@bluefly/openstandardagents 0.4.4 → 0.4.5

package/examples/migrations/swarm-to-ossa/before-handoffs.py

@@ -1,373 +0,0 @@
-"""
-OpenAI Swarm: Advanced Handoff Patterns
-
-This example demonstrates various handoff patterns in OpenAI Swarm:
-1. Conditional handoffs
-2. Bidirectional handoffs (agents can transfer back and forth)
-3. Context preservation across handoffs
-4. Handoff with state changes
-
-Swarm Documentation: https://github.com/openai/swarm
-"""
-
-from swarm import Swarm, Agent
-from typing import Dict, Any, Optional
-
-# Initialize Swarm client
-client = Swarm()
-
-# ============================================================================
-# Shared State (simulated database)
-# ============================================================================
-
-conversation_state = {
-    "order_id": None,
-    "customer_tier": "standard",
-    "issue_resolved": False,
-    "escalation_count": 0
-}
-
-# ============================================================================
-# Transfer Functions with Context
-# ============================================================================
-
-def transfer_to_order_verification():
-    """Transfer to order verification agent to validate order details."""
-    print("[HANDOFF] → Order Verification Agent")
-    return order_verification_agent
-
-
-def transfer_to_payment_processing():
-    """Transfer to payment processing agent after order verification."""
-    print("[HANDOFF] → Payment Processing Agent")
-    return payment_processing_agent
-
-
-def transfer_to_fulfillment():
-    """Transfer to fulfillment agent after payment is processed."""
-    print("[HANDOFF] → Fulfillment Agent")
-    return fulfillment_agent
-
-
-def transfer_to_supervisor():
-    """Escalate to supervisor for complex issues."""
-    global conversation_state
-    conversation_state["escalation_count"] += 1
-    print(f"[ESCALATION] → Supervisor (Escalation #{conversation_state['escalation_count']})")
-    return supervisor_agent
-
-
-def transfer_back_to_sales():
-    """Return customer to sales agent."""
-    print("[HANDOFF] ← Sales Agent")
-    return sales_agent
-
-
-# ============================================================================
-# Conditional Transfer Functions
-# ============================================================================
-
-def conditional_transfer_payment(context_variables: Dict[str, Any]):
-    """
-    Conditionally transfer to payment based on order total.
-    High-value orders go to supervisor for approval.
-    """
-    order_total = context_variables.get("order_total", 0)
-
-    if order_total > 10000:
-        print(f"[CONDITIONAL] High-value order (${order_total}) → Supervisor")
-        return supervisor_agent
-    else:
-        print(f"[CONDITIONAL] Standard order (${order_total}) → Payment Processing")
-        return payment_processing_agent
-
-
-# ============================================================================
-# Agent Functions
-# ============================================================================
-
-def verify_order(order_id: str, customer_id: str) -> str:
-    """Verify order details and customer eligibility."""
-    global conversation_state
-    conversation_state["order_id"] = order_id
-
-    # Simulated verification
-    return f"Order {order_id} verified for customer {customer_id}. Ready for payment."
-
-
-def process_payment(payment_method: str, amount: float) -> str:
-    """Process payment for the order."""
-    # Simulated payment processing
-    return f"Payment of ${amount:.2f} processed via {payment_method}. Transaction ID: TXN-{payment_method}-001"
-
-
-def schedule_fulfillment(order_id: str, delivery_address: str) -> str:
-    """Schedule order fulfillment and delivery."""
-    # Simulated fulfillment
-    return f"Order {order_id} scheduled for delivery to {delivery_address}. Expected delivery: 3-5 business days."
-
-
-def approve_order(order_id: str, approval_note: str) -> str:
-    """Supervisor approves high-value orders."""
-    return f"Order {order_id} approved by supervisor. Note: {approval_note}"
-
-
-# ============================================================================
-# Agent Definitions with Handoff Patterns
-# ============================================================================
-
-# Sales Agent - Entry point
-sales_agent = Agent(
-    name="Sales Agent",
-    instructions="""You are a sales agent helping customers place orders.
