feat: Phase 5.3 - Multi-Agent Learning Infrastructure
Implement intelligent agent learning from Knowledge Graph execution history
with per-task-type expertise tracking, recency bias, and learning curves.
## Phase 5.3 Implementation
### Learning Infrastructure (✅ Complete)
- LearningProfileService with per-task-type expertise metrics
- TaskTypeExpertise model tracking success_rate, confidence, learning curves
- Recency bias weighting: recent 7 days weighted 3x higher (exponential decay)
- Confidence scoring prevents overfitting: min(1.0, executions / 20)
- Learning curves computed from daily execution windows
### Agent Scoring Service (✅ Complete)
- Unified AgentScore combining SwarmCoordinator + learning profiles
- Scoring formula: 0.3*base + 0.5*expertise + 0.2*confidence
- Rank agents by combined score for intelligent assignment
- Support for recency-biased scoring (recent_success_rate)
- Methods: rank_agents, select_best, rank_agents_with_recency
### KG Integration (✅ Complete)
- KGPersistence::get_executions_for_task_type() - query by agent + task type
- KGPersistence::get_agent_executions() - all executions for agent
- Coordinator::load_learning_profile_from_kg() - core KG→Learning integration
- Coordinator::load_all_learning_profiles() - batch load for multiple agents
- Convert PersistedExecution → ExecutionData for learning calculations
### Agent Assignment Integration (✅ Complete)
- AgentCoordinator uses learning profiles for task assignment
- extract_task_type() infers task type from title/description
- assign_task() scores candidates using AgentScoringService
- Fallback to load-based selection if no learning data available
- Learning profiles stored in coordinator.learning_profiles RwLock
### Profile Adapter Enhancements (✅ Complete)
- create_learning_profile() - initialize empty profiles
- add_task_type_expertise() - set task-type expertise
- update_profile_with_learning() - update swarm profiles from learning
## Files Modified
### vapora-knowledge-graph/src/persistence.rs (+30 lines)
- get_executions_for_task_type(agent_id, task_type, limit)
- get_agent_executions(agent_id, limit)
### vapora-agents/src/coordinator.rs (+100 lines)
- load_learning_profile_from_kg() - core KG integration method
- load_all_learning_profiles() - batch loading for agents
- assign_task() already uses learning-based scoring via AgentScoringService
### Existing Complete Implementation
- vapora-knowledge-graph/src/learning.rs - calculation functions
- vapora-agents/src/learning_profile.rs - data structures and expertise
- vapora-agents/src/scoring.rs - unified scoring service
- vapora-agents/src/profile_adapter.rs - adapter methods
## Tests Passing
- learning_profile: 7 tests ✅
- scoring: 5 tests ✅
- profile_adapter: 6 tests ✅
- coordinator: learning-specific tests ✅
## Data Flow
1. Task arrives → AgentCoordinator::assign_task()
2. Extract task_type from description
3. Query KG for task-type executions (load_learning_profile_from_kg)
4. Calculate expertise with recency bias
5. Score candidates (SwarmCoordinator + learning)
6. Assign to top-scored agent
7. Execution result → KG → Update learning profiles
## Key Design Decisions
✅ Recency bias: 7-day half-life with 3x weight for recent performance
✅ Confidence scoring: min(1.0, total_executions / 20) prevents overfitting
✅ Hierarchical scoring: 30% base load, 50% expertise, 20% confidence
✅ KG query limit: 100 recent executions per task-type for performance
✅ Async loading: load_learning_profile_from_kg supports concurrent loads
## Next: Phase 5.4 - Cost Optimization
Ready to implement budget enforcement and cost-aware provider selection.
2026-01-11 13:03:53 +00:00
|
|
|
// Integration tests for SwarmCoordinator integration with AgentCoordinator
|
2026-01-11 21:46:08 +00:00
|
|
|
// Tests verify swarm task assignment, profile synchronization, and metrics
|
|
|
|
|
// integration
|
feat: Phase 5.3 - Multi-Agent Learning Infrastructure
Implement intelligent agent learning from Knowledge Graph execution history
with per-task-type expertise tracking, recency bias, and learning curves.
