proagents 1.6.17 → 1.6.19

package/.proagents/ai-models/task-routing.md

@@ -1,503 +0,0 @@
-# Task Routing
-
-Route tasks to the most appropriate AI model based on complexity, type, and cost.
-
----
-
-## Routing Overview
-
-```
-┌─────────────────────────────────────────────────────────────┐
-│                    Task Routing Pipeline                     │
-├─────────────────────────────────────────────────────────────┤
-│                                                             │
-│  Incoming Task                                              │
-│       │                                                     │
-│       ▼                                                     │
-│  ┌─────────────┐                                           │
-│  │  Analyzer   │ ── Complexity, Type, Context              │
-│  └─────────────┘                                           │
-│       │                                                     │
-│       ▼                                                     │
-│  ┌─────────────┐      ┌─────────────────────────────────┐  │
-│  │   Router    │ ───▶ │ Simple → Haiku/GPT-3.5          │  │
-│  └─────────────┘      │ Medium → Sonnet/GPT-4           │  │
-│       │               │ Complex → Opus/GPT-4-Turbo      │  │
-│       │               └─────────────────────────────────┘  │
-│       ▼                                                     │
-│  ┌─────────────┐                                           │
-│  │  Selected   │                                           │
-│  │   Model     │                                           │
-│  └─────────────┘                                           │
-│                                                             │
-└─────────────────────────────────────────────────────────────┘
-```
-
----
-
-## Routing Configuration
-
-### Basic Setup
-
-```yaml
-ai:
-  routing:
-    enabled: true
-
-    # Default model
-    default: "claude-3-sonnet"
-
-    # Routing strategy
-    strategy: "smart"  # smart, cost_optimized, quality_first
-```
-
-### Task Type Routing
-
-```yaml
-ai:
-  routing:
-    by_task_type:
-      # Code analysis
-      code_analysis:
-        default: "claude-3-opus"
-        reason: "Best for deep code understanding"
-
-      # Code generation
-      code_generation:
-        simple: "claude-3-haiku"
-        moderate: "claude-3-sonnet"
-        complex: "claude-3-opus"
-
-      # Code review
-      code_review:
-        default: "claude-3-sonnet"
-        security_review: "claude-3-opus"
-
-      # Documentation
-      documentation:
-        default: "claude-3-sonnet"
-        api_docs: "claude-3-sonnet"
-        quick_comments: "claude-3-haiku"
-
-      # Bug fixing
-      bug_fix:
-        simple: "claude-3-haiku"
-        complex: "claude-3-sonnet"
-        critical: "claude-3-opus"
-
-      # Testing
-      test_generation:
-        unit_tests: "claude-3-sonnet"
-        integration_tests: "claude-3-opus"
-
-      # Refactoring
-      refactoring:
-        simple: "claude-3-sonnet"
-        architectural: "claude-3-opus"
-```
-
----
-
-## Complexity Detection
-
-### Complexity Indicators
-
-```yaml
-ai:
-  routing:
-    complexity:
-      # Low complexity indicators
-      low:
-        - "single_file_change"
-        - "documentation_only"
-        - "config_change"
-        - "simple_fix"
-        - "formatting"
-        - "less_than_20_lines"
-
-      # Medium complexity indicators
-      medium:
-        - "multiple_files"
-        - "new_component"
-        - "api_integration"
-        - "moderate_logic"
-        - "20_to_100_lines"
-
-      # High complexity indicators
-      high:
-        - "architectural_change"
-        - "cross_module"
-        - "security_related"
-        - "performance_critical"
-        - "more_than_100_lines"
-        - "complex_algorithms"
-        - "state_management"
-```
-
-### Context-Based Routing
-
-```yaml
-ai:
-  routing:
-    by_context:
-      # File patterns
-      file_patterns:
-        "src/auth/**":
-          model: "claude-3-opus"
-          reason: "Security-sensitive code"
-
-        "src/payments/**":
-          model: "claude-3-opus"
-          reason: "Financial code requires high accuracy"
-
-        "*.test.*":
-          model: "claude-3-sonnet"
-          reason: "Test files are moderately complex"
-
-        "docs/**":
-          model: "claude-3-haiku"
-          reason: "Documentation is straightforward"
-
-      # Project phase
-      by_phase:
-        analysis:
-          model: "claude-3-opus"
