Joblyser

Joblyser is an AI-powered job application platform for resume-assisted analysis and content generation (reviews, cover letters, cold emails, and LinkedIn notes).

This repository is a monorepo containing all runtime services and shared migration assets.

Monorepo Layout

Path	Service	Purpose
joblyser-web	Next.js	User-facing web application
joblyser-api	Go	Auth, user/resume APIs, and dispatch proxy
joblyser-dispatch	FastAPI (Python)	Session/task API and Celery task producer
joblyser-worker	Celery (Python)	Asynchronous task consumer
joblyser-ai	FastAPI (Python)	Agent execution, job-description extraction, and RAG
migrations	SQL	Shared Postgres schema migrations
docs	Markdown	Workspace documentation

System Overview

1. The web app uses joblyser-api for auth, user, resume, and API-proxied agent requests, and calls joblyser-dispatch for streamed agent runs and job-description extraction.
2. joblyser-dispatch creates and tracks tasks, then publishes worker.tasks.agent to RabbitMQ's agent-queue.
3. joblyser-worker consumes the task and invokes joblyser-ai with a worker service token.
4. joblyser-ai executes the agent workflow; the worker persists task/chat updates for dispatch to stream back to the web app.

Platform Capabilities

• JD analysis: Skill matching, strengths/gaps, and fit summary.
• Generated assets: Cover letter, cold email, and LinkedIn note generation.
• Session history: Multi-run chat history per server-generated session.
• Resume workflow:
  • upload resume documents
  • select profile primary resume
  • use profile or ad-hoc resume in session
• Editable outputs:
  • cover_letter, linkedin_note, cold_mail support editing
  • edits are persisted via orchestrated API routes with debounced autosave.

Prerequisites

• Bun 1.2.18
• Go 1.25+
• Python 3.13+
• PostgreSQL
• MongoDB
• RabbitMQ
• Chroma and Redis-compatible storage for AI/dispatch/worker flows

Docker Compose Setup

1. Copy .env.example to .env at the repository root.
2. Copy each env/.env.{ai,api,dispatch,worker}.example to its non-example counterpart and set the required values.
3. Resolve the current Compose service-name mismatch (worker.depends_on.agent should refer to ai) before running docker compose up --build.

By default, host port bindings in docker-compose.yml are restricted to 127.0.0.1 for safer local usage. Only set a *_BIND_HOST variable to 0.0.0.0 when you intentionally want external access.

Local Runbook

1. Start infrastructure dependencies:
   • PostgreSQL
   • MongoDB
   • RabbitMQ
2. Apply SQL migrations from migrations/.
3. Start services in this order:
   • joblyser-ai
   • joblyser-worker
   • joblyser-dispatch
   • joblyser-api
   • joblyser-web

For service-specific commands and environment variables, see each service README.

Current Product Behaviors

• Session IDs are server-generated.
• Agent runs are streamed from dispatch as server-sent events.
• Task and chat state are stored outside Postgres by dispatch and worker.

Repository Conventions

• Treat services as independent deployable units.
• The web app uses /api/v1 on both the Go API and dispatch; AI routes are mounted under /api.
• Prefer server-owned identifiers for persistent entities (sessions/chats/jobs).
• Avoid committing secrets; use local env files and templates.

Documentation

• Workspace docs index: docs/README.md
• Developer workflow guide: docs/development-workflow.md