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.
8.8 KiB
8.8 KiB
📚 doc-lifecycle-manager Integration
Dual-Mode: Agent Plugin + Standalone System
Version: 0.1.0 Status: Specification (VAPORA v1.0 Integration) Purpose: Integration of doc-lifecycle-manager as both VAPORA component AND standalone tool
🎯 Objetivo
doc-lifecycle-manager funciona de dos formas:
- Como agente VAPORA: Documenter role usa doc-lifecycle internally
- Como sistema standalone: Proyectos sin VAPORA usan doc-lifecycle solo
Permite adopción gradual: empezar con doc-lifecycle solo, migrar a VAPORA después.
🔄 Dual-Mode Architecture
Mode 1: Standalone (Sin VAPORA)
proyecto-simple/
├── docs/
│ ├── architecture/
│ ├── guides/
│ └── adr/
├── .doc-lifecycle-manager/
│ ├── config.toml
│ ├── templates/
│ └── metadata/
└── .github/workflows/
└── docs-update.yaml # Triggered on push
Usage:
# Manual
doc-lifecycle-manager classify docs/
doc-lifecycle-manager consolidate docs/
doc-lifecycle-manager index --for-rag
# Via CI/CD
.github/workflows/docs-update.yaml:
on: [push]
steps:
- run: doc-lifecycle-manager sync
Capabilities:
- Classify docs by type
- Consolidate duplicates
- Manage lifecycle (draft → published → archived)
- Generate RAG index
- Build presentations (mdBook, Slidev)
Mode 2: As VAPORA Agent (With VAPORA)
proyecto-vapora/
├── .vapora/
│ ├── agents/
│ │ └── documenter/
│ │ ├── config.toml
│ │ └── plugins/
│ │ └── doc-lifecycle-manager/ # Embedded
│ └── ...
├── docs/
└── .coder/
Architecture:
Documenter Agent (Role)
│
├─ Root Files Keeper
│ ├─ README.md
│ ├─ CHANGELOG.md
│ ├─ ROADMAP.md
│ └─ (auto-generated)
│
└─ doc-lifecycle-manager Plugin
├─ Classify documents
├─ Consolidate duplicates
├─ Manage ADRs (from sessions)
├─ Generate presentations
└─ Build RAG index
Workflow:
Task completed
↓
Orchestrator publishes: "task_completed" event
↓
Documenter Agent subscribes to: vapora.tasks.completed
↓
Documenter loads config:
├─ Root Files Keeper (built-in)
└─ doc-lifecycle-manager plugin
↓
Executes (in order):
1. Extract decisions from sessions → doc-lifecycle ADR classification
2. Update root files (README, CHANGELOG, ROADMAP)
3. Classify all docs in docs/
4. Consolidate duplicates
5. Generate RAG index
6. (Optional) Build mdBook + Slidev presentations
↓
Publishes: "docs_updated" event
🔌 Plugin Interface
Documenter Agent Loads doc-lifecycle-manager
pub struct DocumenterAgent {
pub root_files_keeper: RootFilesKeeper,
pub doc_lifecycle: DocLifecycleManager, // Plugin
}
impl DocumenterAgent {
pub async fn execute_task(
&mut self,
task: Task,
) -> anyhow::Result<()> {
// 1. Update root files (always)
self.root_files_keeper.sync_all(&task).await?;
// 2. Use doc-lifecycle for deep doc management (if configured)
if self.config.enable_doc_lifecycle {
self.doc_lifecycle.classify_docs("docs/").await?;
self.doc_lifecycle.consolidate_duplicates().await?;
self.doc_lifecycle.manage_lifecycle().await?;
// 3. Build presentations
if self.config.generate_presentations {
self.doc_lifecycle.generate_mdbook().await?;
self.doc_lifecycle.generate_slidev().await?;
}
// 4. Build RAG index (for search)
self.doc_lifecycle.build_rag_index().await?;
}
Ok(())
}
}
🚀 Migration: Standalone → VAPORA
Step 1: Run Standalone
proyecto/
├── docs/
│ ├── architecture/
│ └── adr/
├── .doc-lifecycle-manager/
│ └── config.toml
└── .github/workflows/docs-update.yaml
# Usage: Manual or via CI/CD
doc-lifecycle-manager sync
Step 2: Install VAPORA
# Initialize VAPORA
vapora init
# VAPORA auto-detects existing .doc-lifecycle-manager/
# and integrates it into Documenter agent
Step 3: Migrate Workflows
# Before (in CI/CD):
- run: doc-lifecycle-manager sync
# After (in VAPORA):
# - Documenter agent runs automatically post-task
# - CLI still available:
vapora doc-lifecycle classify
vapora doc-lifecycle consolidate
vapora doc-lifecycle rag-index
📋 Configuration
Standalone Config
