Collaboration Models Development Guide

This document provides detailed configuration examples, real-world cases, and technical implementation details for collaboration models.

Basic Concepts: Please read Collaboration Models Concepts first to understand the basic concepts.

---

Pipeline Configuration Details

Complete YAML Configuration Example

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: dev-pipeline-team
  namespace: default
spec:
  # Collaboration model: pipeline
  collaborationModel: "pipeline"

  # Team members (defined in execution order)
  members:
    # Step 1: Developer
    - name: "developer"
      role: "leader"
      botRef:
        name: developer-bot
        namespace: default
      prompt: |
        You are a senior software developer.
        Your task is to implement the feature based on the requirements.
        Write clean, well-documented code following best practices.
      # Pass the developer output to the next stage
      contextPassing: "previous_bot"

    # Step 2: Code Reviewer
    - name: "reviewer"
      role: "member"
      botRef:
        name: reviewer-bot
        namespace: default
      prompt: |
        You are a code reviewer.
        Review the code for:
        - Code quality and style
        - Potential bugs and security issues
        - Performance optimization opportunities
        Provide constructive feedback.
      # Pass both the original request and reviewer output
      contextPassing: "original_and_previous"

    # Step 3: Test Engineer
    - name: "tester"
      role: "member"
      botRef:
        name: tester-bot
        namespace: default
      prompt: |
        You are a QA engineer.
        Create comprehensive tests including:
        - Unit tests
        - Integration tests
        - Edge case scenarios
        Execute tests and report results.

    # Step 4: Deployment Expert
    - name: "deployer"
      role: "member"
      botRef:
        name: deployer-bot
        namespace: default
      prompt: |
        You are a DevOps engineer.
        Prepare the deployment:
        - Build the application
        - Create deployment configuration
        - Document deployment steps

Real-World Case: Blog Article Production Pipeline

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: blog-production-pipeline
  namespace: default
spec:
  collaborationModel: "pipeline"
  members:
    - name: "writer"
      role: "leader"
      botRef:
        name: content-writer-bot
        namespace: default
      prompt: "Write an engaging blog post on the given topic with proper structure and flow."

    - name: "editor"
      role: "member"
      botRef:
        name: content-editor-bot
        namespace: default
      prompt: "Edit the content for grammar, clarity, and readability. Improve sentence structure and flow."

    - name: "seo-optimizer"
      role: "member"
      botRef:
        name: seo-bot
        namespace: default
      prompt: "Optimize the content for SEO: add meta descriptions, keywords, and improve headings."

    - name: "publisher"
      role: "member"
      botRef:
        name: publisher-bot
        namespace: default
      prompt: "Format the content for publication and create a publishing checklist."

Context Passing Between Stages

Pipeline members can set contextPassing to control which message this stage passes to the next stage. The default value is none, which preserves the existing no-context behavior.

Value	Description
none	Pass no message
original_user	Pass the original user request
previous_bot	Pass the current stage Bot's final output
original_and_previous	Pass both the original user request and current stage output

Pipeline Best Practices

✅ Recommended

• Each stage has single clear responsibility
• Control pipeline length (max 6 steps recommended)
• Clear input/output expectations in each Bot's prompt
• Add error handling and validation steps

❌ Avoid

• Overly long pipelines (more than 8 steps)
• Overlapping responsibilities between steps
• Missing intermediate validation
• Illogical Bot ordering

Optimized Pipeline Design Example

members:
  - name: "validator" # First validate input
  - name: "processor" # Then process
  - name: "quality-check" # Quality check
  - name: "finalizer" # Finally complete

---

Route Configuration Details

Complete YAML Configuration Example

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: tech-support-route-team
  namespace: default
spec:
  # Collaboration model: route
  collaborationModel: "route"

  # Team members
  members:
    # Leader: Routing Decision Maker
    - name: "router"
      role: "leader"
      botRef:
        name: router-bot
        namespace: default
      prompt: |
        You are a technical support router.
        Analyze the user's question and route it to the appropriate specialist:
        - Frontend issues → frontend-expert
        - Backend issues → backend-expert
        - Database issues → database-expert
        - DevOps/Infrastructure issues → devops-expert

        Provide a brief analysis of why you chose this specialist.

