Guide
JDA quickstart
Follow these steps to prepare Chrome, connect providers, and launch the local Ollama workflow without leaving this landing page.
Install and pin the extension
Start with a clean install so the popup, overlay, and Options page share the same state before following the onboarding tabs below.
- Open the Chrome Web Store listing, click “Add to Chrome”, confirm permissions, and pin the icon to the toolbar.
- Open any vacancy (LinkedIn, hh.ru, Indeed, niche sites) and click the icon once to grant scopes for the popup/overlay.
- Visit Options → General to set language, report format, and clipboard behaviour (mirrors the Lang & I/O onboarding tab).
Wire models, prompts, and rules
Use the Options tabs highlighted in the onboarding block to keep models, prompts, and extraction rules in sync.
- LLM Settings: add Ollama, OpenAI, Gemini, OpenRouter or other endpoints, paste keys, and drag active models into the right order.
- Auto-extraction: duplicate LinkedIn / hh.ru templates, tune DOM chain, CSS, Template blocks, and test with Smart Highlighter.
- CV & Prompts: keep multiple resumes plus system/output templates so popup, overlay, Notion, and exports share one tone.
Run Ollama Local with Llama 3
Matches the Ollama onboarding tab: run the local provider with consent-friendly host/origin settings.
- Install Ollama and run `ollama pull llama3` to download the model.
- Open chrome://extensions, enable Developer Mode, and copy the JDA ID (chrome-extension://<ID>).
- Start `ollama serve` with `OLLAMA_HOST=127.0.0.1:11434` and `OLLAMA_ORIGINS=chrome-extension://<ID>`, then allow http://localhost:11434/* when Chrome prompts you.
Onboarding
Configure JDA in focused steps
Each Options tab has its own workflow. These quick notes explain what to prepare so your first analysis mirrors the production setup.
LLM Settings
Connect providers & models
Add Ollama, OpenAI, Gemini, OpenRouter, Azure OpenAI, Anthropic, Perplexity, DeepSeek, Meta/xAI or Hugging Face — and decide which models stay active.
- Store API keys or local endpoints and keep secrets disabled in exports if needed.
- Pick active models, order them with drag-and-drop, and define fallbacks.
- Adjust per-model prompts/tokens to keep {{GLOBAL_SYSTEM_PROMPT}} in sync.