Contracts

All Pydantic models live in morning_agents/contracts/models.py.

Enums

Severity

class Severity(str, Enum):
    info = "info"
    warning = "warning"
    action_needed = "action_needed"

AgentStatus

class AgentStatus(str, Enum):
    success = "success"
    partial = "partial"
    error = "error"

---

Core Models

Finding

A single piece of intel from an agent.

Field	Type	Description
id	str	Unique ID, e.g. "brew-001"
source_agent	str	Agent name that produced this
category	str	Free-form category tag
severity	Severity	info / warning / action_needed
title	str	One-line summary
detail	str	Longer description
metadata	dict[str, Any]	Agent-specific data (always include tool_id)
timestamp	AwareDatetime	UTC timestamp

Convention: Always include tool_id in metadata. Cross-reference rules key on it.

---

AgentResult

One agent's complete output.

Field	Type	Description
agent_name	str	Agent identifier
agent_display_name	str	Display string with emoji
status	AgentStatus	Success / partial / error
started_at	AwareDatetime	When the agent started
completed_at	AwareDatetime	When it finished
duration_ms	int	Wall time in milliseconds
findings	list[Finding]	All findings
summary	FindingSummary \| None	Computed by compute_summary()
tool_calls	list[ToolCall]	MCP and API calls made
error	str \| None	Error message if status is error

compute_summary() is called by the orchestrator after agent.run() returns.

---

CrossReference

A correlation between findings from different agents.

Field	Type	Description
id	str	Unique ID, e.g. "xref-node-pr-abc12345"
severity	Severity	Severity of the correlation
title	str	One-line description
detail	str	Full explanation
source_findings	list[str]	Finding IDs that triggered this
source_agents	list[str]	Agent names involved
timestamp	AwareDatetime	When the cross-reference was generated

---

BriefingOutput

The top-level output contract. Everything else is nested here.

Field	Type	Description
version	str	App version
briefing_id	str	e.g. "brief-2026-03-23-091500"
generated_at	AwareDatetime	UTC timestamp
duration_ms	int	Total wall time
agent_results	list[AgentResult]	One per agent
cross_references	list[CrossReference]	Correlated findings
summary	BriefingSummary	Counts and stats
execution	ExecutionMeta	DAG stages, dependency graph, retry counts
config	BriefingConfig	Runtime config snapshot

---

Supporting Models

ToolCall records a single MCP tool or API call:

class ToolCall(BaseModel):
    tool: str         # tool name
    server: str       # "homebrew-mcp", "anthropic", etc.
    duration_ms: int
    success: bool

FindingSummary stores per-agent finding counts:

class FindingSummary(BaseModel):
    total: int
    by_severity: dict[str, int]

BriefingSummary stores briefing-level stats:

class BriefingSummary(BaseModel):
    agents_run: int
    agents_succeeded: int
    agents_failed: int
    total_findings: int
    by_severity: dict[str, int]
    mcp_servers_used: int