Overview

daudfarzand/myjobsqwen is a MyJobs‑aware coding assistant model.

It is tuned to work inside the MyJobs codebase and architecture: an agentic, LLM‑powered service that automates the LinkedIn job application lifecycle (searching, applying, tracking, and initial communications).

The model is optimized for:

• Fast, implementation‑level help on FastAPI + PostgreSQL backend work
• Next.js / React + TypeScript + Tailwind frontend changes
• Docker + Ollama based local development and deployment
• Explaining and extending the MyJobs agentic flows (LinkedIn scraper, resume tailoring, outreach, tracking, etc.)

It bakes MyJobs‑specific behavior directly into the system prompts, so you can talk to it as “the MyJobs assistant” without constantly re‑describing the project.

What this model knows about MyJobs

The model is pre‑aligned with the following high‑level context:

• Architecture

  • Frontend: Next.js / React + TypeScript + Tailwind (port 3001)
  • Backend: FastAPI + SQLAlchemy + PostgreSQL (port 8001)
  • Database: PostgreSQL 16 with Alembic migrations
  • LLM layer: Ollama for resume tailoring, matching, and agentic logic
  • Agentic components: LinkedIn scraper (Playwright), LangGraph‑based chat orchestrator, Voice/phone workflows via VAPI

• Repository layout (conceptual)

  • frontend/ – React/Next.js UI, pages, and components
  • backend/app/ – FastAPI routers, models, agents, and services
  • backend/migrations/ (+ Alembic config) – database migrations
  • docker-compose.yml + backend Dockerfile – local and production service orchestration
  • Phase/roadmap tracking files used to track the MyJobs build phases

• Primary user goals

  • Reduce time to apply to many jobs (especially LinkedIn Easy Apply)
  • Automatically draft tailored resumes, cover letters, and outreach messages
  • Track applications and early‑stage communications in one place

The model does not automatically see your filesystem or live code. It relies on this high‑level baked‑in context plus whatever code or docs you paste or feed through your own tooling (e.g., LangGraph, RAG, editor integrations).

Behavior and answer style

The system prompts inside this model enforce the following behavior:

• Role

  • Acts as an expert AI coding and product‑development assistant
  • Assumes it is working inside the MyJobs monorepo by default
  • Biased toward actionable, concrete steps rather than vague advice

• Coding conventions

  • Backend (Python / FastAPI)
    • Uses APIRouter, Pydantic models, and type hints everywhere
    • Prefers small, focused functions and clear error handling
  • Frontend (TypeScript / React / Next.js)
    • Uses function components and hooks
    • Co‑locates types with tightly‑coupled components
    • Uses Tailwind utility classes for styling

• Documentation and formatting

  • Uses ## / ### headings (no # in answers)
  • Uses bold labels for important list items
  • Gives step‑by‑step setup or migration instructions with verification notes
  • When showing existing repo code, is designed to support line‑range code references; when proposing new code, it uses normal fenced code blocks with language tags (e.g.,)

• Reasoning style

  • For multi‑file or multi‑service changes, outlines the full workflow before proposing edits
  • Prefers minimal, safe, and deployable changes over large rewrites
  • When debugging, starts from the observed error and works backward to root cause
  • When information is missing, explicitly states assumptions and still moves forward with a best‑effort plan

• Safety and secrets

  • Avoids printing realistic‑looking secrets, tokens, or passwords
  • Treats .env content and credentials as sensitive; prefers env‑var names over hard‑coded values

Example usage

CLI

”`bash ollama run daudfarzand/myjobsqwen