-          reason: "Deep analysis needs best model"
-
-        implementation:
-          model: "claude-3-sonnet"
-          reason: "Balance of speed and quality"
-
-        documentation:
-          model: "claude-3-haiku"
-          reason: "Fast for documentation tasks"
-```
-
----
-
-## Cost Optimization
-
-### Cost-Based Routing
-
-```yaml
-ai:
-  routing:
-    cost_optimization:
-      enabled: true
-
-      # Daily budget
-      daily_budget: 50.00
-
-      # Cost thresholds
-      thresholds:
-        # Use cheaper model when budget is low
-        budget_low:
-          percentage: 80  # 80% of budget used
-          downgrade_to: "claude-3-haiku"
-          except:
-            - "security_review"
-            - "critical_bugs"
-
-        budget_critical:
-          percentage: 95
-          downgrade_to: "claude-3-haiku"
-          except:
-            - "blocking_tasks"
-
-      # Cost per task type
-      max_cost_per_task:
-        simple_fix: 0.10
-        code_review: 0.50
-        feature_implementation: 2.00
-        full_analysis: 5.00
-```
-
-### Model Cost Tracking
-
-```
-┌─────────────────────────────────────────────────────────────┐
-│ AI Cost Dashboard - Today                                   │
-├─────────────────────────────────────────────────────────────┤
-│                                                             │
-│ Budget: $50.00 | Used: $32.50 (65%) | Remaining: $17.50    │
-│ [████████████████████░░░░░░░░░░] 65%                       │
-│                                                             │
-│ By Model:                                                   │
-│ ├── Opus:   $18.00 (12 requests)                           │
-│ ├── Sonnet: $12.50 (45 requests)                           │
-│ └── Haiku:  $2.00 (120 requests)                           │
-│                                                             │
-│ By Task Type:                                               │
-│ ├── Code Analysis:  $15.00                                 │
-│ ├── Code Review:    $8.00                                  │
-│ ├── Implementation: $6.50                                  │
-│ └── Documentation:  $3.00                                  │
-│                                                             │
-│ Routing Decisions:                                          │
-│ ├── Upgraded to Opus: 5 (complexity)                       │
-│ ├── Downgraded to Haiku: 8 (cost)                         │
-│ └── Default Sonnet: 32                                     │
-│                                                             │
-└─────────────────────────────────────────────────────────────┘
-```
-
----
-
-## Quality-First Routing
-
-### Quality Priority
-
-```yaml
-ai:
-  routing:
-    quality_first:
-      enabled: true
-
-      # Always use best model for these
-      always_use_best:
-        - task_type: "security_review"
-        - task_type: "architecture_decisions"
-        - file_pattern: "src/core/**"
-        - tag: "critical"
-
-      # Quality metrics to consider
-      metrics:
-        # If previous response quality was low
-        retry_with_better:
-          enabled: true
-          threshold: 0.7  # Quality score
-          upgrade_to: "next_tier"
-
-        # Track model performance
-        track_accuracy: true
-        adjust_routing: true
-```
-
-### Performance Tracking
-
-```yaml
-ai:
-  routing:
-    performance:
-      # Track response quality
-      track_quality: true
-
-      # Adjust routing based on results
-      adaptive_routing:
-        enabled: true
-
-        # If model consistently underperforms
-        downgrade_threshold: 0.6
-        upgrade_threshold: 0.9
-
-        # Learning period
-        evaluation_window: 100  # requests
-```
-
----
-
-## Fallback Chains
-
-### Fallback Configuration
-
-```yaml
-ai:
-  routing:
-    fallbacks:
-      # Primary fallback chain
-      chain:
-        - "claude-3-opus"
-        - "claude-3-sonnet"
-        - "gpt-4-turbo"
-        - "claude-3-haiku"
-        - "gpt-3.5-turbo"
-
-      # When to fallback
-      triggers:
-        - "rate_limit"
-        - "model_unavailable"
-        - "timeout"
-        - "error"
-
-      # Retry settings
-      retry:
-        max_attempts: 3
-        delay: "2s"
-        backoff: "exponential"
-
-      # Alert on fallback
-      notify:
-        on_fallback: true
-        on_last_resort: true
-```
-
-### Fallback Flow
-
-```
-┌─────────────────────────────────────────────────────────────┐