## Phase 5.3 Implementation
### Learning Infrastructure (✅ Complete)
- LearningProfileService with per-task-type expertise metrics
- TaskTypeExpertise model tracking success_rate, confidence, learning curves
- Recency bias weighting: recent 7 days weighted 3x higher (exponential decay)
- Confidence scoring prevents overfitting: min(1.0, executions / 20)
- Learning curves computed from daily execution windows
### Agent Scoring Service (✅ Complete)
- Unified AgentScore combining SwarmCoordinator + learning profiles
- Scoring formula: 0.3*base + 0.5*expertise + 0.2*confidence
- Rank agents by combined score for intelligent assignment
- Support for recency-biased scoring (recent_success_rate)
- Methods: rank_agents, select_best, rank_agents_with_recency
### KG Integration (✅ Complete)
- KGPersistence::get_executions_for_task_type() - query by agent + task type
- KGPersistence::get_agent_executions() - all executions for agent
- Coordinator::load_learning_profile_from_kg() - core KG→Learning integration
- Coordinator::load_all_learning_profiles() - batch load for multiple agents
- Convert PersistedExecution → ExecutionData for learning calculations
### Agent Assignment Integration (✅ Complete)
- AgentCoordinator uses learning profiles for task assignment
- extract_task_type() infers task type from title/description
- assign_task() scores candidates using AgentScoringService
- Fallback to load-based selection if no learning data available
- Learning profiles stored in coordinator.learning_profiles RwLock
### Profile Adapter Enhancements (✅ Complete)
- create_learning_profile() - initialize empty profiles
- add_task_type_expertise() - set task-type expertise
- update_profile_with_learning() - update swarm profiles from learning
## Files Modified
### vapora-knowledge-graph/src/persistence.rs (+30 lines)
- get_executions_for_task_type(agent_id, task_type, limit)
- get_agent_executions(agent_id, limit)
### vapora-agents/src/coordinator.rs (+100 lines)
- load_learning_profile_from_kg() - core KG integration method
- load_all_learning_profiles() - batch loading for agents
- assign_task() already uses learning-based scoring via AgentScoringService
### Existing Complete Implementation
- vapora-knowledge-graph/src/learning.rs - calculation functions
- vapora-agents/src/learning_profile.rs - data structures and expertise
- vapora-agents/src/scoring.rs - unified scoring service
- vapora-agents/src/profile_adapter.rs - adapter methods
## Tests Passing
- learning_profile: 7 tests ✅
- scoring: 5 tests ✅
- profile_adapter: 6 tests ✅
- coordinator: learning-specific tests ✅
## Data Flow
1. Task arrives → AgentCoordinator::assign_task()
2. Extract task_type from description
3. Query KG for task-type executions (load_learning_profile_from_kg)
4. Calculate expertise with recency bias
5. Score candidates (SwarmCoordinator + learning)
6. Assign to top-scored agent
7. Execution result → KG → Update learning profiles
## Key Design Decisions
✅ Recency bias: 7-day half-life with 3x weight for recent performance
✅ Confidence scoring: min(1.0, total_executions / 20) prevents overfitting
✅ Hierarchical scoring: 30% base load, 50% expertise, 20% confidence
✅ KG query limit: 100 recent executions per task-type for performance
✅ Async loading: load_learning_profile_from_kg supports concurrent loads
## Next: Phase 5.4 - Cost Optimization
Ready to implement budget enforcement and cost-aware provider selection.
2026-01-11 13:03:53 +00:00
|
|
|
|
|
|
|
|
use std::sync::Arc;
|
|
|
|
|
use std::time::Duration;
|
2026-01-11 21:46:08 +00:00
|
|
|
|
feat: Phase 5.3 - Multi-Agent Learning Infrastructure
Implement intelligent agent learning from Knowledge Graph execution history
with per-task-type expertise tracking, recency bias, and learning curves.