-
-When a customer wants to order:
-1. Collect order details (items, quantities)
-2. Calculate order total
-3. Transfer to Order Verification Agent to validate the order
-
-Be friendly and helpful!""",
-    functions=[transfer_to_order_verification]
-)
-
-# Order Verification Agent
-order_verification_agent = Agent(
-    name="Order Verification Agent",
-    instructions="""You verify order details and customer eligibility.
-
-Process:
-1. Verify the order details provided by the customer
-2. Check customer eligibility
-3. Transfer to Payment Processing Agent when verification is complete
-
-If there are issues with the order, transfer back to Sales Agent.""",
-    functions=[verify_order, transfer_to_payment_processing, transfer_back_to_sales]
-)
-
-# Payment Processing Agent
-payment_processing_agent = Agent(
-    name="Payment Processing Agent",
-    instructions="""You handle payment processing for orders.
-
-Process:
-1. Collect payment information
-2. Process the payment securely
-3. Transfer to Fulfillment Agent when payment is successful
-
-For high-value orders (>$10,000), escalate to Supervisor for approval first.
-If payment fails, transfer back to Sales Agent.""",
-    functions=[process_payment, transfer_to_fulfillment, transfer_to_supervisor, transfer_back_to_sales]
-)
-
-# Fulfillment Agent
-fulfillment_agent = Agent(
-    name="Fulfillment Agent",
-    instructions="""You handle order fulfillment and delivery.
-
-Process:
-1. Schedule order fulfillment
-2. Arrange delivery
-3. Provide tracking information
-
-Mark the issue as resolved when delivery is scheduled.""",
-    functions=[schedule_fulfillment, transfer_back_to_sales]
-)
-
-# Supervisor Agent - Handles escalations
-supervisor_agent = Agent(
-    name="Supervisor Agent",
-    instructions="""You are a supervisor handling escalated cases.
-
-Your role:
-1. Review escalated orders (especially high-value ones)
-2. Approve or modify orders as needed
-3. Resolve complex customer issues
-
-You can transfer to any other agent as needed.""",
-    functions=[
-        approve_order,
-        transfer_to_payment_processing,
-        transfer_to_fulfillment,
-        transfer_back_to_sales
-    ]
-)
-
-# ============================================================================
-# Usage Examples
-# ============================================================================
-
-def example_simple_handoff():
-    """Example 1: Simple linear handoff flow."""
-    print("\n" + "=" * 80)
-    print("Example 1: Simple Handoff Flow (Sales → Verification → Payment → Fulfillment)")
-    print("=" * 80)
-
-    messages = [
-        {"role": "user", "content": "I want to order 2 premium widgets for $500 total"}
-    ]
-
-    context_variables = {
-        "customer_id": "CUST123",
-        "customer_tier": "premium",
-        "order_total": 500
-    }
-
-    # Start with sales agent
-    response = client.run(
-        agent=sales_agent,
-        messages=messages,
-        context_variables=context_variables
-    )
-
-    print(f"Current Agent: {response.agent.name}")
-    print(f"Response: {response.messages[-1]['content']}")
-
-    # Continue to verification
-    messages = response.messages
-    messages.append({"role": "user", "content": "Yes, verify the order please"})
-
-    response = client.run(
-        agent=response.agent,
-        messages=messages,
-        context_variables=context_variables
-    )
-
-    print(f"Current Agent: {response.agent.name}")
-    print(f"Response: {response.messages[-1]['content']}")
-
-
-def example_conditional_handoff():
-    """Example 2: Conditional handoff based on order value."""