    # Expert 1: Frontend Specialist
    - name: "frontend-expert"
      role: "member"
      botRef:
        name: frontend-specialist-bot
        namespace: default
      prompt: |
        You are a frontend development expert specializing in:
        - React, Vue, Angular
        - HTML, CSS, JavaScript/TypeScript
        - UI/UX best practices
        - Browser compatibility

        Provide detailed, actionable solutions to frontend problems.

    # Expert 2: Backend Specialist
    - name: "backend-expert"
      role: "member"
      botRef:
        name: backend-specialist-bot
        namespace: default
      prompt: |
        You are a backend development expert specializing in:
        - Python, Java, Node.js
        - RESTful API design
        - Microservices architecture
        - Performance optimization

        Provide detailed, actionable solutions to backend problems.

    # Expert 3: Database Specialist
    - name: "database-expert"
      role: "member"
      botRef:
        name: database-specialist-bot
        namespace: default
      prompt: |
        You are a database expert specializing in:
        - SQL and NoSQL databases
        - Query optimization
        - Database design and normalization
        - Indexing strategies

        Provide detailed, actionable solutions to database problems.

    # Expert 4: DevOps Specialist
    - name: "devops-expert"
      role: "member"
      botRef:
        name: devops-specialist-bot
        namespace: default
      prompt: |
        You are a DevOps expert specializing in:
        - Docker and Kubernetes
        - CI/CD pipelines
        - Cloud infrastructure (AWS, GCP, Azure)
        - Monitoring and logging

        Provide detailed, actionable solutions to DevOps problems.

Real-World Case: Multi-Language Programming Q&A Platform

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: programming-qa-route-team
  namespace: default
spec:
  collaborationModel: "route"
  members:
    - name: "language-router"
      role: "leader"
      botRef:
        name: language-router-bot
        namespace: default
      prompt: |
        Analyze the programming question and route to the appropriate language expert:
        - Python questions → python-expert
        - JavaScript/TypeScript → js-expert
        - Java/Kotlin → jvm-expert
        - Go → go-expert

    - name: "python-expert"
      role: "member"
      botRef:
        name: python-bot
        namespace: default
      prompt: "You are a Python expert. Answer Python-related questions with code examples and best practices."

    - name: "js-expert"
      role: "member"
      botRef:
        name: javascript-bot
        namespace: default
      prompt: "You are a JavaScript/TypeScript expert. Provide modern ES6+ solutions and TypeScript types."

    - name: "jvm-expert"
      role: "member"
      botRef:
        name: java-bot
        namespace: default
      prompt: "You are a JVM expert. Answer Java and Kotlin questions with attention to performance."

    - name: "go-expert"
      role: "member"
      botRef:
        name: go-bot
        namespace: default
      prompt: "You are a Go expert. Provide idiomatic Go solutions emphasizing concurrency."

Route Best Practices

✅ Recommended

• Clear and accurate routing logic in Leader Bot
• Well-defined expert domains without overlap
• Detailed routing rules for Leader
• Include default routing for unknown cases

❌ Avoid

• Ambiguous routing rules
• Expert domain overlap causing selection difficulty
• Missing default handling path
• Overly complex routing decisions

Clear Routing Rules Example

- name: "router"
  prompt: |
    Route questions based on clear criteria:
    - If about UI/UX/styling → frontend-expert
    - If about API/database/server → backend-expert
    - If about deployment/infrastructure → devops-expert
    - If unclear or mixed → general-expert (default)

---

Coordinate Configuration Details

Both Claude Code and Codex executors use the first bot as the coordinator and the remaining members as native subagents. Codex registers a role for each member with its own instructions, model, and reasoning settings. Members without a model inherit the coordinator's model. For API models, the executor proxy routes each member to its own provider and credentials; credentials are never written to role files. Team members share Skills and MCP servers in the execution environment. Codex configuration or role-file deployment failures fail the task explicitly.

With native Codex login, members can inherit that login and select another model. Members using an independent API provider require an API model for the coordinator as well.

