Compass v1.0.0-alpha

Initial alpha release of Compass, an AI-powered career guidance platform that helps users discover and articulate their skills through conversational experience collection.

Features

Authentication & User Management

• Firebase-based authentication with multiple sign-in methods
• Email/password, Google Sign-In, and anonymous authentication
• Invitation code and invitation link access control (docs)
• User session management across tabs
• Anonymous account conversion to registered account
• Email verification, resend verification link, and password reset flows

Conversational AI Experience Collection

• Multi-agent conversation system powered by Gemini 2.0
• Guided experience discovery through natural dialogue, user is guided step by step through the experience collection process
• Extract useful details from user's descriptions of their experiences (title, company, location, work type, start/end date, top skills etc)
• Skills extraction and contextualization from user descriptions of their responsibilities
• Experience summarization and structured data collection (title, company/receiver, work type, responsibilities, top skills, Q&A)
• RAG-based skills extraction using Tabiya taxonomies (seen and unseen economy) (taxonomy platform github)
• Vector search for semantic skills and occupation matching
• Skill deduplication and normalization
• Multiple conversation sessions; start new sessions at any time

CV Upload & Processing

• CV/resume upload and parsing (PDF/DOCX/TXT)
• AI-powered experience extraction from documents
• Structured data extraction and validation
• Edit parsed experiences and send updates back into the conversation

Experience Management

• Edit and refine collected experiences
  • update skills (add/remove), alternative labels, and summaries
  • update experience details (title, company, location, work type, timeline)
  • delete/restore deleted experiences
• Experience association and categorization
  • categorize experiences into one of 4 categories: Self-Employment, Salary Work, Trainee Work, and Uncategorized
• Skills Report: generate a report from your conversation and extracted skills
• Download the Skills Report in multiple formats (PDF, DOCX)

User Experience & Accessibility

• React-based frontend with Material-UI components
• Real-time chat interface with typing indicators
• Responsive design for desktop and mobile
• Accessibility compliance (WCAG 2.0 A)
• Protected pages based on authentication state
• Offline awareness: notify users when offline and disable online-only features
• Storybook for visualizing UI components without running the full app
• Reactions on AI messages (like/dislike) to collect feedback
• Consent screen for Terms & Conditions and data usage
• SEO optimization

Tests

• Frontend unit tests (Jest) and accessibility tests via Storybook
• Backend unit tests (pytest)
• Backend Integration tests for APIs and feature flows
• Evaluation tests for conversation, skills extraction, and CV processing — see Evaluation Tests Guide (docs)
  • End-to-end app: main conversation across phases and termination behavior
  • Conversation agents: Welcome (simulated/scripted), Collect Experiences, Skills Explorer (first-message, sentence decomposition, responsibilities extraction), Q&A, Agent Director routing
  • CV upload and parsing: LLM-based parser on fixtures and real files (PDF/DOCX/TXT); round-trip edits
  • Linking, inference, and ranking: occupation inference (standard + extended JSONL), contextualization, relevant-occupation classifier; skills linking tool
  • Summarization: with/without existing summary; quality/structure checks
  • Data extraction (collect experiences): field-level extraction robustness
  • ESCO search methodology and metrics: recall@k and evaluation notes
• Smoke tests for core health and version checks

Infrastructure

Multi-Environment Deployment

• Pulumi-based Infrastructure-as-Code
• Configurable multi-environment deployments
• Partner-specific environment setups with multiple environments per partner (dev, test, prod)
• Google Cloud Platform integration

Scalable Architecture

• FastAPI backend with modular design based on GCP Cloud Run
• MongoDB for data persistence
• Vector database for semantic search
• Docker containerization
• API Gateway, Cloud CDN, and load balancing

Feature Flags & Extensibility

• Feature experimentation and A/B testing framework
• Pluggable external features (e.g., Skills Ranking) via integration points
  • Available features:
    • Skills Ranking

Observability & Logging

• Sentry integration for error/warning event tracking (frontend and backend), configurable per environment
• Sentry-based user feedback channels (bug reports, invitation requests, product feedback)
• Centralized logging aggregation on Sentry, configurable per environment
• Application metrics for selected frontend/backend events
• Conversation conclusion feedback capture

Feedback

• Conversation completion feedback (prompted survey at end of session)
• Message reactions (like/dislike) on Compass responses
• Bug reports and product feedback via Sentry user-feedback channels

Internal Tooling

• Conversation and skills export scripts (backend/scripts/export_conversation, backend/scripts/export_discovered_skills)
• Embeddings management utilities (backend/scripts/embeddings)
• Conversation analysis utilities (backend/scripts/conversation_analysis)
• IaC deployment helpers (iac/scripts)

Security & Performance

• Sensitive data (PII) is encrypted in transit and at rest (docs)
• Request/response compression to improve performance on low connectivity
• Frontend bundle optimizations for faster load times in low connectivity
• Lazy loading of components where appropriate
• User data isolation: users cannot access other users' data
• Retry of API requests on transient failures

Notes

This is an alpha release. APIs and behaviors may change in future versions. See project documentation for setup and usage instructions.