All work

2026 · Python · CrewAI · LangGraph

Multi-Agent Campaign Creator

Four specialized agents turn one product brief into a full campaign — CrewAI execution, LangGraph control flow.

Source
Results
4 sequential
Agents
74% · 25 tests
Test coverage
research → copy → visuals → KPIs
Output
30-day
Plan horizon

Problem

Campaign creation is manual and serial — research, copy, art direction, strategy. This automates the whole chain from a single product description, with clean failure routing between stages.

Approach

  • 01Agent pipeline: Research → Copywriter → Art Director → Manager, each passing context forward.
  • 02CrewAI runs the agents; LangGraph optionally owns control flow with conditional success/failure edges.
  • 03Role-specific temperature per agent; multi-format Markdown + JSON output.
  • 04Free-tier Groq routing with auto-retry for rate limits.

Architecture

  1. Product brief → Research agent
  2. Copywriter agent
  3. Art Director agent
  4. Manager → synthesized campaign (MD / JSON)

Challenges

  • Clean failure routing via LangGraph conditional edges instead of a brittle linear chain.
  • Staying inside free-tier rate limits with auto-retry and backoff.

Key features

  • CrewAI + optional LangGraph orchestration
  • Conditional success/failure edge routing
  • Multi-format Markdown + JSON briefs
  • Interactive and demo modes
  • GitHub Actions CI
Back to all work