-    print("\n" + "=" * 80)
-    print("Example 2: Conditional Handoff (High-Value Order → Supervisor)")
-    print("=" * 80)
-
-    messages = [
-        {"role": "user", "content": "I want to place a large enterprise order worth $15,000"}
-    ]
-
-    context_variables = {
-        "customer_id": "CUST456",
-        "customer_tier": "enterprise",
-        "order_total": 15000
-    }
-
-    response = client.run(
-        agent=sales_agent,
-        messages=messages,
-        context_variables=context_variables
-    )
-
-    print(f"Current Agent: {response.agent.name}")
-    print(f"Context: {context_variables}")
-
-
-def example_bidirectional_handoff():
-    """Example 3: Bidirectional handoff (agent can go back)."""
-    print("\n" + "=" * 80)
-    print("Example 3: Bidirectional Handoff (Payment fails → Back to Sales)")
-    print("=" * 80)
-
-    messages = [
-        {"role": "user", "content": "My payment failed, what should I do?"}
-    ]
-
-    context_variables = {
-        "customer_id": "CUST789",
-        "payment_failed": True
-    }
-
-    # Start at payment processing agent
-    response = client.run(
-        agent=payment_processing_agent,
-        messages=messages,
-        context_variables=context_variables
-    )
-
-    print(f"Current Agent: {response.agent.name}")
-    print(f"Response: {response.messages[-1]['content']}")
-
-
-def example_context_preservation():
-    """Example 4: Context preservation across handoffs."""
-    print("\n" + "=" * 80)
-    print("Example 4: Context Preservation Across Handoffs")
-    print("=" * 80)
-
-    initial_context = {
-        "customer_id": "CUST999",
-        "customer_tier": "premium",
-        "order_total": 750,
-        "order_items": ["widget-pro", "widget-plus"],
-        "delivery_address": "123 Main St, City, State 12345"
-    }
-
-    messages = [
-        {"role": "user", "content": "I'm ready to complete my order"}
-    ]
-
-    response = client.run(
-        agent=sales_agent,
-        messages=messages,
-        context_variables=initial_context
-    )
-
-    print(f"Initial Context: {initial_context}")
-    print(f"Current Agent: {response.agent.name}")
-    print("Context is automatically preserved across all handoffs!")
-
-
-# ============================================================================
-# Swarm Handoff Limitations
-# ============================================================================
-
-def demonstrate_limitations():
-    """Demonstrate limitations of Swarm's handoff system."""
-    print("\n" + "=" * 80)
-    print("Swarm Handoff Limitations")
-    print("=" * 80)
-
-    limitations = [
-        "1. NO DECLARATIVE HANDOFF RULES: Handoffs are coded in functions, not declared",
-        "2. NO CONDITIONAL SYNTAX: Must write Python code for conditions",
-        "3. NO HANDOFF OBSERVABILITY: Can't track handoff metrics or patterns",
-        "4. NO HANDOFF POLICIES: Can't enforce rules like 'max 3 handoffs'",
-        "5. MANUAL CONTEXT MANAGEMENT: Must manually pass context_variables",
-        "6. NO HANDOFF APPROVAL: Can't require approval for certain handoffs",
-        "7. NO HANDOFF ROLLBACK: Can't undo a handoff if it was wrong",
-        "8. HARD TO TEST: Handoff logic scattered across functions",
-    ]
-
-    for limitation in limitations:
-        print(f"  ❌ {limitation}")
-
-    print("\n  ✓ OSSA solves ALL of these with declarative handoff configuration!")
-    print("  See after-handoffs.ossa.yaml for the solution.")
-
-
-# ============================================================================
-# Main
-# ============================================================================
-
-if __name__ == "__main__":
-    print("OpenAI Swarm - Advanced Handoff Patterns")
-    print("=" * 80)
-
-    # Run examples
-    example_simple_handoff()
-    example_conditional_handoff()
-    example_bidirectional_handoff()
-    example_context_preservation()
-    demonstrate_limitations()
-
-    print("\n" + "=" * 80)
-    print("Migration to OSSA")
-    print("=" * 80)
-    print("See after-handoffs.ossa.yaml for declarative handoff configuration.")
