Redfox Image Text Extractor
Extract structured text from images through an Aident-managed OCR Action and preserve readable layout.
Adapted fromredfox-data/redfox-communityb6d7c18Curated by @Aident
Works with
name: redfox-pdf-image-text-extractor description: "Extract structured text from images through an Aident-managed OCR Action and preserve readable layout."
Redfox Image Text Extractor
Extract structured text from images through an Aident-managed OCR Action and preserve readable layout.
Aident Action graph
- <action-tag>direct:deepseek_ocr_tools:parse_image_to_text</action-tag>: Pass a reachable image URL and preserve the extracted text without inventing unreadable content.
Use the exact Action names above. Inspect the live schema before preparing inputs. RedFoxHub Actions use Aident-managed access where available, so never request or expose a Redfox API key. Preflight every selected Action, show the quoted cost and affected target for paid or mutating work, and wait for explicit confirmation when required.
Workflow
- Confirm the requested outcome, target platform, topic or account identifiers, time window, audience, output format, and acceptance criteria.
- Translate the brief into the smallest useful Action · Parse Image To Textas guidance: 图文提取.
- Select the matching actionName variant from the live schema. Use stable IDs or canonical URLs when available, keep page sizes bounded, and do not fetch unrelated personal data.
- Run Aident preflight. Execute only the approved calls, carry returned task IDs into the matching result Action, and avoid resubmitting an accepted paid job.
- Normalize the returned records or artifacts into the requested report. Keep source URLs, platform identifiers, time windows, and engagement metrics attached to the claims they support.
- Verify completeness and freshness against the brief. State coverage gaps, empty result sets, unavailable fields, and provider errors instead of filling them with assumptions.
Redfox operating guidance
- Preserve the objective of the upstream Image Text Extractor workflow while using Aident Actions instead of local scripts or direct Redfox credentials.
- Compare like with like: use the same platform, topic, time window, and metric definition before ranking accounts or posts.
- Treat engagement counts, rankings, and trend snapshots as time-sensitive observations. Include the retrieval time and do not present them as permanent facts.
- Separate returned evidence from analysis. Label inferred themes, sentiment, benchmarks, and recommendations as analysis.
- For content generation or rewriting, ground recommendations in the retrieved examples, retain the user's brand and compliance constraints, and never copy source text closely.
- For downloads or generated media, confirm usage rights before execution and return the provider URL or task result without claiming durable storage.
Output
Return the confirmed brief, Actions and actionName variants used, query scope, source-backed findings or artifact URLs, verification evidence, and any unresolved limitation. For ranked outputs, include the metric and time window; for asynchronous work, include the terminal task status.
Attribution
Adapted with Redfox's confirmed authorization for Aident redistribution and adaptation. See for the pinned repository provenance.