ResumeCR7

Never stare at a blank resume again.

Stop trying to remember everything you have done. ResumeCR7 gives you a local place to keep concrete evidence from projects, jobs, education, and skills, then turns that evidence into a tailored, ATS-friendly resume when you need one.

It is not a chatbot wrapper. The core workflow is to maintain your own career record first, then generate outputs from that record. Job descriptions can guide selection and phrasing, but they are not treated as evidence.

Who It Is For

1. I made this tool because I'm too broke to afford any AI resume builder on the market
2. ResumeCR7 is for people who repeatedly need to explain their work clearly:

• Software engineers, students, and builders with project-heavy experience.
• Job seekers who tailor resumes for different roles.
• Freelancers or contractors who reuse the same work history in different formats.
• Anyone whose accomplishments are spread across GitHub, notes, tickets, portfolio pages, and memory.

Why Not Just Use ChatGPT?

ChatGPT can help rewrite a resume, but you still have to remember and paste the right facts every time. ResumeCR7 solves a different problem:

• Your evidence is structured and reusable instead of trapped in one prompt.
• Generated bullets are grounded in facts you authored or explicitly enriched.
• Skill and project selection can run deterministically with baseline fallbacks.
• Local YAML files remain inspectable, editable, and portable.
• The immediate output is a tailored .tex resume artifact, with optional PDF rendering, from the same maintained career record.

Key Features

• Local resume evidence for contact info, skills, projects, experience, and education.
• React/Vite workbench for editing evidence, staging changes, setting a target role, generating LaTeX, downloading PDFs, and enriching linked evidence.
• FastAPI backend with resume evidence CRUD and a /resume-generation facade.
• Grounded skill, project, job-focus, and bullet generation services.
• Deterministic baseline skill and project selection with optional OpenAI-backed methods.
• Local-first desktop shell using Tauri v2 and a packaged FastAPI sidecar.
• Runtime data bootstrap that creates schema-valid local files without overwriting existing user-authored evidence.

Screenshots

Maintain Project Evidence

Generate For A Target Role

Export A Resume

Quick Start

Requirements

For the released Linux AppImage, ResumeCR7 bundles the frontend, the packaged FastAPI backend sidecar, and Python application dependencies. PDF rendering is the only current feature that needs a host toolchain.

To quickly check/install the PDF toolchain on Ubuntu/Debian, you may optain the installation script from the release page or from packaging/linux/, then run:

bash /path/to/install-pdf-dependencies.sh

For local development:

• Python 3.12+
• uv for Python dependency management
• Node.js and npm for the frontend
• Optional: latexmk and a TeX installation for PDF rendering

Local Web Workbench

Install backend and frontend dependencies:

uv sync --extra dev
cd frontend
npm install

Run the backend:

uv run resumecr7-api --reload

In another terminal, run the workbench:

cd frontend
npm run dev

The Vite dev server proxies /api/* to http://127.0.0.1:8000 by default. Open the Vite URL shown by npm.

Desktop App

The packaged desktop build currently targets Linux AppImage. Windows and macOS desktop builds are planned, but not part of the current release.

For released Linux AppImage users, no Python, uv, Node.js, npm, or PyInstaller installation is required. On Ubuntu/Debian, install the PDF rendering toolchain only if you want to use Generate PDF:

bash resumecr7-install-pdf-dependencies.sh

You can verify the PDF toolchain without changing the system:

bash resumecr7-check-pdf-dependencies.sh

From a source checkout, use the matching scripts under packaging/linux/.

Build the Linux desktop app from the frontend directory:

cd frontend
npm install
npm run desktop:build

On Ubuntu, Tauri builds require WebView/GTK dependencies:

sudo apt-get update
sudo apt-get install -y \
  build-essential \
  pkg-config \
  libwebkit2gtk-4.1-dev \
  libgtk-3-dev \
  libayatana-appindicator3-dev \
  librsvg2-dev

The desktop shell starts a bundled backend sidecar on 127.0.0.1, waits for /health, and passes the backend URL to the frontend.

Runtime Data

ResumeCR7 stores local user data under RESUMECR7_DATA_DIR. In local development, the default is the repository user/ directory. In the packaged desktop app, Tauri provides the OS app-data directory for com.resumecr7.desktop.

Current Linux desktop location and planned Windows/macOS locations:

Linux:   $XDG_DATA_HOME/com.resumecr7.desktop or ~/.local/share/com.resumecr7.desktop
macOS:   ~/Library/Application Support/com.resumecr7.desktop
Windows: %APPDATA%\com.resumecr7.desktop

Runtime layout:

RESUMECR7_DATA_DIR/
  resume_evidence/
    user.yaml
    skills.yaml
    projects.yaml
    education.yaml
    experience.yaml
  resume_generation/
    config.yaml
    job_target.yaml
    artifacts/
      resume_result.json
      resume_run_manifest.json
      resume.tex
      resume.pdf
    output/
      resume.tex
      resume.pdf
  logs/
    resumecr7.log
    desktop-sidecar.log

The user/ directory is intentionally ignored by git. Safe fictional examples live under examples/.

Resume generation writes internal artifacts under resume_generation/artifacts/ on every run; newer artifacts replace older artifacts and are useful for local inspection and monitoring. The Config page also stores a user-facing resume output directory in resume_generation/config.yaml. After .tex or PDF generation, ResumeCR7 copies resume.tex and resume.pdf to that configured output directory. PDF rendering always uses the internal artifact .tex as its source, then copies the finished PDF to the user output directory.