-│                    Fallback Flow                             │
-├─────────────────────────────────────────────────────────────┤
-│                                                             │
-│  Request                                                    │
-│     │                                                       │
-│     ▼                                                       │
-│  Claude Opus ──[Rate Limited]──▶ Claude Sonnet             │
-│                                        │                    │
-│                                        ▼                    │
-│                               [Timeout]──▶ GPT-4            │
-│                                              │              │
-│                                              ▼              │
-│                                      [Success] ✅           │
-│                                                             │
-│  Log: "Fallback: Opus → Sonnet → GPT-4 (success)"          │
-│                                                             │
-└─────────────────────────────────────────────────────────────┘
-```
-
----
-
-## Custom Routing Rules
-
-### Rule Definition
-
-```yaml
-ai:
-  routing:
-    custom_rules:
-      # Rule 1: Security-critical
-      - name: "security_critical"
-        priority: 1
-        conditions:
-          any:
-            - file_pattern: "src/auth/**"
-            - file_pattern: "src/security/**"
-            - contains: ["password", "token", "secret", "credential"]
-        route_to: "claude-3-opus"
-        reason: "Security-sensitive code"
-
-      # Rule 2: Quick fixes
-      - name: "quick_fixes"
-        priority: 2
-        conditions:
-          all:
-            - task_type: "bug_fix"
-            - files_changed: "<= 2"
-            - lines_changed: "<= 20"
-        route_to: "claude-3-haiku"
-        reason: "Simple fix, fast model"
-
-      # Rule 3: New features
-      - name: "new_features"
-        priority: 3
-        conditions:
-          all:
-            - task_type: "feature"
-            - complexity: ">= medium"
-        route_to: "claude-3-sonnet"
-        reason: "Balance of quality and speed"
-```
-
----
-
-## Routing API
-
-### Programmatic Routing
-
-```typescript
-// Request with routing hints
-const response = await proagents.ai.request({
-  task: "Review this authentication code",
-  context: {
-    files: ["src/auth/login.ts"],
-    type: "security_review"
-  },
-  routing: {
-    prefer: "quality",      // quality, speed, cost
-    minModel: "sonnet",     // Minimum model tier
-    maxCost: 1.00,          // Maximum cost
-    timeout: 30000          // Timeout in ms
-  }
-});
-```
-
-### Routing Override
-
-```bash
-# Force specific model
-proagents review --model opus
-
-# Cost-optimized routing
-proagents analyze --routing cost
-
-# Quality-first routing
-proagents review --routing quality
-
-# View routing decision
-proagents debug-routing "Review auth code"
-```
-
----
-
-## Routing Logs
-
-### Decision Logging
-
-```yaml
-ai:
-  routing:
-    logging:
-      enabled: true
-
-      # What to log
-      log:
-        - decision
-        - reason
-        - cost
-        - latency
-        - fallbacks
-
-      # Storage
-      storage:
-        path: ".proagents/routing-logs/"
-        retention: "30d"
-```
-
-### Log Example
-
-```json
-{
-  "timestamp": "2024-01-15T10:30:00Z",
-  "task_id": "task-123",
-  "task_type": "code_review",
-  "context": {
-    "files": ["src/auth/login.ts"],
-    "complexity": "high",
-    "lines": 150
-  },
-  "routing_decision": {
-    "selected_model": "claude-3-opus",
-    "reason": "Security-sensitive file + high complexity",
-    "rules_matched": ["security_critical", "high_complexity"],
-    "cost_estimate": 0.45
-  },
-  "actual": {
-    "model_used": "claude-3-opus",
-    "latency_ms": 2500,
-    "tokens_in": 1200,
-    "tokens_out": 800,
-    "actual_cost": 0.42,
-    "quality_score": 0.95
-  }
-}
-```
-
----
-
-## Best Practices
-
-1. **Start Conservative**: Use higher-tier models initially, optimize later
-2. **Monitor Costs**: Track spending per task type
-3. **Quality Metrics**: Measure and track response quality
-4. **Context Matters**: Consider file sensitivity in routing
-5. **Fallbacks Ready**: Always have fallback chain configured
-6. **Review Logs**: Periodically review routing decisions

package/.proagents/ai-training/README.md

@@ -1,155 +0,0 @@
-# AI Training & Project Learning
-
-Train AI on project-specific patterns, preferences, and domain knowledge.
