CV Joint

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

Features

• Uses agentic analysis to
  • Create structured data from Job Posting URLs or text files
  • Create structured CV data from text CV files
  • Optimize CV data for a job posting with relevant experience using RAG
• Generates Markdown versions of domain objects, suitable for Obsidian
• Renders CV data in LaTeX to generate PDFs

Implementation

• Agentic analysis with CrewAI

• Pydantic for domain objects and structured outputs

• Click CLI

• 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
│   ├── crews
│   │   ├── cv_analysis
│   │   ├── cv_optimization
│   │   ├── job_posting_analysis
│   │   └── tools
│   ├── infrastructure
│   ├── models
│   ├── renderers
│   │   └── latex
│   ├── repositories
│   │   └── filesystem.py
│   ├── services
│   │   └── analyzers # crew facades
│   │   └── application.py
│   │   └── converters.py
│   │   └── exporter.py
│   │   └── knowledge_chat.py
│   └── ui
│   │   └── app.py # Gradio
│   │   └── cli.py
└── templates
    ├── cover-letter.tex
    └── cv.tex

Installation

make install

runs uv tool install --editable .

Configuration

cp sample.env .env

Set environment variables in .env:

ANTHROPIC_API_KEY=
DEEPSEEK_API_KEY=
GOOGLE_API_KEY=
OPENAI_API_KEY=
SERPER_API_KEY=

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 (~/.cv-joint/settings.yaml):

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
  job_posting_analysis:
    agents:
      job_analyst:
        model: gpt-4o
  cv_optimization:
    agents:
      cv_strategist:
        model: claude-sonnet-4-20250514
      cv_rewriter:
        model: claude-sonnet-4-20250514

repositories:
  filesystem:
    data_dir: "~/vaults/frobozz/areas/job-search/cv-joint"

Data directory structure:

{data_dir}/
├── collections/
│   ├── job-postings.json
│   ├── cvs.json
│   └── optimized-cvs.json
├── job-postings/{identifier}/
│   ├── job-posting.json
│   ├── job-posting.md  
│   └── cvs/{identifier}/ # optimized for job posting
│       ├── curriculum-vitae.json
│       ├── curriculum-vitae.md
│       ├── cv-transformation-plan.json
│       └── cv-transformation-plan.md
└── cvs/{identifier}/
    ├── curriculum-vitae.json
    └── curriculum-vitae.md

Testing

make test

runs uv run pytest tests/ --tb=short

Gradio Usage

cv-joint

Launches Gradio, runs GRADIO_LAUNCHED_COMMAND, GRADIO_FINISHED_COMMAND after Ctl-C

(I have it run a wrapper app pointing at http://localhost:7860, then close it)

cv-joint open

Launches Gradio then opens the default browser at http://localhost:7860

CLI Usage

# UI
cv-joint           # serve at http://localhost:7860
cv-joint open      # serve and open in browser

# Job postings
cv-joint list job-postings
cv-joint list job-postings --archived
cv-joint list job-postings -q acme
cv-joint archive job-postings/{id}
cv-joint apply job-postings/{id} {cv-id}
cv-joint apply job-postings/{id} {cv-id} --date 2026-05-27
cv-joint reanalyze job-postings/{id}

# CVs
cv-joint list cvs
cv-joint list cvs -q wesley
cv-joint reanalyze cvs/{id} path/to/file

# General
cv-joint remove job-postings/{id}
cv-joint remove job-postings/{id}/cvs/{id}
cv-joint rename job-postings/{id} {new-id}
cv-joint export-markdown
cv-joint export-markdown [job-postings|cvs]
cv-joint show-config