empathy-framework 4.1.1__py3-none-any.whl → 4.4.0__py3-none-any.whl

empathy_os/meta_workflows/models.py

@@ -0,0 +1,567 @@
+"""Core data models for meta-workflow system.
+
+This module defines the data structures for:
+- Form schemas and responses (Socratic questioning)
+- Agent composition rules and specs (dynamic agent generation)
+- Meta-workflow templates and results (orchestration)
+
+Created: 2026-01-17
+Purpose: Enable dynamic workflow creation via forms + agent teams
+"""
+
+import json
+import uuid
+from dataclasses import asdict, dataclass, field
+from datetime import datetime
+from enum import Enum
+from typing import Any
+
+# =============================================================================
+# Enums
+# =============================================================================
+
+
+class QuestionType(str, Enum):
+    """Types of form questions supported."""
+
+    TEXT_INPUT = "text_input"
+    SINGLE_SELECT = "single_select"
+    MULTI_SELECT = "multi_select"
+    BOOLEAN = "boolean"
+
+
+class TierStrategy(str, Enum):
+    """Model tier escalation strategies for agents."""
+
+    CHEAP_ONLY = "cheap_only"  # Use only cheap tier
+    PROGRESSIVE = "progressive"  # Escalate cheap → capable → premium
+    CAPABLE_FIRST = "capable_first"  # Start at capable, escalate to premium if needed
+    PREMIUM_ONLY = "premium_only"  # Use only premium tier
+
+
+# =============================================================================
+# Form Components
+# =============================================================================
+
+
+@dataclass
+class FormQuestion:
+    """A single question in a Socratic form.
+
+    Attributes:
+        id: Unique identifier for this question
+        text: The question text shown to user
+        type: Question type (text_input, single_select, etc.)
+        options: Available options for select questions
+        default: Default value if user doesn't provide one
+        help_text: Additional help text shown to user
+        required: Whether this question must be answered
+    """
+
+    id: str
+    text: str
+    type: QuestionType
+    options: list[str] = field(default_factory=list)
+    default: str | None = None
+    help_text: str | None = None
+    required: bool = True
+
+    def to_ask_user_format(self) -> dict[str, Any]:
+        """Convert to format compatible with AskUserQuestion tool.
+
+        Returns:
+            Dictionary with question data for AskUserQuestion
+        """
+        # Boolean questions convert to Yes/No select
+        if self.type == QuestionType.BOOLEAN:
+            return {
+                "question_id": self.id,
+                "question": self.text,
+                "type": "single_select",
+                "options": ["Yes", "No"],
+                "default": self.default,  # Preserve default for boolean questions
+                "help_text": self.help_text,
+            }
+
+        return {
+            "question_id": self.id,
+            "question": self.text,
+            "type": self.type.value,
+            "options": self.options,
+            "default": self.default,
+            "help_text": self.help_text,
+        }
+
+
+@dataclass
+class FormSchema:
+    """Schema defining a collection of questions for a meta-workflow.
+
+    Attributes:
+        title: Form title
+        description: Form description
+        questions: List of questions to ask
+    """
+
+    title: str
+    description: str
+    questions: list[FormQuestion] = field(default_factory=list)
+
+    def get_question_batches(self, batch_size: int = 4) -> list[list[FormQuestion]]:
+        """Batch questions for asking (AskUserQuestion supports max 4 at once).
+
+        Args:
+            batch_size: Maximum questions per batch (default: 4)
+
+        Returns:
+            List of question batches
+        """
+        batches = []
+        for i in range(0, len(self.questions), batch_size):
+            batches.append(self.questions[i : i + batch_size])
+        return batches
+
+
+@dataclass
+class FormResponse:
+    """User's responses to a form.
+
+    Attributes:
+        template_id: ID of template this response is for
+        responses: Dictionary mapping question_id → user's answer
+        timestamp: When response was submitted
+        response_id: Unique ID for this response
+    """
+
+    template_id: str
+    responses: dict[str, Any] = field(default_factory=dict)
+    timestamp: str = field(default_factory=lambda: datetime.now().isoformat())
+    response_id: str = field(default_factory=lambda: f"resp-{datetime.now().strftime('%Y%m%d-%H%M%S')}")
+
+    def get(self, question_id: str, default: Any = None) -> Any:
+        """Get response for a question.