Evidence Model

Resume evidence is the source of truth. Implemented evidence files are:

• user.yaml: contact/header fields.
• skills.yaml: categorized skills under technology, programming, and concepts.
• projects.yaml: project records with summaries, highlights, active flags, categorized skills, and optional links.
• experience.yaml: work records with role, dates, location, highlights, categorized skills, and optional links.
• education.yaml: education records with degree, dates, location, grade, and relevant coursework.

See examples/resume_evidence/ for safe starter files.

Why FastAPI For Local YAML?

ResumeCR7 is local-first today, and the source of truth is a small set of YAML files. That can look heavier than necessary for an editor that also sends optional LLM requests.

The FastAPI backend is intentional: it gives the app one boundary for schema validation, atomic file writes, generation orchestration, metrics, and future web-service expansion. The React workbench, desktop shell, and any future hosted surface can share the same product behavior instead of reimplementing resume evidence rules in each client.

API Surface

Start the backend:

uv run resumecr7-api --reload

OpenAPI docs are available at http://127.0.0.1:8000/docs.

Primary product routes:

Method	Path	Purpose
GET	/health	Service liveness, effective config, and runtime paths
GET	/metrics-lite	Aggregate request/error/latency/token counters
GET	/resume-evidence	Load all registered resume evidence
GET/PUT	/resume-evidence/user	Read or replace contact evidence
GET/PUT	/resume-evidence/skills	Read or replace skills evidence
GET/POST	/resume-evidence/projects	List or create projects
GET/PUT/DELETE	/resume-evidence/projects/{id}	Read, replace, or delete a project
GET/POST	/resume-evidence/experience	List or create experience records
GET/PUT/DELETE	/resume-evidence/experience/{id}	Read, replace, or delete experience
GET/POST	/resume-evidence/education	List or create education records
GET/PUT/DELETE	/resume-evidence/education/{id}	Read, replace, or delete education
POST	/resume-generation/tex	Run the full resume pipeline, write the .tex artifact, copy the user .tex, and return paths/content
POST	/resume-generation/pdf	Render the current .tex artifact, copy the user PDF, and return PDF bytes
POST	/resume-generation/enrich-link-evidence	Enrich project or experience evidence from links

Lower-level capability routes are also available for testing and integration: /select-skills, /select-projects, /derive-job-focus, /generate-bulletpoints, and /enrich-link-evidence.

Configuration

ResumeCR7 reads environment settings through app/config.py and per-run resume generation settings through resume_generation/config.yaml in the active data directory.

Common environment variables:

RESUMECR7_DATA_DIR=user
RESUMECR7_PACKAGED=false
SKILL_METHOD=baseline        # baseline, embeddings, llm
SKILL_TOP_N=10
PROJ_METHOD=llm              # baseline, llm
DEV_MODE=true
LOG_LEVEL=INFO
LLM_PROVIDER=openai          # openai, qwen
OPENAI_API_KEY=your_key_here
QWEN_API_KEY=your_qwen_key_here
# DASHSCOPE_API_KEY can also provide the Qwen key.
QWEN_BASE_URL=https://dashscope-us.aliyuncs.com/compatible-mode/v1
RESUMECR7_GITHUB_TOKEN=github_pat_for_private_repo_scanning

OPENAI_API_KEY is required for OpenAI-backed embeddings and OpenAI-backed LLM features. QWEN_API_KEY or DASHSCOPE_API_KEY is required when LLM_PROVIDER=qwen for LLM-backed selection, job-focus generation, bullet generation, and enabled link scanning. Baseline paths remain functional without LLM provider keys.

Private GitHub repository link scanning can use RESUMECR7_GITHUB_TOKEN or GITHUB_TOKEN, or a saved token in the local generation config. Use a fine-grained GitHub token scoped to only the repositories you want ResumeCR7 to read, with repository contents read access.

HTTP clients used by LLM-backed features also honor standard proxy environment variables, including HTTP_PROXY, HTTPS_PROXY, ALL_PROXY, and NO_PROXY and their lowercase variants. For HTTP/HTTPS proxies, set proxy URLs with an http:// or https:// scheme. If ALL_PROXY or another proxy variable uses a SOCKS scheme such as socks://, socks5://, or socks5h://, the Python environment must include SOCKS support through the declared httpx[socks] dependency.

Repository Structure

app/                      FastAPI backend and product services
frontend/                 React/Vite workbench and Tauri desktop shell
resume_evidence/          compatibility shims and CLI entrypoints
resume_generation/        compatibility shims and CLI entrypoints
data/                     durable evaluation and skill-pool assets
examples/                 fictional evidence/config examples
tests/                    backend test suite
scripts/                  build, eval, and release helper scripts
docs/
  architecture-overview.md
  development.md
  screenshots/

Start architecture work with docs/architecture-overview.md and docs/development.md.

Contributing

1. Create a focused branch.
2. Install dependencies with uv sync --extra dev and cd frontend && npm install.
3. Add or update tests for non-trivial behavior changes.
4. Run backend tests with uv run pytest.
5. Run frontend checks with cd frontend && npm test && npm run build.
6. For release metadata changes, run uv run python scripts/validate_release.py --tag vX.Y.Z.
7. Open a pull request that explains the user-visible behavior and test coverage.

Roadmap

Near-term work:

• Support for Windows and macOS desktop builds.
• Signed installers before broad desktop distribution.
• Automatic updater integration.
• Async generation-run lifecycle after the local facade shape stabilizes.
• Easier manipulation, editing of generated resume, maybe a real-time editor.
• Quicker generation with multithreaded or multiprocess LLM calls.
• Wider LLM support (currently only OpenAI), including open-source models and local inference.
• More resume templates and auto-formatting. For example, dynamic font sizing, line spacing, and page breaks based on content length. Or an estimation/warning if the current content will not fit on a single page.