Complete YAML Configuration Example

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: market-research-coordinate-team
  namespace: default
spec:
  # Collaboration model: coordinate
  collaborationModel: "coordinate"

  # Team members
  members:
    # Leader: Coordinator
    - name: "coordinator"
      role: "leader"
      botRef:
        name: coordinator-bot
        namespace: default
      prompt: |
        You are a market research coordinator.

        PHASE 1 - Task Decomposition:
        Break down the market research task into parallel workstreams:
        1. Competitor analysis
        2. Customer sentiment analysis
        3. Market trend analysis
        4. Data collection and statistics

        Assign each workstream to the appropriate specialist.

        PHASE 2 - Result Synthesis:
        After receiving all reports, synthesize them into a comprehensive
        market research report with:
        - Executive summary
        - Key findings from each area
        - Strategic recommendations
        - Data visualizations and insights

    # Expert 1: Competitor Analyst
    - name: "competitor-analyst"
      role: "member"
      botRef:
        name: competitor-analyst-bot
        namespace: default
      prompt: |
        You are a competitor analysis specialist.
        Analyze:
        - Main competitors and market share
        - Competitor strategies and positioning
        - Strengths and weaknesses
        - Competitive advantages

        Provide a detailed competitor analysis report.

    # Expert 2: Customer Insights Analyst
    - name: "customer-analyst"
      role: "member"
      botRef:
        name: customer-analyst-bot
        namespace: default
      prompt: |
        You are a customer insights specialist.
        Analyze:
        - Customer demographics and segments
        - Customer pain points and needs
        - Customer satisfaction and feedback
        - Buying behavior patterns

        Provide a detailed customer analysis report.

    # Expert 3: Trend Analyst
    - name: "trend-analyst"
      role: "member"
      botRef:
        name: trend-analyst-bot
        namespace: default
      prompt: |
        You are a market trend specialist.
        Analyze:
        - Industry trends and future outlook
        - Emerging technologies and innovations
        - Regulatory and policy changes
        - Market opportunities and threats

        Provide a detailed trend analysis report.

    # Expert 4: Data Analyst
    - name: "data-analyst"
      role: "member"
      botRef:
        name: data-analyst-bot
        namespace: default
      prompt: |
        You are a data analytics specialist.
        Analyze:
        - Market size and growth rates
        - Statistical trends and patterns
        - Revenue forecasts
        - Key performance indicators

        Provide a detailed data analysis report with visualizations.

Real-World Case: Comprehensive Code Review Team

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: comprehensive-code-review-team
  namespace: default
spec:
  collaborationModel: "coordinate"
  members:
    - name: "review-coordinator"
      role: "leader"
      botRef:
        name: review-coordinator-bot
        namespace: default
      prompt: |
        Coordinate a comprehensive code review:
        1. Distribute code to specialized reviewers
        2. Collect all review feedback
        3. Synthesize into a final review report with prioritized action items

    - name: "security-reviewer"
      role: "member"
      botRef:
        name: security-bot
        namespace: default
      prompt: "Review code for security vulnerabilities, injection risks, and authentication issues."

    - name: "performance-reviewer"
      role: "member"
      botRef:
        name: performance-bot
        namespace: default
      prompt: "Review code for performance issues, optimization opportunities, and scalability."

    - name: "quality-reviewer"
      role: "member"
      botRef:
        name: quality-bot
        namespace: default
      prompt: "Review code quality, maintainability, design patterns, and best practices."

    - name: "test-reviewer"
      role: "member"
      botRef:
        name: test-bot
        namespace: default
      prompt: "Review test coverage, test quality, and identify missing test scenarios."

Coordinate Best Practices

✅ Recommended

• Leader has clear task decomposition strategy
• Expert Bot responsibilities don't overlap
• Leader needs strong aggregation capability
• Control parallel Bot count (3-5 optimal)

❌ Avoid

• Unbalanced task decomposition
• Too many parallel Bots (more than 7)
• Leader lacks aggregation guidance
• Inconsistent expert output formats

Structured Coordination Example

- name: "coordinator"
  prompt: |
    STEP 1: Decompose task into 4 parallel workstreams
    STEP 2: Assign to specialists
    STEP 3: Collect all reports in structured format
    STEP 4: Synthesize into unified report with:
      - Executive summary
      - Key findings per specialist
      - Recommendations

