CV Joint

Track job postings, create targeted CVs, render them with LaTeX, achieve constant velocity.

Features

• Create structured data from Job Posting URLs
• Create structured CV data from text CV files
• RAG knowledge base equipped chat
• Optimize CV data for job postings (in progress)
• Create PDFs from CV data with LaTeX (in progress)

Implementation

• Analysis and structured output with CrewAI

• Pydantic for defining domain objects and structured outputs

• Gradio tabbed UI

• CVs are rendered using Jinja2 with custom delimiters that play well with LaTeX:

  	customized	standard jinja2
  Statements	(# #)	{% %}
  Expressions	(( ))	{{ }}
  Comments	%( )%	{# #}
  Line Comment	%%	##

Simplified project structure:

.
├── src
│   ├── renderers
│   │   └── latex # LaTeX PDF rendering
│   ├── config
│   ├── converters
│   │   └── markdown.py
│   ├── crews
│   │   ├── cv_analysis
│   │   ├── cv_optimization
│   │   └── job_posting_analysis
│   ├── infrastructure
│   ├── models
│   ├── repositories
│   │   └── filesystem.py
│   ├── services
│   │   └── analyzers # crew facades
│   │   └── application.py
│   └── ui
│   │   └── app.py # Gradio
│   │   └── cli.py
└── templates
    ├── cover-letter.tex
    └── cv.tex

Installation

uv tool install --editable . # --editable: reflect code changes when run

If cv-joint is not found after installation:

uv tool update-shell

Set environment variables:

cp sample.env .env # edit .env and add API keys

Configure the rag-knowledge MCP server for RAG functions (see below).

Configuration override hierarchy:

1. src/*/config/settings.yaml (defaults)
2. ~/.cv-joint/settings.yaml (user dotfile)
3. src/*/config/settings.local.yaml (machine-specific overrides, gitignored)

Strings beginning with ~/ will undergo tilde expansion.

Example user settings, unnecessary defaults except where noted:

chat:
  model: "gpt-4o"

mcpServers:
  rag-knowledge: # null in default settings
    command: "~/.local/bin/uv"
    args:
      - "run"
      - "--directory"
      - "~/code/mcp-server-project"
      - "python"
      - "rag_knowledge_mcp.py"
    env:
      LOG_LEVEL: "INFO"
    x-tool-name: "rag_search_knowledge"

crews:
  cv_analysis:
    agents:
      cv_analyst:
        model: "gpt-4o-mini"

repositories:
  filesystem:
    data_dir: "./data"

Data directory structure:

data/
├── collections/
│   ├── job-postings.json
│   └── cvs.json
├── job-postings/{identifier}/
│   ├── job-posting.json
│   └── cvs/{identifier}/
│       ├── record.json
│       ├── transformation-plan.json
│       └── cv.json
└── cvs/{identifier}/
    └── cv.json

To dump all merged (tilde expanded) configuration:

cv-joint show-config

Testing

uv run pytest tests/ --tb=short # or: make test

Usage

cv-joint           # serve at http://localhost:7860
cv-joint launch    # same
cv-joint open      # serve and open in browser
cv-joint --help    # show additional functionality