+
+        Args:
+            question_id: ID of question
+            default: Default value if not found
+
+        Returns:
+            User's response or default
+        """
+        return self.responses.get(question_id, default)
+
+
+# =============================================================================
+# Agent Components
+# =============================================================================
+
+
+@dataclass
+class AgentCompositionRule:
+    """Rule defining when and how to create an agent.
+
+    Attributes:
+        role: Agent's role/purpose
+        base_template: Base agent template to use
+        tier_strategy: Model tier escalation strategy
+        tools: List of tools agent can use
+        required_responses: Conditions that must be met (question_id → required value)
+        config_mapping: Map form responses to agent config (form_key → config_key)
+        success_criteria: List of success criteria for this agent
+    """
+
+    role: str
+    base_template: str
+    tier_strategy: TierStrategy
+    tools: list[str] = field(default_factory=list)
+    required_responses: dict[str, str | list[str]] = field(default_factory=dict)
+    config_mapping: dict[str, str] = field(default_factory=dict)
+    success_criteria: list[str] = field(default_factory=list)
+
+    def should_create(self, response: FormResponse) -> bool:
+        """Check if agent should be created based on form responses.
+
+        Args:
+            response: User's form responses
+
+        Returns:
+            True if all required conditions are met
+        """
+        for key, required_value in self.required_responses.items():
+            user_response = response.get(key)
+
+            # Multi-select: check if required value in list
+            if isinstance(user_response, list):
+                # If required_value is a list, check if ANY match
+                if isinstance(required_value, list):
+                    if not any(rv in user_response for rv in required_value):
+                        return False
+                # Single required value must be in user's selections
+                else:
+                    if required_value not in user_response:
+                        return False
+
+            # Single value: exact match required
+            else:
+                if isinstance(required_value, list):
+                    # User must have picked one of the allowed values
+                    if user_response not in required_value:
+                        return False
+                else:
+                    # Exact match
+                    if user_response != required_value:
+                        return False
+
+        return True
+
+    def create_agent_config(self, response: FormResponse) -> dict[str, Any]:
+        """Create agent configuration from form responses.
+
+        Args:
+            response: User's form responses
+
+        Returns:
+            Agent configuration dictionary
+        """
+        config = {}
+        for form_key, config_key in self.config_mapping.items():
+            value = response.get(form_key)
+            if value is not None:
+                config[config_key] = value
+        return config
+
+
+@dataclass
+class AgentSpec:
+    """Specification for a dynamically created agent instance.
+
+    Attributes:
+        role: Agent's role/purpose
+        base_template: Base agent template used
+        tier_strategy: Model tier escalation strategy
+        tools: List of tools agent can use
+        config: Agent-specific configuration
+        success_criteria: List of success criteria
+        agent_id: Unique identifier for this agent instance
+    """
+
+    role: str
+    base_template: str
+    tier_strategy: TierStrategy
+    tools: list[str] = field(default_factory=list)
+    config: dict[str, Any] = field(default_factory=dict)
+    success_criteria: list[str] = field(default_factory=list)
+    agent_id: str = field(default_factory=lambda: f"agent-{uuid.uuid4().hex[:12]}")
+
+
+# =============================================================================
+# Meta-Workflow Components
+# =============================================================================
+
+
+@dataclass
+class MetaWorkflowTemplate:
+    """Template defining a complete meta-workflow.
+
+    Attributes:
+        template_id: Unique identifier for this template
+        name: Human-readable name
+        description: Description of what this workflow does
+        version: Template version
+        tags: Tags for categorization
+        author: Template author
+        form_schema: Form questions to ask user
+        agent_composition_rules: Rules for creating agents
+        estimated_cost_range: Estimated cost range (min, max)
+        estimated_duration_minutes: Estimated duration in minutes
+    """
+
+    template_id: str
+    name: str
+    description: str
+    form_schema: FormSchema
+    agent_composition_rules: list[AgentCompositionRule] = field(default_factory=list)
+    version: str = "1.0.0"
+    tags: list[str] = field(default_factory=list)
+    author: str = "system"
+    estimated_cost_range: tuple[float, float] = (0.05, 0.50)
+    estimated_duration_minutes: int = 5
+
+    def to_json(self) -> str:
+        """Serialize template to JSON string.