-    print("\nOSSA Benefits:")
-    print("  ✓ Declarative handoff rules (YAML, not code)")
-    print("  ✓ Conditional handoffs with expression syntax")
-    print("  ✓ Automatic context propagation (no manual passing)")
-    print("  ✓ Handoff observability (metrics, tracing, logging)")
-    print("  ✓ Handoff policies (max handoffs, approval rules)")
-    print("  ✓ Easy testing and validation")

package/examples/v0.3.6/full-featured-v0.3.6.ossa.yaml

@@ -1,304 +0,0 @@
-apiVersion: ossa/v0.3.6
-kind: Agent
-metadata:
-  name: gitlab-mr-reviewer-pro
-  version: 1.0.0
-  description: "Enterprise-grade MR reviewer with token efficiency, composite identity, and knowledge graph"
-  efficiency_tier: economy  # Aggressive optimization
-
-  # Agent Catalog Metadata (v0.3.6+)
-  catalog:
-    published: true
-    catalog_id: gitlab-mr-reviewer-pro
-    visibility: internal
-    categories:
-      - code-quality
-      - security
-      - gitlab
-      - code-review
-    tags:
-      - gitlab-duo
-      - token-efficiency
-      - knowledge-graph
-      - composite-identity
-      - enterprise
-    icon_url: https://example.com/icons/mr-reviewer.svg
-    documentation_url: https://docs.example.com/mr-reviewer
-    pricing:
-      model: enterprise
-      subscription_monthly: 499
-    ratings:
-      average: 4.8
-      count: 127
-
-  labels:
-    cost-tier: economy
-    security-level: high
-    compliance: sox-gdpr
-    environment: production
-
-spec:
-  role: "Expert code reviewer with deep codebase understanding, security focus, and 95% cost efficiency"
-
-  # Composite Identity (GitLab Duo v0.3.6+)
-  identity:
-    provider: gitlab
-
-    composite:
-      enabled: true
-      merge_strategy: restrictive
-
-      primary_identity:
-        username: mr-review-bot
-        email: mr-review-bot@company.com
-        roles:
-          - developer
-
-      secondary_identity:
-        user_id: $GITLAB_USER_ID
-        username: $GITLAB_USERNAME
-        email: user@example.com
-
-      permissions:
-        read:
-          - repositories
-          - merge_requests
-          - pipelines
-        write:
-          - merge_request_comments
-        execute:
-          - code_analysis
-          - security_scan
-
-    authentication:
-      method: oauth2
-      scopes:
-        - api
-        - read_repository
-        - write_repository
-      auto_refresh: true
-
-    token_source:
-      env_var: GITLAB_COMPOSITE_TOKEN
-
-  llm:
-    provider: anthropic
-    model: claude-opus-4
-    temperature: 0.2
-    max_tokens: 8192
-
-  # Token Efficiency (v0.3.6+) - 95% savings
-  token_efficiency:
-    enabled: true
-
-    # Context Management: 70% savings
-    context_management:
-      strategy: adaptive
-      pruning:
-        enabled: true
-        threshold: 0.3
-        max_tokens: 50000
-        preserve_recent: 10
-      summarization:
-        enabled: true
-        model: haiku
-        compression_ratio: 0.2
-        trigger_threshold: 100000
-      caching:
-        enabled: true
-        ttl: 600
-        strategy: hybrid
-
-    # Prompt Caching: 90% savings
-    prompt_caching:
-      enabled: true
-      provider: anthropic
-      strategy: hybrid
-      cache_breakpoints:
-        - system_prompt
-        - knowledge_base
-        - recent_history
-        - tools
-      min_tokens: 1024
-      ttl: 300
-      estimated_savings: 90
-
-    # Token Budgets: Hard limits
-    token_budget:
-      enabled: true
-      input_limit: 150000
-      output_limit: 8192
-      total_limit: 160000
-      enforcement: adaptive
-      overflow_behavior: summarize
-
-    # Tool Output Limits
-    tool_output_limits:
-      enabled: true
-      max_tokens_per_tool: 10000
-      max_total_tokens: 100000
-      truncation_strategy: smart
