OpenAI links

ChatGPT capabilities

  1. Among other things......
  2. ....
  3. Tools and Special Modes
  4. Deep research
  5. Image Input and Generation
  6. File Uploads (Documents)
  7. Data Analysis
  8. Memory
  9. Scheduled Tasks

Data analysis with Chatgpt

  1. excel
  2. pdf
  3. pandas and charts

ChatGPT Agent

  1. Visual browser for interacting with websites
  2. Code interpreter for running code and analyzing data
  3. Connectors for accessing read-only data sources
  4. Terminal for executing supported commands
  1. scheduling
  2. tools

General Help

  1. entry point for all MAJOR areas liks
  2. Enterprise
  3. Codex
  4. ChatGPT
  5. ....

Function calling explained here

Advice on building agents: New tools

  1. The new Responses API?(opens in a new window), combining the simplicity of the Chat Completions API with the tool use capabilities of the Assistants API for building agents
  2. Built-in tools including web search?(opens in a new window), file search?(opens in a new window), and computer use?(opens in a new window)
  3. The new Agents SDK?(opens in a new window) to orchestrate single-agent and multi-agent workflows
  4. Integrated observability tools?(opens in a new window) to trace and inspect agent workflow execution

Agentic Session Management

  1. part of Agents SDK docs
  2. along with streaming
  3. uses SQL dbs to maintain them

Agent streaming is documented here

Broader streaming is documented here

Webhooks is documented here

  1. Short answer
  2. ....
  3. You need to host them on web service
  4. invoked by the web call
  5. it can reply right away
  6. it can stream back
  7. it can be long running with webhooks or write to external
  8. ....
  9. Session managed by agents
  10. they run in their own servers
  11. They use tools internally including other agents..
  12. ....
  13. ChatGPT will not do the multi-agent loop
  14. Use CustomGPT that uses an HttpAction (not MCP)
  15. MCPs are limited to reads in OpenAI

On synced connectors....

  1. ...how only some data is indexed
  2. for the rest it relies on the source search