A Developer Built Dossier for His Wife and Open-Sourced It

Dossier began as a job-search app that a developer built for his wife, then released as open source after she found a role. The self-hosted TypeScript application stores career history in PostgreSQL and uses a selected large language model to draft resumes and cover letters for individual postings. Its GitHub repository has 706 stars in a category dominated by paid, hosted services.

Model routing gives users control over where sensitive career data goes. Dossier supports local inference through Ollama or LM Studio, plus hosted services such as OpenAI-compatible APIs, Anthropic, and Mistral. Self-hosting covers the application and database; hosted models and optional job-board or Google integrations still transmit relevant data to those providers.

Grounded drafts from one profile

Each user maintains one profile containing contact details, skills, employment, projects, education, and publications. An existing PDF, Word, or Markdown resume can populate the initial record. For each application, Dossier parses the pasted job description, identifies its requirements, compares them with the profile, and generates a tailored resume and cover letter in the selected tone.

Generation is constrained by the stored profile. Dossier presents unmatched requirements as gaps and limits draft claims to recorded experience, skills, and achievements. Because model output remains probabilistic, applicants should verify every generated document before using it.

Drafts remain editable through a chat panel and conventional form controls. Users can regenerate sections, restore earlier AI edits, or revise individual cover-letter fields such as the greeting and sign-off. The server renders finished documents to PDF through Chromium, with 12 bundled themes and a theme builder.