+
+        Returns:
+            JSON string representation
+        """
+        data = {
+            "template_id": self.template_id,
+            "name": self.name,
+            "description": self.description,
+            "version": self.version,
+            "tags": self.tags,
+            "author": self.author,
+            "estimated_cost_range": list(self.estimated_cost_range),
+            "estimated_duration_minutes": self.estimated_duration_minutes,
+            "form_schema": {
+                "title": self.form_schema.title,
+                "description": self.form_schema.description,
+                "questions": [
+                    {
+                        "id": q.id,
+                        "text": q.text,
+                        "type": q.type.value,
+                        "options": q.options,
+                        "default": q.default,
+                        "help_text": q.help_text,
+                        "required": q.required,
+                    }
+                    for q in self.form_schema.questions
+                ],
+            },
+            "agent_composition_rules": [
+                {
+                    "role": rule.role,
+                    "base_template": rule.base_template,
+                    "tier_strategy": rule.tier_strategy.value,
+                    "tools": rule.tools,
+                    "required_responses": rule.required_responses,
+                    "config_mapping": rule.config_mapping,
+                    "success_criteria": rule.success_criteria,
+                }
+                for rule in self.agent_composition_rules
+            ],
+        }
+        return json.dumps(data, indent=2)
+
+    @staticmethod
+    def from_json(json_str: str) -> "MetaWorkflowTemplate":
+        """Deserialize template from JSON string.
+
+        Args:
+            json_str: JSON string representation
+
+        Returns:
+            MetaWorkflowTemplate instance
+
+        Raises:
+            ValueError: If JSON is invalid or missing required fields
+        """
+        try:
+            data = json.loads(json_str)
+        except json.JSONDecodeError as e:
+            raise ValueError(f"Invalid JSON: {e}") from e
+
+        # Parse form schema
+        form_schema = FormSchema(
+            title=data["form_schema"]["title"],
+            description=data["form_schema"]["description"],
+            questions=[
+                FormQuestion(
+                    id=q["id"],
+                    text=q["text"],
+                    type=QuestionType(q["type"]),
+                    options=q.get("options", []),
+                    default=q.get("default"),
+                    help_text=q.get("help_text"),
+                    required=q.get("required", True),
+                )
+                for q in data["form_schema"]["questions"]
+            ],
+        )
+
+        # Parse agent composition rules
+        rules = [
+            AgentCompositionRule(
+                role=rule["role"],
+                base_template=rule["base_template"],
+                tier_strategy=TierStrategy(rule["tier_strategy"]),
+                tools=rule.get("tools", []),
+                required_responses=rule.get("required_responses", {}),
+                config_mapping=rule.get("config_mapping", {}),
+                success_criteria=rule.get("success_criteria", []),
+            )
+            for rule in data.get("agent_composition_rules", [])
+        ]
+
+        return MetaWorkflowTemplate(
+            template_id=data["template_id"],
+            name=data["name"],
+            description=data["description"],
+            version=data.get("version", "1.0.0"),
+            tags=data.get("tags", []),
+            author=data.get("author", "system"),
+            form_schema=form_schema,
+            agent_composition_rules=rules,
+            estimated_cost_range=tuple(data.get("estimated_cost_range", [0.05, 0.50])),
+            estimated_duration_minutes=data.get("estimated_duration_minutes", 5),
+        )
+
+
+@dataclass
+class AgentExecutionResult:
+    """Result from executing a single agent.
+
+    Attributes:
+        agent_id: ID of agent that executed
+        role: Agent's role
+        success: Whether agent succeeded
+        cost: Cost of execution
+        duration: Duration in seconds
+        tier_used: Model tier used
+        output: Agent's output/result
+        error: Error message if failed
+    """
+
+    agent_id: str
+    role: str
+    success: bool
+    cost: float
+    duration: float
+    tier_used: str
+    output: str | dict[str, Any]
+    error: str | None = None
+
+
+@dataclass
+class MetaWorkflowResult:
+    """Result from executing a complete meta-workflow.
