alita-sdk 0.3.379__py3-none-any.whl → 0.3.627__py3-none-any.whl

alita_sdk/community/inventory/__init__.py

@@ -0,0 +1,236 @@
+"""
+Inventory Module for Knowledge Graph Construction and Retrieval.
+
+This module provides two distinct capabilities:
+
+1. **Ingestion Pipeline** - A workflow for building/updating knowledge graphs
+   from source code repositories. NOT a toolkit - it's a defined process.
+   
+   Usage:
+       from alita_sdk.community.inventory import (
+           IngestionPipeline, 
+           ingest_repository,
+           PYTHON_PRESET,
+           TYPESCRIPT_PRESET,
+           get_preset
+       )
+       
+       # Full pipeline with config
+       pipeline = IngestionPipeline(
+           llm=llm,
+           graph_path="./graph.json",
+           source_toolkits={'github': github_toolkit}
+       )
+       result = pipeline.run(source='github', branch='main')
+       
+       # Or one-shot convenience function
+       result = ingest_repository(
+           llm=llm,
+           graph_path="./graph.json",
+           source_toolkit=github_toolkit,
+           source_name="github"
+       )
+
+2. **Retrieval Toolkit** - A pure query toolkit for retrieving context from 
+   a pre-built knowledge graph. Can be added to any agent.
+   
+   Usage:
+       from alita_sdk.community.inventory import InventoryRetrievalToolkit
+       
+       # As a toolkit for agents
+       toolkit = InventoryRetrievalToolkit.get_toolkit(
+           graph_path="./graph.json",
+           base_directory="/path/to/source"  # For local content retrieval
+       )
+       tools = toolkit.get_tools()
+
+Entity Taxonomy (8 layers, 49 types):
+- Product Layer: feature, product, user_story, requirement, epic
+- Domain Layer: domain, subdomain, business_capability, value_stream, process
+- Service Layer: service, microservice, api, api_endpoint, message_queue, event
+- Code Layer: module, package, class, function, method, interface, trait, enum, type, variable, constant
+- Data Layer: database, table, collection, schema, model, entity, field, index, query, migration
+- Testing Layer: test_suite, test_case, test_fixture, mock, stub, assertion
+- Delivery Layer: pipeline, job, stage, environment, deployment, artifact, container
+- Organization Layer: team, repository, project, workspace, organization
+
+Relationship Taxonomy (8 categories, 34 types):
+- Structural: CONTAINS, IMPORTS, EXTENDS, IMPLEMENTS, USES, DEPENDS_ON, INSTANTIATES, COMPOSED_OF
+- Behavioral: CALLS, INVOKES, TRIGGERS, HANDLES, SUBSCRIBES_TO, PUBLISHES_TO, RETURNS
+- Data Lineage: READS_FROM, WRITES_TO, TRANSFORMS, QUERIES, STORES_IN, REFERENCES
+- UI/Product: RENDERS, ROUTES_TO, NAVIGATES_TO, DISPLAYS
+- Testing: TESTS, MOCKS, COVERS, ASSERTS
+- Ownership: OWNED_BY, MAINTAINED_BY, CREATED_BY
+- Temporal: PRECEDES, FOLLOWS, SCHEDULED_BY
+- Semantic: RELATED_TO, SIMILAR_TO, ALIAS_OF
+"""
+
+import logging
+from typing import List, Optional, Dict, Any
+
+# Configuration
+from .config import (
+    IngestionConfig,
+    GuardrailsConfig,
+    generate_config_template,
+)
+
+# Ingestion Pipeline - workflow for graph building
+from .ingestion import (
+    IngestionPipeline,
+    IngestionResult,
+    ingest_repository,
+)
+
+# Retrieval Toolkit - for querying graphs
+from .retrieval import InventoryRetrievalApiWrapper
+
+# Toolkit utilities - for configuration and instantiation
+from .toolkit_utils import (
+    load_toolkit_config,
+    get_llm_for_config,
+    get_source_toolkit,
+)
+
+# Core graph types
+from .knowledge_graph import KnowledgeGraph, Citation
+
+# Extractors (for advanced use)
