[186] Implement ai multi users

[Andriamanampisoa]

What type of PR is this? (check all applicable)

• ✨ Feature
• 🛑 Bug
• ⚠️ Anomaly
• 📝 Doc
• 🎨 Style
• 🧑‍💻 Refactor
• 🛠️ Setup
• 🏗️ Build
• 🔥 Perfs
• ✅ Test
• 🔁 CI
• ⏩ Revert

Description

This pull request introduces a new "simulation context" protocol step for AI interview simulations, ensuring that company and job offer details are sent to the STS (Speech-to-Speech) microservice before audio streaming begins. This context is now required for accurate LLM prompting. The changes include protocol documentation updates, new WebSocket and backend logic to handle and relay this context, and a switch to using OpenRouter as the default LLM backend for the STS service.

Linked GitHub Ticket

Closes EpitechPromo2027/G-EIP-600-NAN-6-1-eip-tugdual.de-reviers#186

Workspace

• 🖥️ Web
• 🛠️ Server
• 🔁 CI
• 🤖 Ai
• 📱 App

[cursor]

PR Summary

High Risk
Large cross-stack change touching auth (WS tokens, internal API key), Redis capacity, and real-time audio/LLM paths; misconfiguration or slot bugs could block or overload simulations.

Overview
Adds Redis-backed simulation capacity on NestJS: concurrent slot limits, a wait queue, JWT WebSocket entrypoints, session/heartbeat/cancel APIs, and internal STS routes for session context and history sync. Interview creation now returns ready vs queued instead of always calling the AI server immediately.

Introduces simulation_context end-to-end: protocol docs, C++ WebSocket handling that forwards company/job context to STS (with ack), and STS per-interview_id briefs, turn history, and optional Nest fallback.

STS switches default LLM to OpenRouter (new client module, env-driven settings); Docker drops GPU reservation for STS by default. The web app polls the queue, shows a banner, and sends heartbeats while active.

^{Reviewed by Cursor Bugbot for commit ee7a0d0. Bugbot is set up for automated code reviews on this repo. Configure here.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
talk-up-ai-dev	Ready	Preview, Comment	Jun 24, 2026 10:41am

[railway-app]

🚅 Deployed to the TalkUp.AI-pr-142 environment in talk-up-ai

Service	Status	Web	Updated (UTC)
Backend	🚨 Crashed (View Logs)		Jun 20, 2026 at 8:48 am

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit ee7a0d0. Configure here.}

[cursor]

STS warnings stall AI server

Medium Severity

process_sts_job now ignores every STS WebSocket message lacking request_id, but the STS TTS failure path still sends a warning without that field. After a synthesis error, the AI server keeps reading until the long stream timeout instead of completing the pending chunk callback.

Additional Locations (1)

• ai/microservices/speech-to-speech/engine/stsProcess.py#L139-L148

 

^{Reviewed by Cursor Bugbot for commit ee7a0d0. Configure here.}