What's New at Applivo

Applivo now routes AI requests through a unified orchestration layer spanning three providers — Groq, Gemini, and OpenRouter. This enables higher reliability, broader model coverage, and intelligent provider switching without any user-facing disruption.

Users can now select their preferred AI model directly from Settings. Models are grouped by plan tier with real-time availability gating.

A new routing engine determines the optimal model for every request based on user preference, plan permissions, real-time availability, and latency signals — without any manual intervention.

• User preference
• Plan permissions
• Real-time availability
• Performance signals
• Latency optimization
• Cost awareness

When a model becomes unavailable due to rate limits, provider outages, or capacity restrictions, Applivo automatically reroutes to the next best alternative. Three strategies available:

New reasoning modes control inference depth, injected into every AI request across the platform.

Four specialized generation frameworks injected into every cover letter, resume optimization, follow-up, and AI chat response.

Persistent Settings Engine

All AI preferences now survive page refresh, logout/login, and deployment restarts. Provider, model, reasoning profile, generation style, and fallback strategy are stored per-user in the database.

• Settings V2 API
• User Settings Database
• Audit Log System (full change history)
• Session Management
• Active session listing
• Logout from all devices

New users now receive 7-day Starter access with full AI features — no credit card required upfront.

Applivo now uses a two-pass AI system for job matching. Pass 1 runs a fast rule-based keyword filter using LLaMA-8B — jobs with less than 30% skill overlap are scored and skipped instantly with no LLM cost. Pass 2 sends qualified jobs to LLaMA-70B for deep structured extraction: required skills, tech stack, ATS keywords, responsibilities, match scores, and personalized recommendations.

Celery Beat fires the full pipeline every 6 hours: scrape, analyze, gate, queue, apply, notify, follow-up. A Priority Queue ensures Premium subscribers' applications are processed before lower-tier users.

• 7 named Celery queues
• In-memory min-heap backed by Redis
• Concurrency level 4
• Daily application limits enforced

AI Career Assistant Chat

Razorpay Subscription Billing

Full Razorpay payment flow integrated with HMAC-SHA256 verification. Automated 30-day subscription provisioning on successful capture.

6-Container Docker Compose Stack