+from .extractors import (
+    ENTITY_TAXONOMY,
+    RELATIONSHIP_TAXONOMY,
+    EntityExtractor,
+    RelationExtractor,
+    FactExtractor,
+    DocumentClassifier,
+    EntitySchemaDiscoverer,
+)
+
+# Toolkit wrapper for agent integration
+from .toolkit import InventoryRetrievalToolkit
+
+# Ingestion presets
+from .presets import (
+    PYTHON_PRESET,
+    PYTHON_PRESET_WITH_TESTS,
+    JAVASCRIPT_PRESET,
+    TYPESCRIPT_PRESET,
+    REACT_PRESET,
+    NEXTJS_PRESET,
+    JAVA_PRESET,
+    SPRING_BOOT_PRESET,
+    MAVEN_PRESET,
+    GRADLE_PRESET,
+    DOTNET_PRESET,
+    CSHARP_PRESET,
+    ASPNET_PRESET,
+    FULLSTACK_JS_PRESET,
+    MONOREPO_PRESET,
+    DOCUMENTATION_PRESET,
+    PRESETS,
+    get_preset,
+    list_presets,
+    combine_presets,
+)
+
+logger = logging.getLogger(__name__)
+
+name = "inventory"
+
+
+def get_tools(tool: dict, tools_list: Optional[List[dict]] = None):
+    """
+    Get inventory retrieval tools for agent integration.
+    
+    This function is called by the toolkit loader to get the
+    retrieval tools for querying a pre-built knowledge graph.
+    
+    NOTE: For ingestion, use the IngestionPipeline directly, not through
+    the agent toolkit system. Ingestion is a workflow, not an agent task.
+    
+    Args:
+        tool: The inventory toolkit configuration dict
+        tools_list: Optional list of all toolkit configs in the agent
+    
+    Returns:
+        List of BaseTool instances for knowledge graph retrieval
+    """
+    settings = tool.get('settings', {})
+    
+    # For retrieval, we need the graph path
+    graph_path = settings.get('graph_path')
+    if not graph_path:
+        logger.warning("Inventory toolkit requires graph_path setting for retrieval")
+    
+    toolkit = InventoryRetrievalToolkit.get_toolkit(
+        selected_tools=settings.get('selected_tools', []),
+        toolkit_name=tool.get('toolkit_name'),
+        # Graph location
+        graph_path=graph_path,
+        # For local content retrieval
+        base_directory=settings.get('base_directory'),
+        # Source toolkits for remote content retrieval (optional)
+        source_toolkits=settings.get('source_toolkits', {}),
+    )
+    return toolkit.get_tools()
+
+
+__all__ = [
+    # Module name
+    'name',
+    'get_tools',
+    
+    # Configuration
+    'IngestionConfig',
+    'GuardrailsConfig',
+    'generate_config_template',
+    
+    # Ingestion (workflow)
+    'IngestionPipeline',
+    'IngestionResult',
+    'ingest_repository',
+    
+    # Retrieval (toolkit)
+    'InventoryRetrievalToolkit',
+    'InventoryRetrievalApiWrapper',
+    
+    # Toolkit utilities
+    'load_toolkit_config',
+    'get_llm_for_config',
+    'get_source_toolkit',
+    
+    # Core types
+    'KnowledgeGraph',
+    'Citation',
+    
+    # Extractors
+    'ENTITY_TAXONOMY',
+    'RELATIONSHIP_TAXONOMY',
+    'EntityExtractor',
+    'RelationExtractor',
+    'FactExtractor',
+    'DocumentClassifier',
+    'EntitySchemaDiscoverer',
+    
+    # Presets
+    'PYTHON_PRESET',
+    'PYTHON_PRESET_WITH_TESTS',
+    'JAVASCRIPT_PRESET',
+    'TYPESCRIPT_PRESET',
+    'REACT_PRESET',
+    'NEXTJS_PRESET',
+    'JAVA_PRESET',
+    'SPRING_BOOT_PRESET',
+    'MAVEN_PRESET',
+    'GRADLE_PRESET',
+    'DOTNET_PRESET',
+    'CSHARP_PRESET',
+    'ASPNET_PRESET',
+    'FULLSTACK_JS_PRESET',
+    'MONOREPO_PRESET',
+    'DOCUMENTATION_PRESET',
+    'PRESETS',
+    'get_preset',
+    'list_presets',
+    'combine_presets',
+]

alita_sdk/community/inventory/config.py

@@ -0,0 +1,257 @@
+"""
+Configuration for Inventory Ingestion Pipeline.
