Release Notes for ms-nodejs-backend — AI Profile Filling and Challenge Identity Enhancements

Overview

This release delivers a new AI-powered feature to generate and enrich user expert profiles by integrating OpenAI’s GPT-5 with web search capabilities. It further refactors AI interaction code to adopt the latest OpenAI client APIs with schema-based JSON responses, enhancing data accuracy and consistency. Additionally, it introduces a secure challenge identity sharing endpoint with permission checks, and upgrades OpenAI dependencies for improved performance and compatibility.

New Features

• Added aiFillProfile API endpoint that:

  • Automatically generates detailed expert profile descriptions for users by querying OpenAI GPT-5 with web search enabled.
  • Extracts and updates the top 5 verified user skills as expert fields.
  • Updates user social media and media-related URLs (LinkedIn, website, Twitter, Medium, video).
  • Returns clear success or error messages reflecting AI generation outcome.

• Introduced showIdentity endpoint to securely share a challenge participant’s identity with strict authorization checks ensuring only organization members or challenge creators can share identities.

Improvements / Changes

• Refactored askchatgpt method to:

  • Use new OpenAI Node.js client instead of deprecated OpenAIApi and Configuration classes.
  • Support optional web search via GPT-5 with structured JSON schema responses when requested.
  • Use completions API with GPT-4o model when web search is not required.
  • Enhance output parsing and data validation for AI responses using JSON schemas.

• Updated internal user profile updating logic to:

  • Replace direct JSON parsing with asynchronous JSON parsing of AI responses.
  • Map AI-generated skills to the platform’s expert fields, removing and recreating existing fields accurately.
  • Handle optional social media URLs gracefully, defaulting to empty strings when not found.
  • Return the updated user object directly in API responses.

• Refactored challenge-related imports and services for better modularity and maintainability, removing unused OpenAI direct calls in challenge workflow code.

• Simplified AI suggested expert retrieval logic for challenges, handling up to three suggested experts directly and enriching expert response details with availability and pricing info.

• Removed legacy logic that determined whether a challenge needed a “squad” of experts via complex AI prompts, replacing it with streamlined calls to fetch suggested experts.

Bug Fixes

• Fixed access control in challenge identity sharing to return appropriate 404 for non-existent challenges and 400 when unauthorized access is attempted.
• Fixed parsing errors and improved error handling in AI profile filling endpoint to return clear messages on failures.

Technical / Developer Notes

• OpenAI SDK dependency upgraded from version 4.52.5 to 6.6.0 for better support of GPT-5 and new response creation APIs.
• Chat interactions now use new client.responses.create() API with JSON schema support, enabling structured AI responses tailored to platform needs.
• AI-generated profile data must conform strictly to a JSON schema with required properties: description, skills array (with name, description, and since), and social URLs.
• aiFillProfile process deletes and recreates expert fields to keep user skills in sync with AI updates.
• Challenge expert response retrieval enriched with detailed availability, pricing, acceptance statuses, and reasons.

Breaking Changes

• Deprecated previous direct usage of OpenAIApi and Configuration classes; all AI calls now use the new OpenAI client and split between completions and responses APIs.
• AI profile fill responses and skills structure updated to align strictly with new JSON schema—developers should update integrations accordingly.

Acknowledgements

Thanks to the engineering team for integrating advanced AI capabilities into the user profile workflows, enhancing expert data accuracy and automating profile enrichment seamlessly.