+
+    Attributes:
+        run_id: Unique ID for this execution
+        template_id: ID of template used
+        timestamp: When execution started
+        form_responses: User's form responses
+        agents_created: List of agents that were created
+        agent_results: Results from agent executions
+        total_cost: Total cost of execution
+        total_duration: Total duration in seconds
+        success: Whether workflow succeeded
+        error: Error message if failed
+    """
+
+    run_id: str
+    template_id: str
+    timestamp: str
+    form_responses: FormResponse
+    agents_created: list[AgentSpec] = field(default_factory=list)
+    agent_results: list[AgentExecutionResult] = field(default_factory=list)
+    total_cost: float = 0.0
+    total_duration: float = 0.0
+    success: bool = True
+    error: str | None = None
+
+    def to_dict(self) -> dict[str, Any]:
+        """Convert result to dictionary for serialization.
+
+        Returns:
+            Dictionary representation
+        """
+        return {
+            "run_id": self.run_id,
+            "template_id": self.template_id,
+            "timestamp": self.timestamp,
+            "form_responses": asdict(self.form_responses),
+            "agents_created": [asdict(agent) for agent in self.agents_created],
+            "agent_results": [asdict(result) for result in self.agent_results],
+            "total_cost": self.total_cost,
+            "total_duration": self.total_duration,
+            "success": self.success,
+            "error": self.error,
+        }
+
+    def to_json(self) -> str:
+        """Serialize result to JSON string.
+
+        Returns:
+            JSON string representation
+        """
+        return json.dumps(self.to_dict(), indent=2)
+
+    @staticmethod
+    def from_dict(data: dict[str, Any]) -> "MetaWorkflowResult":
+        """Create result from dictionary.
+
+        Args:
+            data: Dictionary representation
+
+        Returns:
+            MetaWorkflowResult instance
+        """
+        form_responses = FormResponse(
+            template_id=data["form_responses"]["template_id"],
+            responses=data["form_responses"]["responses"],
+            timestamp=data["form_responses"]["timestamp"],
+            response_id=data["form_responses"]["response_id"],
+        )
+
+        agents_created = [
+            AgentSpec(
+                role=agent["role"],
+                base_template=agent["base_template"],
+                tier_strategy=TierStrategy(agent["tier_strategy"]),
+                tools=agent["tools"],
+                config=agent["config"],
+                success_criteria=agent["success_criteria"],
+                agent_id=agent["agent_id"],
+            )
+            for agent in data.get("agents_created", [])
+        ]
+
+        agent_results = [
+            AgentExecutionResult(
+                agent_id=result["agent_id"],
+                role=result["role"],
+                success=result["success"],
+                cost=result["cost"],
+                duration=result["duration"],
+                tier_used=result["tier_used"],
+                output=result["output"],
+                error=result.get("error"),
+            )
+            for result in data.get("agent_results", [])
+        ]
+
+        return MetaWorkflowResult(
+            run_id=data["run_id"],
+            template_id=data["template_id"],
+            timestamp=data["timestamp"],
+            form_responses=form_responses,
+            agents_created=agents_created,
+            agent_results=agent_results,
+            total_cost=data.get("total_cost", 0.0),
+            total_duration=data.get("total_duration", 0.0),
+            success=data.get("success", True),
+            error=data.get("error"),
+        )
+
+
+# =============================================================================
+# Insight/Analytics Components
+# =============================================================================
+
+
+@dataclass
+class PatternInsight:
+    """An insight learned from analyzing workflow patterns.
+
+    Attributes:
+        insight_type: Type of insight (cost, tier_performance, agent_count, etc.)
+        description: Human-readable description
+        confidence: Confidence level (0.0 to 1.0)
+        data: Supporting data for this insight
+        sample_size: Number of runs this insight is based on
+    """
+
+    insight_type: str
+    description: str
+    confidence: float
+    data: dict[str, Any] = field(default_factory=dict)
+    sample_size: int = 0
+
+    def to_dict(self) -> dict[str, Any]:
+        """Convert to dictionary.
+
+        Returns:
+            Dictionary representation
+        """
+        return asdict(self)