## Phase 5.3 Implementation
### Learning Infrastructure (✅ Complete)
- LearningProfileService with per-task-type expertise metrics
- TaskTypeExpertise model tracking success_rate, confidence, learning curves
- Recency bias weighting: recent 7 days weighted 3x higher (exponential decay)
- Confidence scoring prevents overfitting: min(1.0, executions / 20)
- Learning curves computed from daily execution windows
### Agent Scoring Service (✅ Complete)
- Unified AgentScore combining SwarmCoordinator + learning profiles
- Scoring formula: 0.3*base + 0.5*expertise + 0.2*confidence
- Rank agents by combined score for intelligent assignment
- Support for recency-biased scoring (recent_success_rate)
- Methods: rank_agents, select_best, rank_agents_with_recency
### KG Integration (✅ Complete)
- KGPersistence::get_executions_for_task_type() - query by agent + task type
- KGPersistence::get_agent_executions() - all executions for agent
- Coordinator::load_learning_profile_from_kg() - core KG→Learning integration
- Coordinator::load_all_learning_profiles() - batch load for multiple agents
- Convert PersistedExecution → ExecutionData for learning calculations
### Agent Assignment Integration (✅ Complete)
- AgentCoordinator uses learning profiles for task assignment
- extract_task_type() infers task type from title/description
- assign_task() scores candidates using AgentScoringService
- Fallback to load-based selection if no learning data available
- Learning profiles stored in coordinator.learning_profiles RwLock
### Profile Adapter Enhancements (✅ Complete)
- create_learning_profile() - initialize empty profiles
- add_task_type_expertise() - set task-type expertise
- update_profile_with_learning() - update swarm profiles from learning
## Files Modified
### vapora-knowledge-graph/src/persistence.rs (+30 lines)
- get_executions_for_task_type(agent_id, task_type, limit)
- get_agent_executions(agent_id, limit)
### vapora-agents/src/coordinator.rs (+100 lines)
- load_learning_profile_from_kg() - core KG integration method
- load_all_learning_profiles() - batch loading for agents
- assign_task() already uses learning-based scoring via AgentScoringService
### Existing Complete Implementation
- vapora-knowledge-graph/src/learning.rs - calculation functions
- vapora-agents/src/learning_profile.rs - data structures and expertise
- vapora-agents/src/scoring.rs - unified scoring service
- vapora-agents/src/profile_adapter.rs - adapter methods
## Tests Passing
- learning_profile: 7 tests ✅
- scoring: 5 tests ✅
- profile_adapter: 6 tests ✅
- coordinator: learning-specific tests ✅
## Data Flow
1. Task arrives → AgentCoordinator::assign_task()
2. Extract task_type from description
3. Query KG for task-type executions (load_learning_profile_from_kg)
4. Calculate expertise with recency bias
5. Score candidates (SwarmCoordinator + learning)
6. Assign to top-scored agent
7. Execution result → KG → Update learning profiles
## Key Design Decisions
✅ Recency bias: 7-day half-life with 3x weight for recent performance
✅ Confidence scoring: min(1.0, total_executions / 20) prevents overfitting
✅ Hierarchical scoring: 30% base load, 50% expertise, 20% confidence
✅ KG query limit: 100 recent executions per task-type for performance
✅ Async loading: load_learning_profile_from_kg supports concurrent loads
## Next: Phase 5.4 - Cost Optimization
Ready to implement budget enforcement and cost-aware provider selection.
2026-01-11 13:03:53 +00:00
|
|
|
use vapora_agents::registry::AgentMetadata;
|
2026-01-11 21:32:56 +00:00
|
|
|
use vapora_agents::{AgentCoordinator, AgentRegistry, ProfileAdapter};
|
feat: Phase 5.3 - Multi-Agent Learning Infrastructure
Implement intelligent agent learning from Knowledge Graph execution history
with per-task-type expertise tracking, recency bias, and learning curves.
