OpenAI API Integration - Supercharge your projects with AI

Learn to call the OpenAI completion API from Postman and JavaScript, set bearer token authorization, choose models like text-davinci-003, and interpret response data such as text and finish reasons.

Explore how temperature controls randomness and creativity in completions, with guidance to keep it under 1 for reliability and use top p to adjust coherence.

Leverage the n and best of parameters to generate multiple completions and pick the best by log probability per token, with examples like a joke generator and token limits.

Learn to limit costs with max tokens, structure prompts to return only needed information, and prevent early cutoffs by applying token limits in OpenAI API requests.

Enable completion streaming with server-sent events (TSS) by setting the stream to true. Receives JSON completion objects in partial messages, with a final done; a Chrome extension demonstrates practical application.

Explore how the echo parameter includes the prompt in the completion to aid debugging, and use the stop parameter to bound output with early termination.

Explore how log probs reveals token-level probabilities and how logit bias shapes completions by biasing or banning specific tokens, with practical examples and limits.

Discover how presence penalty and frequency penalty control token repetition in OpenAI completions, comparing effects across the whole text versus per line, with practical tuning cautions.

Set the user parameter to uniquely identify each user, such as a username or email, to help OpenAI monitor accounts and provide actionable feedback. The parameter doesn't affect completion results.

Explore suffix and insertion with text DaVinci three to insert completions between two text sequences, enabling template and code fill-ins, with a browser extension example overwriting part of an email.

Learn how to edit and translate text with OpenAI's edits endpoint, using input and instruction parameters and models like text-davinci-edit-1 through the v1/edits path.

Explore chat completions in OpenAI's API, using GPT 3.5 turbo, with a messages-based format to support multi-turn conversations and cost-efficient AI responses.

Create a project structure for a chrome extension using a React boilerplate, configure package and manifest metadata, install Bootstrap, and persist the OpenAI API key with Chrome local storage.

Attach an AI assistant button to Gmail's email composition forms by creating a content script, handling dynamic loading with interval retries, and updating the extension through npm build and reload.

Configure an email assistant by wiring a button to the OpenAI completion API, load and refresh the API key from Chrome local storage, and instantiate the OpenAI client.

Learn how to stream completions from the openai api by enabling stream, using server-sent events with zk js as an event source replacement, and handling authorization headers and json payloads.

Expand your email assistant by building a feature selection menu with divs, css classes, and an svg icon, then implement click handling to toggle the menu's display.

Attach a click listener to the complete button, isolate logic, toggle the assistant menu, and add a loading spinner with helpers to toggle elements and highlight errors.

Tidy up the code by moving dom helper functions into dom utils.js, add show, hide, and highlight element helpers, and fix a toggle bug from server messages.

Learn how to auto correct emails using the OpenAI edit endpoint, fixing spelling, grammar, formatting, and punctuation, and apply this through an email assistant workflow.

Discover how OpenAI's Dall-E creates high-quality images from prompts, edits with descriptions and transparency masks, and generates variations, while noting 1:1 aspect ratio and resolution limits and upgrades.

Generate images with the DALL·E endpoint by providing prompts, choosing up to ten images, and selecting 256, 512, or 1024 pixel sizes with a 1:1 aspect ratio.

Explore how to edit images with OpenAI's image edits API by applying alpha masks, preparing square PNGs, and using prompts like 'A lady holding a vase' to generate updated visuals.

Explore generating image variations with Dolly using the /v1/images/variations endpoint and form data, and adjust settings like size, variation count, and user id.

Learn how OpenAI Whisper, an API for transcribing audio with multilingual speech recognition and English translations, enables searchable transcripts for calls and YouTube videos.

Learn to create audio transcriptions using the whisper API: set up a v1 audio transcriptions request, upload a file, specify model, language, and response format, and adjust temperature for reliability.

Use the audio API to generate English transcriptions and translations from recordings in other languages with whisper, automatically detecting language, via the /v1/audio/translations endpoint, demonstrated with Bosnian audio.

Learn to handle whisper's 25MB limit by compressing audio to 128 kbps mp3, uploading to the whisper API, and splitting into chunks or using online tools.

Explore how the moderation api filters unwanted content and enforces usage policies in your projects, detecting and classifying hate, threatening, self-harm, sexual, sexual minors, and violence or graphic content.

Learn to use the moderations API by duplicating requests, setting input, and reading the flagged status and categories to detect hateful or self-harm content, while using the latest moderation model.

Discover how OpenAI completion models blend internet data and how fine tuning with training data customizes models like Da Vinci, Curie, Babbage, and Ada to improve accuracy and latency.

Prepare training data in jsonl format as prompt-completion pairs for fine-tuning, use the OpenAI CLI to preprocess data, remove duplicates, and set the stop parameter; the more examples, the better.

Create and deploy a fine-tuned OpenAI model using the CLI, upload training data to the files API, monitor progress, and customize completions with max tokens and stop conditions.

Explore how OpenAI embeddings generate vectors to measure distance between text strings and relatedness, enabling search, clustering, recommendations, anomaly detection, and classification for developers.

Generate embeddings with OpenAI in Postman using the Ada2 model, with simple inputs, and explore how embeddings visualize wine and digits datasets for clustering and classification.