-
----
-
-## Overview
-
-ProAgents learns from your project to provide increasingly accurate and relevant suggestions over time.
-
-```
-┌─────────────────────────────────────────────────────────────┐
-│                    Learning Architecture                     │
-├─────────────────────────────────────────────────────────────┤
-│                                                             │
-│  Data Sources                                               │
-│  ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐          │
-│  │Codebase │ │  PRs &  │ │ User    │ │ Domain  │          │
-│  │Patterns │ │ Reviews │ │Feedback │ │  Docs   │          │
-│  └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘          │
-│       │           │           │           │                │
-│       └───────────┴─────┬─────┴───────────┘                │
-│                         │                                   │
-│                         ▼                                   │
-│              ┌─────────────────────┐                       │
-│              │   Learning Engine   │                       │
-│              └──────────┬──────────┘                       │
-│                         │                                   │
-│       ┌─────────────────┼─────────────────┐                │
-│       ▼                 ▼                 ▼                │
-│  ┌─────────┐     ┌─────────┐     ┌─────────┐              │
-│  │ Pattern │     │ Domain  │     │  User   │              │
-│  │ Model   │     │ Model   │     │ Prefs   │              │
-│  └─────────┘     └─────────┘     └─────────┘              │
-│                                                             │
-│  Applied To:                                                │
-│  • Code suggestions                                        │
-│  • Architecture decisions                                  │
-│  • Naming conventions                                      │
-│  • Review comments                                         │
-│                                                             │
-└─────────────────────────────────────────────────────────────┘
-```
-
----
-
-## Quick Start
-
-### Enable Learning
-
-```yaml
-# proagents.config.yaml
-
-learning:
-  enabled: true
-  auto_learn: true
-```
-
-### View What's Learned
-
-```bash
-# Show learning summary
-proagents learning status
-
-# Show specific patterns
-proagents learning patterns
-
-# Show user preferences
-proagents learning preferences
-```
-
----
-
-## Learning Types
-
-| Type | Description | Example |
-|------|-------------|---------|
-| **Code Patterns** | Common code structures | "Always use async/await over .then()" |
-| **Naming Conventions** | Variable/function naming | "Components use PascalCase" |
-| **Architecture** | Design patterns used | "Services in /src/services/" |
-| **Domain Knowledge** | Business terminology | "User = registered customer" |
-| **User Preferences** | Individual preferences | "Prefers detailed explanations" |
-| **Team Standards** | Team conventions | "All PRs need 2 approvers" |
-
----
-
-## Documentation Files
-
-| File | Description |
-|------|-------------|
-| [pattern-learning.md](./pattern-learning.md) | Code pattern recognition |
-| [domain-knowledge.md](./domain-knowledge.md) | Business domain learning |
-| [user-preferences.md](./user-preferences.md) | User preference tracking |
-| [continuous-learning.md](./continuous-learning.md) | Ongoing improvement |
-| [training-data.md](./training-data.md) | Training data management |
-
----
-
-## Configuration
-
-```yaml
-# proagents.config.yaml
-
-learning:
-  enabled: true
-
-  # What to learn from
-  sources:
-    codebase: true
-    pull_requests: true
-    code_reviews: true
-    user_feedback: true
-    documentation: true
-
-  # What to learn
-  learn:
-    code_patterns: true
-    naming_conventions: true
-    architecture_patterns: true
-    testing_patterns: true
-    domain_knowledge: true
-    user_preferences: true
-
-  # Auto-apply learnings
-  auto_apply:
-    code_suggestions: true
-    review_comments: true
-    documentation: true
-
-  # Storage
-  storage:
-    path: ".proagents/learning"
-    sync_to_cloud: false
-
-  # Privacy
-  privacy:
-    anonymize: true
-    exclude_secrets: true
-    exclude_patterns:
-      - "*.env"
-      - "**/secrets/**"
-```
-
----
-
-## Commands
-
-| Command | Description |
-|---------|-------------|
-| `proagents learning status` | Show learning status |
-| `proagents learning patterns` | Show learned patterns |
-| `proagents learning train` | Trigger manual training |
-| `proagents learning reset` | Reset learning data |
-| `proagents learning export` | Export learning data |
-| `proagents learning import` | Import learning data |