## Phase 5.3 Implementation
### Learning Infrastructure (✅ Complete)
- LearningProfileService with per-task-type expertise metrics
- TaskTypeExpertise model tracking success_rate, confidence, learning curves
- Recency bias weighting: recent 7 days weighted 3x higher (exponential decay)
- Confidence scoring prevents overfitting: min(1.0, executions / 20)
- Learning curves computed from daily execution windows
### Agent Scoring Service (✅ Complete)
- Unified AgentScore combining SwarmCoordinator + learning profiles
- Scoring formula: 0.3*base + 0.5*expertise + 0.2*confidence
- Rank agents by combined score for intelligent assignment
- Support for recency-biased scoring (recent_success_rate)
- Methods: rank_agents, select_best, rank_agents_with_recency
### KG Integration (✅ Complete)
- KGPersistence::get_executions_for_task_type() - query by agent + task type
- KGPersistence::get_agent_executions() - all executions for agent
- Coordinator::load_learning_profile_from_kg() - core KG→Learning integration
- Coordinator::load_all_learning_profiles() - batch load for multiple agents
- Convert PersistedExecution → ExecutionData for learning calculations
### Agent Assignment Integration (✅ Complete)
- AgentCoordinator uses learning profiles for task assignment
- extract_task_type() infers task type from title/description
- assign_task() scores candidates using AgentScoringService
- Fallback to load-based selection if no learning data available
- Learning profiles stored in coordinator.learning_profiles RwLock
### Profile Adapter Enhancements (✅ Complete)
- create_learning_profile() - initialize empty profiles
- add_task_type_expertise() - set task-type expertise
- update_profile_with_learning() - update swarm profiles from learning
## Files Modified
### vapora-knowledge-graph/src/persistence.rs (+30 lines)
- get_executions_for_task_type(agent_id, task_type, limit)
- get_agent_executions(agent_id, limit)
### vapora-agents/src/coordinator.rs (+100 lines)
- load_learning_profile_from_kg() - core KG integration method
- load_all_learning_profiles() - batch loading for agents
- assign_task() already uses learning-based scoring via AgentScoringService
### Existing Complete Implementation
- vapora-knowledge-graph/src/learning.rs - calculation functions
- vapora-agents/src/learning_profile.rs - data structures and expertise
- vapora-agents/src/scoring.rs - unified scoring service
- vapora-agents/src/profile_adapter.rs - adapter methods
## Tests Passing
- learning_profile: 7 tests ✅
- scoring: 5 tests ✅
- profile_adapter: 6 tests ✅
- coordinator: learning-specific tests ✅
## Data Flow
1. Task arrives → AgentCoordinator::assign_task()
2. Extract task_type from description
3. Query KG for task-type executions (load_learning_profile_from_kg)
4. Calculate expertise with recency bias
5. Score candidates (SwarmCoordinator + learning)
6. Assign to top-scored agent
7. Execution result → KG → Update learning profiles
## Key Design Decisions
✅ Recency bias: 7-day half-life with 3x weight for recent performance
✅ Confidence scoring: min(1.0, total_executions / 20) prevents overfitting
✅ Hierarchical scoring: 30% base load, 50% expertise, 20% confidence
✅ KG query limit: 100 recent executions per task-type for performance
✅ Async loading: load_learning_profile_from_kg supports concurrent loads
## Next: Phase 5.4 - Cost Optimization
Ready to implement budget enforcement and cost-aware provider selection.
2026-01-11 13:03:53 +00:00
|
|
|
|
|
|
|
|
/// Helper to create a test agent
|
|
|
|
|
fn create_test_agent(id: &str, role: &str) -> AgentMetadata {
|
|
|
|
|
AgentMetadata::new(
|
|
|
|
|
role.to_string(),
|
|
|
|
|
format!("Agent {}", id),
|
|
|
|
|
"claude".to_string(),
|
|
|
|
|
"claude-sonnet-4".to_string(),
|
|
|
|
|
vec!["coding".to_string(), "testing".to_string()],
|
|
|
|
|
)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_swarm_coordinator_integration_with_registry() {
|
|
|
|
|
// Setup: Create registry and coordinator
|
|
|
|
|
let registry = Arc::new(AgentRegistry::new(10));
|
|
|
|
|
|
|
|
|
|
// Register multiple agents
|
|
|
|
|
let agent1 = create_test_agent("1", "developer");
|
|
|
|
|
let agent2 = create_test_agent("2", "developer");
|
|
|
|
|
registry.register_agent(agent1).unwrap();
|
|
|
|
|
registry.register_agent(agent2).unwrap();
|
|
|
|
|
|
|
|
|
|
// Create coordinator (internally creates and initializes SwarmCoordinator)
|
|
|
|
|
let coordinator = AgentCoordinator::with_registry(Arc::clone(®istry));
|
|
|
|
|
|
|
|
|
|
// Assign a task - should use swarm coordinator
|
|
|
|
|
let result = coordinator
|
|
|
|
|
.assign_task(
|
|
|
|
|
"developer",
|
|
|
|
|
"Test task".to_string(),
|
|
|
|
|
"Implement a feature".to_string(),
|
|
|
|
|
"{}".to_string(),
|
|
|
|
|
80,
|
|
|
|
|
)
|
|
|
|
|
.await;
|
|
|
|
|
|
|
|
|
|
assert!(result.is_ok(), "Task assignment should succeed");
|
|
|
|
|
let assigned_agent_id = result.unwrap();
|
2026-01-11 21:32:56 +00:00
|
|
|
assert!(
|
|
|
|
|
!assigned_agent_id.is_empty(),
|
|
|
|
|
"Agent ID should not be empty"
|
|
|
|
|
);
|
feat: Phase 5.3 - Multi-Agent Learning Infrastructure
Implement intelligent agent learning from Knowledge Graph execution history
with per-task-type expertise tracking, recency bias, and learning curves.
