Report a bug

What went wrong?

OpenAI API from A to Z — AI Academy

AI Academy/Courses/OpenAI API from A to Z

⌘K

UK Sign in Sign up

IntermediateИнженерияgpt

OpenAI API from A to Z

A complete course on the OpenAI API for developers: Chat Completions, model selection (gpt-4o, o1/o3), tokens and cost, structured outputs, vision and audio, Assistants API v2, function calling, streaming, embeddings, and production patterns. Real TypeScript and Python code with current parameter names (2025).

5modules

16lessons

380 mintotal time

Developers integrating OpenAI into a productaudience

Progress

0%0 / 16 lessons

Sign in to save your progress.

Module 1

API Basics

Installing the SDK, your first Chat Completions call, choosing a model for the task and budget, counting tokens and cost.

SDK install and first call

Install the openai SDK, create a client, make your first chat completion with gpt-4o. Parse the response, finish_reason, and usage.

Choosing a model in 2025

gpt-4o, gpt-4o-mini, o1/o3 (reasoning), o4-mini — when to use which. Real cost table and selection criteria.

Tokens, context, and cost

Count tokens with tiktoken, context windows (128K for gpt-4o), truncation strategies, price-per-1M-tokens table.

Module 2

Chat Completions

System messages and sampling, structured outputs with a strict schema, gpt-4o multimodality: vision, Whisper, TTS.

System messages and sampling parameters

The system role, temperature (0 vs 0.7), top_p, frequency_penalty, presence_penalty, and seed for reproducibility.

Structured Outputs and JSON Schema

response_format: { type: 'json_schema', json_schema: {...} } with strict: true vs the old JSON mode. Validation with Zod.

Multimodality: vision, Whisper, TTS

gpt-4o with images (URL vs base64), transcription via Whisper, speech synthesis with TTS and audio streaming.

Module 3

Assistants API v2

Assistants, Threads, Runs; File Search with a vector store; Code Interpreter; combining tools and tool_choice.

Assistants, Threads, Runs

Lifecycle: create an assistant, add a message to a thread, start a run, poll for completion. Full code.

File Search and vector store

Upload a file to a vector store, attach it to an assistant, get answers with citations. Full setup in code.

Code Interpreter

What it can do (Python, charts, CSV analysis), sandbox limits, when it's worth the cost.

Combining tools and tool_choice

Parallel tool_calls, returning multiple tool results at once, tool_choice to force a call.

Module 4

Function Calling

Function definitions, parsing and executing tool_calls, parallel calls, designing reliable tool descriptions.

Function calling: the full cycle

name/description/parameters, tool_calls in the response, parsing and executing, sending results back. Full TypeScript example.

Parallel function calls

When the model calls multiple functions at once, how to execute them in parallel and merge results.

Patterns: designing reliable functions

get_weather, search_database, send_email — how to write description and parameters so the model calls them correctly.

Module 5

Production

Streaming with AbortController, embeddings and semantic search (pgvector), rate limits, backoff, cost monitoring.

Streaming responses

stream: true, iterating over chunks, parsing delta, aborting with AbortController.

Embeddings and semantic search

text-embedding-3-small vs -large, cosine similarity, semantic search, a pgvector example.

Production: limits, retries, monitoring

Rate limits (RPM/TPM), exponential backoff, the usage dashboard, cost alerts, request id for debugging.

AAI AcademyEN Start learningCatalog

Overview

AI Academy Progress Achievements Leaderboard Insights Notes Notifications Team API Pricing Settings

Learning

Tracks8 Courses51

Libraries

Prompts30 Workflows12 Tools22 Compare Blog

Updated · May 14, 2026