---

Collaborate Configuration Details

Complete YAML Configuration Example

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: product-brainstorm-team
  namespace: default
spec:
  # Collaboration model: collaborate
  collaborationModel: "collaborate"

  # Team members (all participate equally)
  members:
    # Product Manager Perspective
    - name: "product-manager"
      role: "member"
      botRef:
        name: pm-bot
        namespace: default
      prompt: |
        You are a product manager participating in a brainstorming session.
        Focus on:
        - User needs and market fit
        - Feature prioritization
        - Business value and ROI
        - User experience

        Engage actively with other team members' ideas and build upon them.

    # Technical Architect Perspective
    - name: "architect"
      role: "member"
      botRef:
        name: architect-bot
        namespace: default
      prompt: |
        You are a software architect participating in a brainstorming session.
        Focus on:
        - Technical feasibility
        - System design and scalability
        - Integration challenges
        - Technical debt considerations

        Provide technical insights and collaborate with the team.

    # UX Designer Perspective
    - name: "ux-designer"
      role: "member"
      botRef:
        name: ux-bot
        namespace: default
      prompt: |
        You are a UX designer participating in a brainstorming session.
        Focus on:
        - User interface and interaction design
        - User journey and experience
        - Accessibility and usability
        - Visual design principles

        Contribute design perspectives and iterate on ideas.

    # Data Analyst Perspective
    - name: "data-analyst"
      role: "member"
      botRef:
        name: data-bot
        namespace: default
      prompt: |
        You are a data analyst participating in a brainstorming session.
        Focus on:
        - Data-driven insights
        - Metrics and KPIs
        - User behavior analysis
        - A/B testing opportunities

        Provide analytical perspective and support decisions with data.

    # Marketing Expert Perspective
    - name: "marketing-expert"
      role: "member"
      botRef:
        name: marketing-bot
        namespace: default
      prompt: |
        You are a marketing expert participating in a brainstorming session.
        Focus on:
        - Market positioning
        - Target audience
        - Go-to-market strategy
        - Competitive differentiation

        Contribute marketing insights and collaborate on strategy.

Real-World Case: System Incident Response Team

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: incident-response-team
  namespace: default
spec:
  collaborationModel: "collaborate"
  members:
    - name: "backend-engineer"
      role: "member"
      botRef:
        name: backend-sre-bot
        namespace: default
      prompt: |
        You are a backend engineer responding to a system incident.
        Check: API services, application logs, error rates, service dependencies.
        Share findings and collaborate to identify root cause.

    - name: "database-admin"
      role: "member"
      botRef:
        name: dba-bot
        namespace: default
      prompt: |
        You are a database administrator responding to a system incident.
        Check: Database performance, query performance, connections, locks, replication status.
        Share findings and collaborate to identify root cause.

    - name: "frontend-engineer"
      role: "member"
      botRef:
        name: frontend-sre-bot
        namespace: default
      prompt: |
        You are a frontend engineer responding to a system incident.
        Check: Client-side errors, network requests, browser console, CDN status.
        Share findings and collaborate to identify root cause.

    - name: "devops-engineer"
      role: "member"
      botRef:
        name: devops-bot
        namespace: default
      prompt: |
        You are a DevOps engineer responding to a system incident.
        Check: Infrastructure health, container status, network connectivity, resource usage.
        Share findings and collaborate to identify root cause.

Real-World Case: Innovation Product Design Workshop

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: innovation-workshop-team
  namespace: default
spec:
  collaborationModel: "collaborate"
  members:
    - name: "creative-thinker"
      role: "member"
      botRef:
        name: creative-bot
        namespace: default
      prompt: "Generate innovative and out-of-the-box ideas. Challenge assumptions and explore possibilities."

    - name: "practical-analyst"
      role: "member"
      botRef:
        name: analyst-bot
        namespace: default
      prompt: "Evaluate ideas for feasibility, cost, and practicality. Provide realistic assessments."

    - name: "customer-advocate"
      role: "member"
      botRef:
        name: customer-bot
        namespace: default
      prompt: "Represent the customer perspective. Ensure ideas truly solve user problems."

