OpenAI Chat Assistant Embedded in Teachable – Developer Requirements Objective: Develop an embeddable chat widget (code hosted and deployed on my server) for use within Teachable. The widget must: • Simulate a chat interaction between a real student, an AI-generated instructor, and AI-generated classmates. • Use the OpenAI API (API key will be provided) to generate concise replies mimicking student/class interactions. • Match Teachable’s design (including fonts, colors, and buttons). - Reuse the user interface from an existing similar application I already have (UI assets and layout can be copied or adapted). • Log the entire conversation and generate a searchable, timestamped PDF report stored on the server. ________________________________________ Functional Requirements: Chat Behavior: • Each session must identify whether the user is the instructor or the student (only one instructor role exists). Only students can enter the chat, Open AI Assitant acts as the instructor or as a student depending on who the person who enters chat talks to. • Students must enter their email before entering the chat. This email: o Serves as an identifier for the PDF transcript. o Enables search and retrieval of the conversation later. • The conversation: o Must be saved to a PDF that includes date/time stamps aligned with a Node app already developed (used to track login times). o Must be downloadable via a separate panel/bar within the UI using the student email as the search query. o Must include time and date of each message, matching the timestamps from the external Node app (synchronization required). • Chat replies: o Must be brief (maximum 30 words per message). o Must be limited to 2–8 exchanges per session (one exchange = one student message + one response from a classmate or instructor). o Should end automatically once the exchange limit is reached or after a timeout (e.g., inactivity or simulated message like “Instructor Alex got disconnected”). Participants: • The instructor will always appear under one name (e.g., “Instructor Alex”). • Classmates should be simulated using either: o Names fetched via the Teachable API (if available). • Simulated classmates should appear and disappear (e.g., "Ana just connected", "Tom has disconnected") to mimic a dynamic class. o Provide a pool of 5–10 fake classmate names, rotating online/offline states so the conversation can be started with them. o Students should be able to initiate conversations with currently connected classmates. ________________________________________ PDF Report Generation: • Every student chat must be: o Logged. o Converted to a PDF. o Stored on the server (existing infrastructure from previous projects can be reused). • The PDF: o Must contain the entire conversation and the student’s email. o Must be searchable by email using a designated search field. o May either: Append new chats to an existing PDF for the student. Or be saved as a separate file per session.