# .doc-lifecycle-manager/config.toml
[lifecycle]
doc_root = "docs/"
adr_path = "docs/adr/"
archive_days = 180
[classification]
enabled = true
auto_consolidate_duplicates = true
detect_orphaned_docs = true
[rag]
enabled = true
chunk_size = 500
overlap = 50
index_path = ".doc-lifecycle-manager/index.json"
[presentations]
generate_mdbook = true
generate_slidev = true
mdbook_out = "book/"
slidev_out = "slides/"
[lifecycle_rules]
[[rule]]
path_pattern = "docs/guides/*"
lifecycle = "guide"
retention_days = 0 # Never delete
[[rule]]
path_pattern = "docs/experimental/*"
lifecycle = "experimental"
retention_days = 30
VAPORA Integration Config
# .vapora/.vapora.toml
[documenter]
# Embedded doc-lifecycle config
doc_lifecycle_enabled = true
doc_lifecycle_config = ".doc-lifecycle-manager/config.toml" # Reuse
[root_files]
auto_update = true
generate_changelog_from_git = true
generate_roadmap_from_tasks = true
🎯 Commands (Both Modes)
Standalone Mode
# Classify documents
doc-lifecycle-manager classify docs/
# Consolidate duplicates
doc-lifecycle-manager consolidate
# Manage lifecycle
doc-lifecycle-manager lifecycle prune --older-than 180d
# Build RAG index
doc-lifecycle-manager rag-index --output index.json
# Generate presentations
doc-lifecycle-manager mdbook build
doc-lifecycle-manager slidev build
VAPORA Integration
# Via documenter agent (automatic post-task)
# Or manual:
vapora doc-lifecycle classify
vapora doc-lifecycle consolidate
vapora doc-lifecycle rag-index
# Root files (via Documenter)
vapora root-files sync
# Full documentation update
vapora document sync --all
📊 Lifecycle States (doc-lifecycle)
Draft
├─ In-progress documentation
├─ Not indexed
└─ Not published
Published
├─ Ready for users
├─ Indexed for RAG
├─ Included in presentations
└─ Linked in README
Updated
├─ Recently modified
├─ Re-indexed for RAG
└─ Change log entry created
Archived
├─ Outdated
├─ Removed from presentations
├─ Indexed but marked deprecated
└─ Can be recovered
🔐 RAG Integration
doc-lifecycle → RAG Index
{
"doc_id": "ADR-015-batch-workflow",
"title": "ADR-015: Batch Workflow System",
"doc_type": "adr",
"lifecycle_state": "published",
"created_date": "2025-11-09",
"last_updated": "2025-11-10",
"vector_embedding": [0.1, 0.2, ...], // 1536-dim
"content_preview": "Decision: Use Rust for batch orchestrator...",
"tags": ["orchestrator", "workflow", "architecture"],
"source_session": "sess-2025-11-09-143022",
"related_adr": ["ADR-010", "ADR-014"],
"search_keywords": ["batch", "workflow", "orchestrator"]
}
RAG Search (Via VAPORA Agent Search)
# Search documentation
vapora search "batch workflow architecture"
# Results from doc-lifecycle RAG index:
# 1. ADR-015-batch-workflow.md (0.94 relevance)
# 2. batch-workflow-guide.md (0.87)
# 3. orchestrator-design.md (0.71)
🎯 Implementation Checklist
Standalone Components
- Document classifier (by type, domain, lifecycle)
- Duplicate detector & consolidator
- Lifecycle state management (Draft→Published→Archived)
- RAG index builder (chunking, embeddings)
- mdBook generator
- Slidev generator
- CLI interface
VAPORA Integration
- Documenter agent loads doc-lifecycle-manager
- Plugin interface (DocLifecycleManager trait)
- Event subscriptions (vapora.tasks.completed)
- Config reuse (.doc-lifecycle-manager/ detected)
- Seamless startup (no additional config)
Migration Tools
- Detect existing .doc-lifecycle-manager/
- Auto-configure Documenter agent
- Preserve existing RAG indexes
- No data loss during migration
📊 Success Metrics
✅ Standalone doc-lifecycle works independently ✅ VAPORA auto-detects and loads doc-lifecycle ✅ Documenter agent uses both Root Files + doc-lifecycle ✅ Migration takes < 5 minutes ✅ No duplicate work (each tool owns its domain) ✅ RAG indexing automatic and current
Version: 0.1.0 Status: ✅ Integration Specification Complete Purpose: Seamless doc-lifecycle-manager dual-mode integration with VAPORA