-      large_output_handling: summarize
-
-    # Streaming for reduced latency
-    streaming:
-      enabled: true
-      chunk_size: 512
-
-    # Knowledge Graph: 98% savings
-    knowledge_graph:
-      enabled: true
-      max_context_tokens: 15000
-      indexing_strategy: selective
-
-    # Checkpoint Compression
-    checkpoint_compression:
-      enabled: true
-      strategy: delta
-      compression_ratio: 0.15
-
-    # Estimated Savings
-    estimated_savings:
-      percentage: 95
-      baseline_cost: 15
-      optimized_cost: 0.75
-
-  # Knowledge Sources with Compliance Enforcement
-  knowledge_sources:
-    - name: codebase-graph
-      provider: custom
-      config:
-        type: knowledge_graph
-        mcp_server: gitlab-kg
-
-  mandatory_knowledge_sources:
-    - source_id: react-best-practices
-      enforcement_level: query
-      triggers:
-        - "**/*.tsx"
-        - "**/*.jsx"
-    - source_id: typescript-standards
-      enforcement_level: validate
-      triggers:
-        - "**/*.ts"
-        - "**/*.tsx"
-    - source_id: security-patterns
-      enforcement_level: block
-      triggers:
-        - "**/*auth*.ts"
-        - "**/*security*.ts"
-
-  # MCP Extensions
-  extensions:
-    mcp:
-      servers:
-        - name: gitlab-kg
-          transport: http
-          oauth2:
-            enabled: true
-            registration_endpoint: https://gitlab.com/oauth/applications
-            scopes:
-              - api
-              - read_repository
-
-      tools:
-        - name: query_codebase
-          description: "Semantic code search"
-          server: gitlab-kg
-
-        - name: get_relationships
-          description: "Code relationships"
-          server: gitlab-kg
-
-        - name: security_scan
-          description: "Security vulnerability scan"
-          server: gitlab-kg
-
-  tools:
-    - type: mcp
-      name: gitlab-mr-tools
-      description: "GitLab MR operations"
-
-    - type: function
-      name: static-analysis
-      description: "ESLint, TypeScript, Prettier checks"
-
-    - type: function
-      name: security-scan
-      description: "SAST, dependency vulnerabilities"
-
-  autonomy:
-    level: assisted
-    approval_required:
-      - merge_request_approval
-      - code_modification
-    allowed_actions:
-      - add_comment
-      - request_changes
-      - run_analysis
-    blocked_actions:
-      - approve_mr
-      - merge_mr
-      - modify_code
-
-  observability:
-    enabled: true
-    tracing:
-      enabled: true
-      provider: opentelemetry
-      sampling_rate: 1
-    metrics:
-      track_tokens: true
-      track_cost: true
-      track_latency: true
-      track_quality: true
-    logging:
-      level: info
-      structured: true
-
-  safety:
-    sandboxing:
-      enabled: true
-      mode: strict
-    content_filtering:
-      enabled: true
-      blocklist:
-        - credentials
-        - api-keys
-        - passwords
-    rate_limiting:
-      enabled: true
-      requests_per_minute: 30
-
-  # Runtime Configuration
-  runtime:
-    max_turns: 20
-    checkpoint_interval: 5
-    environments:
-      production:
-        max_turns: 15
-        timeout: 300
-      staging:
-        max_turns: 20
-        timeout: 600
-
-  # Completion Signals
-  completion:
-    enabled: true
-    default_signal: complete
-    signals:
-      - signal: complete
-        description: "Review completed successfully"
-      - signal: blocked
-        description: "Security issues found"
-      - signal: escalate
-        description: "Complex issues require human review"
-
-  # Checkpointing
-  checkpointing:
-    enabled: true
-    strategy: interval
-    storage:
-      backend: agent-brain
-      location: redis://checkpoint-store:6379
-      retention: 7d