## Phase 5.3 Implementation
### Learning Infrastructure (✅ Complete)
- LearningProfileService with per-task-type expertise metrics
- TaskTypeExpertise model tracking success_rate, confidence, learning curves
- Recency bias weighting: recent 7 days weighted 3x higher (exponential decay)
- Confidence scoring prevents overfitting: min(1.0, executions / 20)
- Learning curves computed from daily execution windows
### Agent Scoring Service (✅ Complete)
- Unified AgentScore combining SwarmCoordinator + learning profiles
- Scoring formula: 0.3*base + 0.5*expertise + 0.2*confidence
- Rank agents by combined score for intelligent assignment
- Support for recency-biased scoring (recent_success_rate)
- Methods: rank_agents, select_best, rank_agents_with_recency
### KG Integration (✅ Complete)
- KGPersistence::get_executions_for_task_type() - query by agent + task type
- KGPersistence::get_agent_executions() - all executions for agent
- Coordinator::load_learning_profile_from_kg() - core KG→Learning integration
- Coordinator::load_all_learning_profiles() - batch load for multiple agents
- Convert PersistedExecution → ExecutionData for learning calculations
### Agent Assignment Integration (✅ Complete)
- AgentCoordinator uses learning profiles for task assignment
- extract_task_type() infers task type from title/description
- assign_task() scores candidates using AgentScoringService
- Fallback to load-based selection if no learning data available
- Learning profiles stored in coordinator.learning_profiles RwLock
### Profile Adapter Enhancements (✅ Complete)
- create_learning_profile() - initialize empty profiles
- add_task_type_expertise() - set task-type expertise
- update_profile_with_learning() - update swarm profiles from learning
## Files Modified
### vapora-knowledge-graph/src/persistence.rs (+30 lines)
- get_executions_for_task_type(agent_id, task_type, limit)
- get_agent_executions(agent_id, limit)
### vapora-agents/src/coordinator.rs (+100 lines)
- load_learning_profile_from_kg() - core KG integration method
- load_all_learning_profiles() - batch loading for agents
- assign_task() already uses learning-based scoring via AgentScoringService
### Existing Complete Implementation
- vapora-knowledge-graph/src/learning.rs - calculation functions
- vapora-agents/src/learning_profile.rs - data structures and expertise
- vapora-agents/src/scoring.rs - unified scoring service
- vapora-agents/src/profile_adapter.rs - adapter methods
## Tests Passing
- learning_profile: 7 tests ✅
- scoring: 5 tests ✅
- profile_adapter: 6 tests ✅
- coordinator: learning-specific tests ✅
## Data Flow
1. Task arrives → AgentCoordinator::assign_task()
2. Extract task_type from description
3. Query KG for task-type executions (load_learning_profile_from_kg)
4. Calculate expertise with recency bias
5. Score candidates (SwarmCoordinator + learning)
6. Assign to top-scored agent
7. Execution result → KG → Update learning profiles
## Key Design Decisions
✅ Recency bias: 7-day half-life with 3x weight for recent performance
✅ Confidence scoring: min(1.0, total_executions / 20) prevents overfitting
✅ Hierarchical scoring: 30% base load, 50% expertise, 20% confidence
✅ KG query limit: 100 recent executions per task-type for performance
✅ Async loading: load_learning_profile_from_kg supports concurrent loads
## Next: Phase 5.4 - Cost Optimization
Ready to implement budget enforcement and cost-aware provider selection.
