# CV-fit Product Spec

## Positioning

CV-fit is a personal application copilot. It helps a user understand how well their profile fits a job, where the profile is weak, and what to change before applying.

It should not start as an auto-apply bot. Trust, evidence, and manual review are part of the core value proposition.

## Product Layers

### Layer 1: Core Analysis
- ingest profile or CV
- ingest a single job posting
- extract requirements
- compare requirements against profile evidence
- present score, gaps, and concrete adaptation guidance

### Layer 2: Guided Optimization
- rewrite or improve CV bullets
- suggest missing evidence to add
- tailor summary and positioning for a target job
- prepare an application package for manual review

### Layer 3: Continuous Scouting
- sync profile from LinkedIn
- poll job sources against saved criteria
- scan emails for job links
- notify user about strong matches
- identify recurring profile weaknesses

## MVP Scope

### User Inputs
- CV upload or pasted profile text
- job description text or URL pasted manually
- optional target language

### Core Outputs
- fit score
- must/should/could requirement mapping
- gap analysis
- recommendations for CV changes
- radar diagram for visual overlap

### Main Screens

#### 1. Landing / Dashboard
- explain product value
- show recent analyses
- entry point to new comparison

#### 2. New Comparison
- paste job description or URL
- upload CV or paste profile
- choose analysis language

#### 3. Match Report
- overall score
- visual overlap
- evidence-backed strengths
- missing or weak areas

#### 4. CV Adaptation
- prioritized edits
- example rewrites
- export-ready reviewed version

#### 5. Review
- final manual approval
- export or copy results

## Later Vision

### Profile Automation
- LinkedIn OAuth connection
- initial profile import
- periodic refresh of profile data
- user-controlled sync settings

### Job Discovery Automation
- saved searches per platform
- source connectors for LinkedIn, StepStone, and others
- scheduled polling
- match scoring against latest profile

### Email Ingestion
- scan inbox for job-related emails
- extract job links
- deduplicate and score them

### Notifications
- push, email, or in-app alerts for high-fit jobs
- alerts for recurring qualification gaps

## Prompt Architecture

### Generic Prompt
Use the generic fit-check prompt as the default system behavior:
- consistent scoring
- evidence discipline
- reusable across all users

### Personalized Prompt
Treat the personalized prompt as a user template layer:
- references user-specific profile context
- can include preferred tone and document strategy
- useful for premium or saved application profiles

## UX Principles

- web-first and mobile-friendly from day one
- no hidden automation in the MVP
- every score must be traceable to evidence
- recommendations must be actionable, not generic
- manual review remains mandatory before application output

## Technical Shape

### Frontend
- responsive web app
- PWA as a later extension
- mobile browser support for iPhone and Android

### Backend Later
- document ingestion
- job extraction pipeline
- auth connectors
- scheduled polling jobs
- notifications and user state

## Suggested Build Order

1. Responsive website with product framing and MVP flow
2. Single comparison workflow using pasted inputs
3. Structured analysis result UI with score and charts
4. CV adaptation step with manual approval
5. User accounts and saved analyses
6. LinkedIn sync and source polling
7. Email ingestion and notifications