ChatGPT (OpenAI)

Send one or more images to gpt-4o Vision and receive a detailed interpretation, including object identification, OCR text, scene context, or anomaly flags.

Run any block of text through gpt-4o for deep analysis — sentiment, topic classification, key entity extraction, or custom rubric scoring — and return structured results.

Run any text payload through OpenAI's moderation API to detect hate speech, self-harm, sexual content, or violence before it reaches your users or downstream systems.

Maintain a stateful multi-turn dialogue with gpt-4o or gpt-4o-mini, passing the full message history so the model can answer follow-up questions with complete context.

Generate a high-dimensional vector representation of any text using text-embedding-3-large or text-embedding-3-small for semantic search, clustering, or similarity matching.

Use DALL·E's inpainting capability to modify a specific masked region of an existing image — replacing backgrounds, removing objects, or extending scenes — without touching the rest of the composition.

Transcribe audio or video files to accurate text using Whisper-large-v3, supporting 57 languages and returning word-level timestamps when requested.

Generate a short mp4 video clip from a text prompt using OpenAI's Sora video generation model, defining the scene, style, and duration.

Use gpt-4o with a JSON schema to parse unstructured text — emails, PDFs, form submissions — and return clean, typed records ready for database writes or downstream processing.

Create a new image from a text description using DALL·E 3, specifying size, style (vivid or natural), and quality level.

Fire a one-shot prompt at any OpenAI model — gpt-4o, gpt-4o-mini, o1, o3-mini — and return the raw completion, ideal for single-step generation tasks that don't need conversation history.

Transform any text into natural-sounding audio using OpenAI's TTS-1 or TTS-1-HD model, choosing from six voices and returning an mp3 or opus file.

Translate any text into a target language using gpt-4o, preserving tone and domain-specific terminology better than rule-based engines.

Compose a complete, context-aware email — subject line, greeting, body, call-to-action, and sign-off — from a brief description or set of data inputs.

Retrieve a specific past ChatGPT conversation by ID or search criteria, returning the full message thread for audit, reuse, or downstream analysis.

Look up a previously uploaded file in the OpenAI Files API by name or ID, confirming its status and returning its metadata before using it in a fine-tune or assistant run.

Retrieve a specific stored response object from the OpenAI Responses API by ID, useful for auditing, re-processing, or resuming an interrupted workflow.

Retrieve the status and download URL of a previously submitted Sora video generation job, polling until the render is complete.

Query a pre-built vector index of OpenAI embeddings to find the semantically closest items to a given query text, returning ranked results with similarity scores.

Look up an existing file in the OpenAI Files API by name; if none is found, upload it and return the new file object — ensuring idempotent file management in long-running pipelines.

Fires when a ChatGPT chat completion request returns a final response, enabling downstream steps to process the output immediately.

Fires when an OpenAI Assistants API run reaches a terminal state (completed, failed, or requires_action), allowing the agent to retrieve the output or handle tool calls.

Fires when an OpenAI Batch API job reaches completed or failed status, signalling that up to 50,000 queued completions are ready for download.

Fires when an uploaded file in the OpenAI Files API transitions to 'processed' status, confirming it is ready for use in fine-tuning or assistant retrieval.

Fires when the OpenAI account's token usage or cost crosses a configured threshold within the billing period, enabling proactive spend control.

Fires when the OpenAI Moderation API returns a category score above the configured threshold for submitted content, routing it for human review or automated action.

Fires when a fine-tuning job reaches succeeded status, confirming the custom model is deployed and available via its new model ID.