2026-01-11 13:03:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_profile_adapter_creates_valid_profiles() {
|
|
|
|
|
// Setup: Create agents and adapter
|
|
|
|
|
let agent1 = create_test_agent("1", "developer");
|
|
|
|
|
let agent2 = create_test_agent("2", "reviewer");
|
|
|
|
|
|
|
|
|
|
// Create profiles from agents
|
|
|
|
|
let profile1 = ProfileAdapter::create_profile(&agent1);
|
|
|
|
|
let profile2 = ProfileAdapter::create_profile(&agent2);
|
|
|
|
|
|
|
|
|
|
// Verify profile structure - ID is UUID mapped from agent.id
|
|
|
|
|
assert_eq!(profile1.id, agent1.id);
|
|
|
|
|
assert_eq!(profile2.id, agent2.id);
|
|
|
|
|
|
|
|
|
|
// Verify capabilities are mapped
|
|
|
|
|
assert!(!profile1.capabilities.is_empty());
|
|
|
|
|
assert!(!profile2.capabilities.is_empty());
|
|
|
|
|
|
|
|
|
|
// Verify default success rate is neutral
|
|
|
|
|
assert_eq!(profile1.success_rate, 0.5);
|
|
|
|
|
assert_eq!(profile2.success_rate, 0.5);
|
|
|
|
|
|
|
|
|
|
// Verify availability is based on agent status
|
|
|
|
|
assert!(profile1.availability);
|
|
|
|
|
assert!(profile2.availability);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_batch_profile_creation() {
|
|
|
|
|
// Setup: Create multiple agents
|
|
|
|
|
let agents = vec![
|
|
|
|
|
create_test_agent("1", "developer"),
|
|
|
|
|
create_test_agent("2", "reviewer"),
|
|
|
|
|
create_test_agent("3", "tester"),
|
|
|
|
|
];
|
|
|
|
|
|
|
|
|
|
// Batch create profiles
|
|
|
|
|
let profiles = ProfileAdapter::batch_create_profiles(agents);
|
|
|
|
|
|
|
|
|
|
// Verify all profiles created
|
|
|
|
|
assert_eq!(profiles.len(), 3);
|
|
|
|
|
|
|
|
|
|
// Verify each profile has correct properties
|
2026-02-14 20:10:55 +00:00
|
|
|
for profile in &profiles {
|
feat: Phase 5.3 - Multi-Agent Learning Infrastructure
Implement intelligent agent learning from Knowledge Graph execution history
with per-task-type expertise tracking, recency bias, and learning curves.
## Phase 5.3 Implementation
### Learning Infrastructure (✅ Complete)
- LearningProfileService with per-task-type expertise metrics
- TaskTypeExpertise model tracking success_rate, confidence, learning curves
- Recency bias weighting: recent 7 days weighted 3x higher (exponential decay)
- Confidence scoring prevents overfitting: min(1.0, executions / 20)
- Learning curves computed from daily execution windows
### Agent Scoring Service (✅ Complete)
- Unified AgentScore combining SwarmCoordinator + learning profiles
- Scoring formula: 0.3*base + 0.5*expertise + 0.2*confidence
- Rank agents by combined score for intelligent assignment
- Support for recency-biased scoring (recent_success_rate)
- Methods: rank_agents, select_best, rank_agents_with_recency
### KG Integration (✅ Complete)
- KGPersistence::get_executions_for_task_type() - query by agent + task type
- KGPersistence::get_agent_executions() - all executions for agent
- Coordinator::load_learning_profile_from_kg() - core KG→Learning integration
- Coordinator::load_all_learning_profiles() - batch load for multiple agents
- Convert PersistedExecution → ExecutionData for learning calculations
### Agent Assignment Integration (✅ Complete)
- AgentCoordinator uses learning profiles for task assignment
- extract_task_type() infers task type from title/description
- assign_task() scores candidates using AgentScoringService
- Fallback to load-based selection if no learning data available
- Learning profiles stored in coordinator.learning_profiles RwLock
### Profile Adapter Enhancements (✅ Complete)
- create_learning_profile() - initialize empty profiles
- add_task_type_expertise() - set task-type expertise
- update_profile_with_learning() - update swarm profiles from learning
## Files Modified
### vapora-knowledge-graph/src/persistence.rs (+30 lines)
- get_executions_for_task_type(agent_id, task_type, limit)
- get_agent_executions(agent_id, limit)
### vapora-agents/src/coordinator.rs (+100 lines)
- load_learning_profile_from_kg() - core KG integration method
- load_all_learning_profiles() - batch loading for agents
- assign_task() already uses learning-based scoring via AgentScoringService
### Existing Complete Implementation
- vapora-knowledge-graph/src/learning.rs - calculation functions
- vapora-agents/src/learning_profile.rs - data structures and expertise
- vapora-agents/src/scoring.rs - unified scoring service
- vapora-agents/src/profile_adapter.rs - adapter methods
## Tests Passing
- learning_profile: 7 tests ✅
- scoring: 5 tests ✅
- profile_adapter: 6 tests ✅
- coordinator: learning-specific tests ✅
## Data Flow
1. Task arrives → AgentCoordinator::assign_task()
2. Extract task_type from description
3. Query KG for task-type executions (load_learning_profile_from_kg)
4. Calculate expertise with recency bias
5. Score candidates (SwarmCoordinator + learning)
6. Assign to top-scored agent
7. Execution result → KG → Update learning profiles
## Key Design Decisions
✅ Recency bias: 7-day half-life with 3x weight for recent performance
✅ Confidence scoring: min(1.0, total_executions / 20) prevents overfitting
✅ Hierarchical scoring: 30% base load, 50% expertise, 20% confidence
✅ KG query limit: 100 recent executions per task-type for performance
✅ Async loading: load_learning_profile_from_kg supports concurrent loads
## Next: Phase 5.4 - Cost Optimization
Ready to implement budget enforcement and cost-aware provider selection.
