Documentation

File names such as ai.yaml or data/config/ai.yaml mean your own copy in data/config/; the shipped defaults in config/ are read-only (see configuration.md). Most settings can also be changed on the Settings page, or in its Advanced tab.

Start with docker compose up -d --build and the setup wizard at /setup (see setup.md). After the wizard, the tuning guide (open it in a browser) explains the pages and how to adjust what you get, with annotated mock-ups.

Guide	What it covers
installation.md	Install from scratch: requirements, every step on macOS and Linux, the AI worker as a service, Telegram, phone access, updating, backups, uninstalling, troubleshooting
faq.md	Short answers: cost, AI options, privacy, sources, markets, phone access, Windows
setup.md	First start: the setup wizard (/setup) and pnpm run setup, the answers file, setup.yaml, reopening the wizard
usage.md	Start here for daily use: opening the dashboard, health checks (doctor, Sources page, alerts, logs), the pages, reading a job description, querying jobs from a terminal
configuration.md	Where configuration lives: shipped defaults config/ and your own data/config/, which one is read, pnpm config:init / migrate / check, config_version upgrades
market-packs.md	The market you search in: choosing it, what follows it (places, remote scope, local language, base currency, channels), pack files, writing a pack, pnpm market:show / market:regress
languages.md	Interface, AI output, translation, Telegram and command-line languages (language.yaml), adding an interface language
settings.md	The Settings page: AI provider / model / effort dropdowns, manual buttons (collect now, test channels, test AI, Telegram test), rate limits
sources.md	The source list: sources.yaml, source ids, the seven states, page actions, import and export, pnpm sources:list
deploy.md	First-deploy checklist (macOS-oriented), scripts/ops/deploy.sh, pnpm run doctor, upgrade, rollback, a 7-day unattended acceptance table
upgrade.md	Upgrading to a release tag (deploy.sh --ref), what the deploy does, rolling back, uninstalling
host-setup.md	Docker at login, external volumes, power settings, Tailscale, CLI logins on the host, installing the ai-worker, BotFather
multiple-instances.md	One person = one instance: running a second one beside the first (project, volume and port names), what else to keep apart
backup-restore.md	Automatic pg_dump backups, retention, restore drills, disaster recovery
privacy.md	What is stored, what leaves your machine (AI providers, job sites, Telegram), provider data policies, how to delete everything
commercial-use.md	Using Job Radar in a company or team: AGPL, shared instances, scraping and AI-subscription terms
ai-providers.md	AI via a subscription CLI, an API key, or both: who runs what where, the host ai-worker, execution / host_worker / translation settings, troubleshooting
ai-usage.md	AI usage and cost statistics, model catalog (models.yaml)
extraction.md	Rules for work mode, German level, agency detection and description cleanup; pnpm reprocess
signals.md	Compensation and benefit signals: tiers, vocabularies, the two-step extraction, the board
salary.md	Salary level on the job page (benchmark of your collected postings, links to salary sites) and the optional AI salary lookup with web search, salary.yaml
breakdown.md	Job breakdown: manual or automatic, card states and error codes, quote checking, breakdown.yaml
ai-traps.md	AI trap detection: what is recognised, which sources only give text, how the text sent to an AI is cleaned, the marks on the pages and in Telegram, traps.yaml, pnpm traps:scan
insights.md	The /insights page: how every card is computed, insights.yaml
interest-suggestions.md	Weekly AI suggestions: evidence rules, approve / reject, suggest settings
resume-kit.md	Resume tailoring hand-off: resume.yaml, exporting a JD, the /tailor command
telegram.md	Creating a bot, finding your chat id
gmail-alerts.md	Two-step verification, app password, the label, subscribing to job alerts
linkedin-session.md	pnpm auth:linkedin, daily page limits, what to do on a verification page
jsearch.md, tavily.md, apify.md, brightdata.md	Keys, quotas and budgets of the paid or keyed channels
development.md	For contributors: code layout, tests, e2e, conventions
ci.md	The CI pipeline explained job by job and step by step: what each runs and why, timings, running it locally, reading failures, e2e projects, Dependabot, release check