CrewAI Product Planner

A multi-agent product planning assistant built with CrewAI on EdgeOne Makers — a PM, Tech Lead, and Reviewer collaborate to turn your product idea into a PRD and Tech Spec through interactive Q&A.

Framework: CrewAI · Category: Quick Start · Language: Python

Overview

CrewAI Product Planner simulates a product team: a Product Manager interviews you to gather requirements, then collaborates with a Tech Lead to produce a PRD and Technical Specification. A Reviewer provides improvement suggestions at each stage. The entire process is conversational — you guide the direction through multiple-choice options or free-text input.

• Multi-agent orchestration — three agents (PM, Tech Lead, Reviewer) with distinct roles working in sequence via CrewAI Flows
• Interactive Q&A — the PM asks clarifying questions before drafting; you choose from options or type custom answers
• Iterative refinement — after the initial draft, continue providing feedback until you're satisfied with the final documents
• Session persistence — conversation state is recoverable across instances via external store sync
• Streaming output — real-time SSE streaming of agent responses with per-agent attribution

Environment Variables

Variable	Required	Description
AI_GATEWAY_API_KEY	Yes	Model gateway API key. Use your Makers Models API Key, or any OpenAI-compatible provider key.
AI_GATEWAY_BASE_URL	Yes	Gateway base URL. For Makers Models, use https://ai-gateway.edgeone.link/v1.

This template follows the OpenAI-compatible standard — you can point these variables at Makers Models or any other compatible gateway / provider.

How to get AI_GATEWAY_API_KEY

1. Open the Makers Console.
2. Sign in and enable Makers.
3. Go to Makers → Models → API Key and create a key.
4. Copy it into AI_GATEWAY_API_KEY (set AI_GATEWAY_BASE_URL to https://ai-gateway.edgeone.link/v1).

Built-in models (@makers/deepseek-v4-flash, @makers/hy3-preview, @makers/minimax-m2.7) are free and rate-limited — great for prototyping. For production, bind your own provider key (BYOK) in the console.

Local Development

Prerequisites: Node.js, npm, Python 3.11+

npm install
cp .env.example .env
edgeone makers dev

The CLI automatically installs Python dependencies from agents/requirements.txt.

Open http://localhost:8080/agent-metrics for the local observability panel.

Project Structure

crewai-planner-python/
├── agents/
│   ├── stream.py              # /stream — main conversation endpoint (SSE)
│   ├── _lib/
│   │   ├── flow.py            # TurnFlow: CrewAI Flow with pause/resume
│   │   ├── llm.py             # LLM singleton initialization
│   │   ├── persistence.py     # In-memory + store-backed state persistence
│   │   ├── feedback_provider.py # Async feedback provider (raises HumanFeedbackPending)
│   │   └── logger.py          # Shared logger factory
│   ├── _crews/
│   │   ├── agents.yaml        # Agent role definitions (PM, TL, Reviewer)
│   │   ├── discovery_crew/    # Requirements gathering crew
│   │   ├── planning_crew/     # PRD + Tech Spec generation crew
│   │   └── iteration_crew/    # Feedback iteration crew
│   └── requirements.txt       # Python dependencies
├── cloud-functions/
│   ├── history.py             # /history — retrieve conversation messages
│   └── delete.py              # /delete — delete conversation data
├── src/                       # Frontend (React + Tailwind)
├── edgeone.json               # Agent runtime configuration
└── package.json

Files prefixed with _ are private modules — not exposed as public routes by EdgeOne.

How It Works

The agent runs as a session-mode runtime: requests sharing the same conversation_id are routed to the same instance.

Workflow

1. First turn — user enters a product name. The PM agent asks a clarifying question and the Flow pauses (via @human_feedback).
2. Discovery phase — the PM asks 2-3 rounds of questions (A/B/C options). After enough context is gathered (or 3 rounds), the Flow transitions to drafting.
3. Drafting phase — the PM writes a PRD, the Tech Lead writes a Technical Specification, and the Reviewer suggests improvements. The Flow pauses for user feedback.
4. Iteration phase — user provides feedback (or selects "looks good"). The PM and TL respond to each piece of feedback. This loop continues until the user confirms finalization.
5. Finalization — PM and TL produce the final versions of both documents.

Key Mechanisms

• CrewAI Flow + @human_feedback: each pause raises HumanFeedbackPending; the next HTTP request resumes via resume_async(feedback).
• Streaming: CrewAI's FlowStreamingOutput delivers token-by-token SSE events with agent attribution.
• Persistence: in-memory dict stores Flow state; sync_pending_to_store() backs it up to the platform store so state is recoverable across instances.
• Session history: /history (cloud function) reads stored messages; /delete clears conversation data.

Routes

Route	Method	Description
/stream	POST	Main conversation turn (SSE streaming)
/history	POST	Retrieve conversation messages
/delete	POST	Delete conversation and flow state

The conversation_id is passed via the makers-conversation-id request header.

Resources

• Makers Agents Documentation
• Quick Start: Agent Development
• Makers Models

License

MIT