2026-01-11 13:03:53 +00:00
|
|
|
assert!(!profile.id.is_empty());
|
|
|
|
|
assert!(!profile.capabilities.is_empty());
|
|
|
|
|
assert!(profile.success_rate >= 0.0 && profile.success_rate <= 1.0);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_task_assignment_selects_available_agent() {
|
|
|
|
|
// Setup: Create registry with agents
|
|
|
|
|
let registry = Arc::new(AgentRegistry::new(10));
|
|
|
|
|
|
|
|
|
|
let agent1 = create_test_agent("1", "developer");
|
|
|
|
|
let agent2 = create_test_agent("2", "developer");
|
|
|
|
|
|
|
|
|
|
registry.register_agent(agent1).unwrap();
|
|
|
|
|
registry.register_agent(agent2).unwrap();
|
|
|
|
|
|
|
|
|
|
let coordinator = AgentCoordinator::with_registry(Arc::clone(®istry));
|
|
|
|
|
|
|
|
|
|
// Assign multiple tasks
|
|
|
|
|
let result1 = coordinator
|
|
|
|
|
.assign_task(
|
|
|
|
|
"developer",
|
|
|
|
|
"Task 1".to_string(),
|
|
|
|
|
"Description".to_string(),
|
|
|
|
|
"{}".to_string(),
|
|
|
|
|
80,
|
|
|
|
|
)
|
|
|
|
|
.await;
|
|
|
|
|
|
|
|
|
|
let result2 = coordinator
|
|
|
|
|
.assign_task(
|
|
|
|
|
"developer",
|
|
|
|
|
"Task 2".to_string(),
|
|
|
|
|
"Description".to_string(),
|
|
|
|
|
"{}".to_string(),
|
|
|
|
|
80,
|
|
|
|
|
)
|
|
|
|
|
.await;
|
|
|
|
|
|
|
|
|
|
// Both should succeed
|
|
|
|
|
assert!(result1.is_ok());
|
|
|
|
|
assert!(result2.is_ok());
|
|
|
|
|
|
|
|
|
|
// Both should have assigned agents
|
|
|
|
|
assert!(!result1.unwrap().is_empty());
|
|
|
|
|
assert!(!result2.unwrap().is_empty());
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_coordinator_without_agent_fails() {
|
|
|
|
|
// Setup: Create registry and coordinator with no agents
|
|
|
|
|
let registry = Arc::new(AgentRegistry::new(10));
|
|
|
|
|
let coordinator = AgentCoordinator::with_registry(registry);
|
|
|
|
|
|
|
|
|
|
// Try to assign task with no available agents
|
|
|
|
|
let result = coordinator
|
|
|
|
|
.assign_task(
|
|
|
|
|
"nonexistent",
|
|
|
|
|
"Task".to_string(),
|
|
|
|
|
"Description".to_string(),
|
|
|
|
|
"{}".to_string(),
|
|
|
|
|
80,
|
|
|
|
|
)
|
|
|
|
|
.await;
|
|
|
|
|
|
|
|
|
|
// Should fail
|
|
|
|
|
assert!(result.is_err());
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_profile_sync_task_spawns() {
|
|
|
|
|
// Setup: Create registry and coordinator
|
|
|
|
|
let registry = Arc::new(AgentRegistry::new(10));
|
|
|
|
|
|
|
|
|
|
let agent = create_test_agent("1", "developer");
|
|
|
|
|
registry.register_agent(agent).unwrap();
|
|
|
|
|
|
|
|
|
|
// Create coordinator (spawns background profile sync task)
|
|
|
|
|
let _coordinator = AgentCoordinator::with_registry(Arc::clone(®istry));
|
|
|
|
|
|
|
|
|
|
// Give background task time to initialize
|
|
|
|
|
tokio::time::sleep(Duration::from_millis(100)).await;
|
|
|
|
|
|
|
|
|
|
// Registry should still have agents
|
|
|
|
|
