Skip to content

EXAMPLES.md

Latest commit

 

History

History
269 lines (199 loc) · 8.67 KB

EXAMPLES.md

File metadata and controls

269 lines (199 loc) · 8.67 KB

AgentHub - Use Cases & Examples

Real-world examples of how AgentHub orchestrates multi-AI collaboration.

Table of Contents

Example 1: Parallel Code Review

Scenario: MAIN agent wrote a new feature and wants thorough review before merging.

Workflow

# Terminal 1 (MAIN) - Initiates review
python3 src/cli/team.py say --from MAIN --to A --type review --text "Review auth.py:120-180"
python3 src/cli/team.py say --from MAIN --to B --type review --text "Review auth.py:120-180"
python3 src/cli/team.py say --from MAIN --to C --type review --text "Review auth.py:120-180"

# A, B, C review in parallel...

# Terminal 1 (MAIN) - Collects results
python3 src/cli/team.py inbox

Message Flow

MAIN                    A                       B                       C
 │                      │                       │                       │
├───────review─────────>│                       │                       │
├───────review─────────────────────────────────>│                       │
├───────review─────────────────────────────────────────────────────────>│
 │                      │                       │                       │
 │                      │<───────report─────────│                       │
 │                      │                                               │
 │<─────report──────────┘                       │                       │
 │                                              │                       │
 │<───────────────────────report────────────────┘                       │
 │                                                                      │
 │<──────────────────────────────────────────report────────────────────┘
 │
 └──> Consolidates feedback → Requests fixes → Final verification

Time Savings

Approach Time Speedup
Sequential review (A→B→C) 30 min 1x
Parallel review (A‖B‖C) 10 min 3x

Example 2: Feature Development

Scenario: Implement user authentication with OAuth2.

Step 1: MAIN Breaks Down Task

# MAIN assigns subtasks
python3 src/cli/team.py say --from MAIN --to A --type assign --text "Task: Design OAuth2 flow diagram"
python3 src/cli/team.py say --from MAIN --to B --type assign --text "Task: Implement backend OAuth endpoints"
python3 src/cli/team.py say --from MAIN --to C --type assign --text "Task: Create frontend login page"
python3 src/cli/team.py say --from MAIN --to D --type assign --text "Task: Write unit tests"

Step 2: Parallel Execution

Agent Task Output
A Architecture docs/oauth-flow.md
B Backend src/auth/oauth.py
C Frontend src/pages/login.tsx
D Tests tests/test_oauth.py

Step 3: Integration

# MAIN verifies each component
python3 src/cli/team.py say --from MAIN --to A --type verify --text "Verify flow diagram matches implementation"
python3 src/cli/team.py say --from MAIN --to B --type verify --text "Verify endpoints handle all cases"

Example 3: Documentation Generation

Scenario: Write comprehensive API documentation.

Workflow

# MAIN outlines structure
python3 src/cli/team.py say --from MAIN --to all --type assign --text "
  API Documentation Structure:
  - A: Overview & Quick Start
  - B: Authentication endpoints
  - C: User endpoints
  - D: Error handling & rate limits
"

# Agents write in parallel...

# MAIN reviews and consolidates
python3 src/cli/team.py say --from MAIN --to A --type review --text "Review B's section for consistency"
python3 src/cli/team.py say --from MAIN --to B --type review --text "Review C's section"

Result

Complete documentation generated in ~15 minutes instead of 60+ minutes.

Example 4: Bug Investigation

Scenario: Production issue: "API returns 500 on user profile update"

Investigation Workflow

# MAIN describes bug
python3 src/cli/team.py say --from MAIN --to all --type assign --text "
  BUG: Profile update fails with 500 error
  Reproduce: PUT /api/users/{id} with valid data
  Expected: 200 OK
  Actual: 500 Internal Server Error
  Logs: 'NullPointerException in UserValidator'
"

# Agents investigate different angles
python3 src/cli/team.py say --from MAIN --to A --type assign --text "Check validator code"
python3 src/cli/team.py say --from MAIN --to B --type assign --text "Check recent commits"
python3 src/cli/team.py say --from MAIN --to C --type assign --text "Check test coverage"
python3 src/cli/team.py say --from MAIN --to D --type assign --text "Propose hotfix"

Root Cause Analysis (Parallel)

Agent Finding
A Missing null check in UserValidator:42
B Introduced in commit abc123
C No test for null email case
D Hotfix: Add if email is None check

Resolution

# MAIN reviews proposed fix
python3 src/cli/team.py say --from MAIN --to D --type review --text "Review hotfix PR"

# MAIN assigns verification
python3 src/cli/team.py say --from MAIN --to A --type verify --text "Verify fix resolves issue"

Example 5: Architecture Design

Scenario: Design microservices architecture for e-commerce platform.

Phase 1: Research (Parallel)

python3 src/cli/team.py say --from MAIN --to A --type assign --text "Research: Best practices for service communication"
python3 src/cli/team.py say --from MAIN --to B --type assign --text "Research: Database per service vs shared database"
python3 src/cli/team.py say --from MAIN --to C --type assign --text "Research: Authentication patterns (JWT, session, etc.)"
python3 src/cli/team.py say --from MAIN --to D --type assign --text "Research: Deployment strategies (Kubernetes, Docker Compose)"

Phase 2: Design Review (Collaborative)

Each agent presents findings → Others critique → MAIN synthesizes

Phase 3: Documentation

# MAIN assigns documentation sections
python3 src/cli/team.py say --from MAIN --to A --type assign --text "Write: Architecture overview diagram"
python3 src/cli/team.py say --from MAIN --to B --type assign --text "Write: Service boundaries & contracts"
python3 src/cli/team.py say --from MAIN --to C --type assign --text "Write: Data flow diagrams"
python3 src/cli/team.py say --from MAIN --to D --type assign --text "Write: Deployment guide"

Message Protocol Examples

Review Request

{
  "from": "MAIN",
  "to": "A",
  "type": "review",
  "text": "Review this PR for security issues",
  "context": {
    "pr_url": "https://github.com/..."
  }
}

Task Assignment

{
  "from": "MAIN",
  "to": "B",
  "type": "assign",
  "text": "Implement rate limiting middleware",
  "context": {
    "deadline": "2026-01-21",
    "priority": "high"
  }
}

Question

{
  "from": "C",
  "to": "MAIN",
  "type": "clarify",
  "text": "Should rate limit be per-IP or per-user?"
}

Tips for Effective Collaboration

  1. Be Specific: Include file paths, line numbers, error messages
  2. Set Context: Share background info in message text
  3. Check Inbox Regularly: Run inbox command to see new messages
  4. Use Parallel Execution: Assign independent tasks to multiple agents
  5. Verify Results: Use verify message type before marking done

Real-World Performance

Task Type Single AI AgentHub (5 agents) Improvement
Code Review 45 min 15 min 3x faster
Feature Implementation 2 hours 45 min 2.7x faster
Documentation 90 min 25 min 3.6x faster
Bug Investigation 60 min 20 min 3x faster
Architecture Design 3 hours 1 hour 3x faster

Results based on internal testing with similar tasks

Need more examples? Open a Discussion!