backend: implement GitHub platform auto-discovery from GitHub API

[ShantKhatri]

Summary

Product doc section 5.10 specifies 'Platform Auto-Discovery' — when a user enters their GitHub username, DevCard should attempt to auto-discover their linked social profiles from GitHub public metadata and pre-fill the link form.

Context

The GitHub connect OAuth flow already exists in apps/backend/src/routes/connect.ts. After a successful GitHub OAuth, the backend receives an access token. Use this to call the GitHub user API and extract linked profiles.

Tasks

• after successful GitHub OAuth in connect.ts, call GET https://api.github.com/user using the received token.
• extract blog, twitter_username, company, bio, and html_url fields.
• create a new route GET /api/connect/github/autodiscover (authenticated) that:
  • calls GitHub API with the stored token for the user.
  • returns discovered platforms as { platform: string, username: string, confidence: 'high' | 'low' }[].
  • twitter_username → twitter (high confidence).
  • blog URL → portfolio (high) or infer npm/hashnode/devto from URL pattern (low).
• do NOT auto-add links — only suggest. The client UI confirms.
• cache the discovery result for 1 hour per user in Redis (apps/backend/src/plugins/redis.ts if it exists, or add it).
• write tests mocking the GitHub API response and asserting discovery output.

Acceptance Criteria

• endpoint returns structured suggestions based on real GitHub API data.
• no links are auto-added without user confirmation.
• caching prevents redundant API calls within TTL.
• graceful handling when GitHub token is revoked or API returns 401.

Difficulty

advanced — requires external API integration, token management, pattern matching, and caching layer.

[SdSarthak]

Hi @nowbind,

I'd like to work on this issue under GirlScript Summer of Code 2026.

GitHub platform auto-discovery is a great use of structured data + pattern matching, and it's directly in my domain. I have experience with API integrations, NLP-based URL pattern inference, and building confidence-scoring pipelines.

My plan: after fetching the GitHub user object, I'd extract twitter_username, blog, company, and bio, then run the blog field through a URL pattern matcher to infer platforms like dev.to, hashnode, npm, or a personal portfolio — scoring each suggestion by confidence. Redis caching per-user will keep the GitHub API calls minimal.

Could you please assign this to me?

Thanks!

[Amritasahu04]

Hello @ShantKhatri , I’d like to work on this issue under GSSoC 2026.
I went through the issue details and studied the existing flow around connect.ts. I understand that this feature involves GitHub OAuth integration, authenticated backend routes, GitHub API consumption, profile inference logic, Redis-based caching, and test coverage.

I’m really interested in the platform auto-discovery workflow described in the product docs, and I’d love to work through the integration and implementation of this feature step by step. My approach would be to first understand the current token handling flow, then incrementally build the discovery endpoint, inference logic, caching layer, and mocked tests while ensuring the feature remains suggestion-based and non-destructive.

Please assign this issue to me if it’s available.

[ShantKhatri]

Hello @ShantKhatri , I’d like to work on this issue under GSSoC 2026. I went through the issue details and studied the existing flow around connect.ts. I understand that this feature involves GitHub OAuth integration, authenticated backend routes, GitHub API consumption, profile inference logic, Redis-based caching, and test coverage.

I’m really interested in the platform auto-discovery workflow described in the product docs, and I’d love to work through the integration and implementation of this feature step by step. My approach would be to first understand the current token handling flow, then incrementally build the discovery endpoint, inference logic, caching layer, and mocked tests while ensuring the feature remains suggestion-based and non-destructive.

Please assign this issue to me if it’s available.

Happy to assign this to you, but could you please start with good-first. Assigning this for now.

[Amritasahu04]

Hello @ShantKhatri , I’d like to work on this issue under GSSoC 2026. I went through the issue details and studied the existing flow around connect.ts. I understand that this feature involves GitHub OAuth integration, authenticated backend routes, GitHub API consumption, profile inference logic, Redis-based caching, and test coverage.
I’m really interested in the platform auto-discovery workflow described in the product docs, and I’d love to work through the integration and implementation of this feature step by step. My approach would be to first understand the current token handling flow, then incrementally build the discovery endpoint, inference logic, caching layer, and mocked tests while ensuring the feature remains suggestion-based and non-destructive.
Please assign this issue to me if it’s available.

Happy to assign this to you, but could you please start with good-first. Assigning this for now.

Working on it! I’ll start with a good-first issue as suggested and work my way up!
Thankyouu

[ShantKhatri]

Hello @ShantKhatri , I’d like to work on this issue under GSSoC 2026. I went through the issue details and studied the existing flow around connect.ts. I understand that this feature involves GitHub OAuth integration, authenticated backend routes, GitHub API consumption, profile inference logic, Redis-based caching, and test coverage.
I’m really interested in the platform auto-discovery workflow described in the product docs, and I’d love to work through the integration and implementation of this feature step by step. My approach would be to first understand the current token handling flow, then incrementally build the discovery endpoint, inference logic, caching layer, and mocked tests while ensuring the feature remains suggestion-based and non-destructive.
Please assign this issue to me if it’s available.

Happy to assign this to you, but could you please start with good-first. Assigning this for now.

Working on it! I’ll start with a good-first issue as suggested and work my way up! Thankyouu

@Amritasahu04 If you have raised any PR for this issue, then please do link that here.

[github-actions]

Hey @ShantKhatri (Project Admin) and @Harxhit (Maintainer),

This issue (previously assigned to @Amritasahu04) has been automatically unassigned because no linked pull request was found after the scheduled check.

If work is in progress, please open a PR and link it to this issue to keep the assignment.

[udaycodespace]

Hi @ShantKhatri, as per the Discord guidelines, sharing my approach before claiming this.

I see it just got unassigned. I'd like to pick this up.

Plan:

• Read the existing OAuth flow in connect.ts to understand how the token is stored
• Build GET /api/connect/github/autodiscover that calls GET https://api.github.com/user with the stored token
• Extract twitter_username (high confidence), blog (run through URL pattern matcher to infer dev.to / hashnode / npm / portfolio), html_url
• Return suggestions as { platform, username, confidence }[], no auto-adding
• Cache per user in Redis with 1hr TTL
• Handle 401 from GitHub (revoked token) gracefully
• Write tests mocking the GitHub API response

1. I've done external API integration and pattern matching in my college projects so the flow is familiar.
2. This is a step up in complexity and that's exactly why I want it.

ETA: tomorrow morning.

Can you assign this under GSSoC '26?