let agents = registry.list_all();
|
|
|
|
|
assert_eq!(agents.len(), 1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_profile_load_calculation() {
|
|
|
|
|
// Setup: Create agent with known task count
|
|
|
|
|
let agent = create_test_agent("1", "developer");
|
|
|
|
|
|
|
|
|
|
let profile = ProfileAdapter::create_profile(&agent);
|
|
|
|
|
|
|
|
|
|
// Verify load is normalized (0.0-1.0)
|
|
|
|
|
assert!(profile.current_load >= 0.0 && profile.current_load <= 1.0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_multiple_role_assignment() {
|
|
|
|
|
// Setup: Create registry with agents of different roles
|
|
|
|
|
let registry = Arc::new(AgentRegistry::new(10));
|
|
|
|
|
|
|
|
|
|
let developer = create_test_agent("dev", "developer");
|
|
|
|
|
let reviewer = create_test_agent("rev", "reviewer");
|
|
|
|
|
|
|
|
|
|
registry.register_agent(developer).unwrap();
|
|
|
|
|
registry.register_agent(reviewer).unwrap();
|
|
|
|
|
|
|
|
|
|
let coordinator = AgentCoordinator::with_registry(Arc::clone(®istry));
|
|
|
|
|
|
|
|
|
|
// Assign task for developer
|
|
|
|
|
let dev_result = coordinator
|
|
|
|
|
.assign_task(
|
|
|
|
|
"developer",
|
|
|
|
|
"Code task".to_string(),
|
|
|
|
|
"Write code".to_string(),
|
|
|
|
|
"{}".to_string(),
|
|
|
|
|
80,
|
|
|
|
|
)
|
|
|
|
|
.await;
|
|
|
|
|
|
|
|
|
|
// Assign task for reviewer
|
|
|
|
|
let rev_result = coordinator
|
|
|
|
|
.assign_task(
|
|
|
|
|
"reviewer",
|
|
|
|
|
"Review task".to_string(),
|
|
|
|
|
"Review code".to_string(),
|
|
|
|
|
"{}".to_string(),
|
|
|
|
|
80,
|
|
|
|
|
)
|
|
|
|
|
.await;
|
|
|
|
|
|
|
|
|
|
// Both should succeed
|
|
|
|
|
assert!(dev_result.is_ok());
|
|
|
|
|
assert!(rev_result.is_ok());
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[tokio::test]
|
|
|
|
|
async fn test_swarm_statistics_available() {
|
|
|
|
|
// Setup: Create registry and coordinator with agents
|
|
|
|
|
let registry = Arc::new(AgentRegistry::new(10));
|
|
|
|
|
|
|
|
|
|
let agent1 = create_test_agent("1", "developer");
|
|
|
|
|
let agent2 = create_test_agent("2", "developer");
|
|
|
|
|
|
|
|
|
|
registry.register_agent(agent1).unwrap();
|
|
|
|
|
registry.register_agent(agent2).unwrap();
|
|
|
|
|
|
|
|
|
|
let coordinator = AgentCoordinator::with_registry(Arc::clone(®istry));
|
|
|
|
|
|
|
|
|
|
// Give swarm time to initialize
|
|
|
|
|
tokio::time::sleep(Duration::from_millis(100)).await;
|
|
|
|
|
|
|
|
|
|
// Coordinator should have functioning swarm coordinator
|
|
|
|
|
let result = coordinator
|
|
|
|
|
.assign_task(
|
|
|
|
|
"developer",
|
|
|
|
|
"Task".to_string(),
|
|
|
|
|
"Description".to_string(),
|
|
|
|
|
"{}".to_string(),
|
|
|
|
|
80,
|
|
|
|
|
)
|
|
|
|
|
.await;
|
|
|
|
|
|
|
|
|
|
// Should successfully assign task (verifies swarm is functional)
|
|
|
|
|
assert!(result.is_ok());
|
|
|
|
|
}
|