research & Insights → Content Harvest.
- Card: ContentHarvestNode.
- Properties: researchNodes.tsx.
- Controls: Platforms; 25/50/100/250 per platform; Top/Recent; transcripts; Safe search; save dataset.
- Opened surfaces: Structured result preview.
Tips
- Test 25 items per platform before scaling to 250.
- Keep the query focused enough that Top and Recent still describe the same subject.
Limitations and important notes
- Transcript availability depends on the selected platform and item.
- Metrics are snapshots supplied by sources and may be incomplete or change later.
Troubleshooting
The collection is noisy
Use more specific keyword phrases, remove irrelevant platforms, and validate a smaller run before increasing the count.