    - name: "tech-innovator"
      role: "member"
      botRef:
        name: tech-bot
        namespace: default
      prompt: "Explore cutting-edge technologies that could enable innovative solutions."

Collaborate Best Practices

✅ Recommended

• Define clear collaboration goals
• Each Bot has clear perspective/role
• Set discussion rounds or stopping conditions
• Encourage Bots to reference and respond to each other

❌ Avoid

• Lack of clear goals causing divergence
• Vague role definitions
• No discussion termination conditions
• Bots talking past each other without interaction

Effective Collaboration Setup Example

members:
  - name: "architect"
    prompt: |
      Role: Technical Architect
      Goal: Design scalable solution
      Interaction: Build on others' ideas, raise technical concerns

  - name: "product"
    prompt: |
      Role: Product Manager
      Goal: Ensure user value
      Interaction: Connect features to user needs, prioritize

  # Include facilitation instructions
  - name: "facilitator"
    role: "leader"
    prompt: |
      Guide discussion towards concrete action items.
      Summarize when reaching consensus.
      Keep discussion focused on the goal.

---

Performance Comparison

Model	Average Time	Parallelism	Resource Consumption	Predictability
Pipeline	N × T	Low (serial)	Low	High
Route	T + routing time	Low (single path)	Low	High
Coordinate	T + aggregation time	High (parallel)	High	Medium
Collaborate	Uncertain	High (concurrent)	High	Low

Note: N = Number of Bots, T = Average processing time per Bot

---

General Prompt Design Guide

Good Prompt Structure

prompt: |
  # Good Prompt Structure:
  1. Role definition: You are a [role]...
  2. Responsibilities: Your responsibilities include...
  3. Input description: You will receive...
  4. Output requirements: Provide output in the format...
  5. Quality criteria: Ensure [quality criteria]...

Team Size Recommendations

• Pipeline: 3-6 Bots
• Route: 1 Leader + 3-8 experts
• Coordinate: 1 Leader + 3-5 experts
• Collaborate: 3-8 Bots

---

More Real-World Cases

News Insights Platform (Coordinate)

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: news-insight-team
  namespace: default
spec:
  collaborationModel: "coordinate"
  members:
    - name: "coordinator"
      role: "leader"
      botRef: { name: coordinator-bot, namespace: default }
      prompt: "Coordinate news analysis: assign data collection, sentiment analysis, and trend identification. Synthesize comprehensive news report."

    - name: "news-collector"
      role: "member"
      botRef: { name: collector-bot, namespace: default }
      prompt: "Collect news from multiple sources on the given topic. Provide summaries with sources and timestamps."

    - name: "sentiment-analyzer"
      role: "member"
      botRef: { name: sentiment-bot, namespace: default }
      prompt: "Analyze sentiment and tone of news articles. Identify positive, negative, and neutral coverage."

    - name: "trend-identifier"
      role: "member"
      botRef: { name: trend-bot, namespace: default }
      prompt: "Identify emerging trends and patterns in news coverage. Highlight key themes and developments."

Full-Stack Development Team (Pipeline)

apiVersion: agent.wecode.io/v1
kind: Team
metadata:
  name: fullstack-dev-team
  namespace: default
spec:
  collaborationModel: "pipeline"
  members:
    - name: "requirements-analyst"
      role: "leader"
      botRef: { name: analyst-bot, namespace: default }
      prompt: "Analyze requirements and create detailed technical specifications."

    - name: "backend-developer"
      role: "member"
      botRef: { name: backend-bot, namespace: default }
      prompt: "Implement backend API based on specifications. Use Python FastAPI."

    - name: "frontend-developer"
      role: "member"
      botRef: { name: frontend-bot, namespace: default }
      prompt: "Build frontend interface using React. Integrate with backend API."

    - name: "integration-tester"
      role: "member"
      botRef: { name: tester-bot, namespace: default }
      prompt: "Test full-stack integration. Report any issues or bugs."

---

Related Documentation

• Collaboration Models Concepts - Basic concepts and selection guide
• Core Concepts - Understand Bot, Team, and other core concepts
• YAML Specification - Complete configuration format guide