+
+Since the ingestion runs within Alita, the LLM and embeddings are provided
+by the Alita client. Configuration only needs model names, not providers.
+
+Usage:
+    # From YAML config file
+    config = IngestionConfig.from_yaml("./ingestion-config.yml")
+    
+    # Programmatic
+    config = IngestionConfig(
+        llm_model="gpt-4o-mini",
+        embedding_model="text-embedding-3-small",
+        guardrails=GuardrailsConfig(
+            max_tokens_per_doc=8000,
+            max_entities_per_doc=50,
+        )
+    )
+    
+    # Use in pipeline (Alita client provides LLM/embeddings)
+    pipeline = IngestionPipeline(
+        llm=alita.get_langchain_llm(config.llm_model),
+        embedding=alita.get_embeddings(config.embedding_model),
+        graph_path=config.graph_path,
+        guardrails=config.guardrails,
+    )
+"""
+
+import os
+import logging
+from typing import Any, Optional, Dict, List
+from pydantic import BaseModel, Field
+
+logger = logging.getLogger(__name__)
+
+
+class GuardrailsConfig(BaseModel):
+    """Guardrails configuration for safe and controlled extraction."""
+    
+    # Token/content limits
+    max_tokens_per_doc: int = Field(
+        default=8000, 
+        description="Maximum tokens per document before chunking"
+    )
+    max_entities_per_doc: int = Field(
+        default=50, 
+        description="Maximum entities to extract from a single document"
+    )
+    max_relations_per_doc: int = Field(
+        default=100,
+        description="Maximum relations to extract per document"
+    )
+    
+    # Content filtering
+    content_filter_enabled: bool = Field(
+        default=True,
+        description="Enable content filtering for PII/secrets"
+    )
+    filter_patterns: List[str] = Field(
+        default_factory=lambda: [
+            r'(?i)(password|secret|api[_-]?key|token)\s*[=:]\s*["\'][^"\']+["\']',
+            r'\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b',
+            r'-----BEGIN [A-Z]+ PRIVATE KEY-----',
+        ],
+        description="Regex patterns to filter from content before LLM processing"
+    )
+    
+    # Rate limiting
+    rate_limit_requests_per_minute: Optional[int] = Field(
+        default=None,
+        description="Max LLM requests per minute (None = unlimited)"
+    )
+    rate_limit_tokens_per_minute: Optional[int] = Field(
+        default=None,
+        description="Max tokens per minute (None = unlimited)"
+    )
+    
+    # Error handling
+    max_retries: int = Field(default=3, description="Max retries on LLM errors")
+    retry_delay_seconds: float = Field(default=1.0, description="Delay between retries")
+    skip_on_error: bool = Field(
+        default=True, 
+        description="Skip document on extraction error vs fail pipeline"
+    )
+    
+    # Validation
+    validate_entity_types: bool = Field(
+        default=True,
+        description="Validate extracted entities against taxonomy"
+    )
+    validate_relation_types: bool = Field(
+        default=True,
+        description="Validate extracted relations against taxonomy"
+    )
+    
+    # Deduplication
+    deduplicate_entities: bool = Field(
+        default=True,
+        description="Merge duplicate entities by name+type+file"
+    )
+    
+    # Confidence thresholds
+    entity_confidence_threshold: float = Field(
+        default=0.5,
+        description="Minimum confidence for entity extraction"
+    )
+    relation_confidence_threshold: float = Field(
+        default=0.5,
+        description="Minimum confidence for relation extraction"
+    )
+
+
+class IngestionConfig(BaseModel):
+    """
+    Configuration for the ingestion pipeline.
+    
+    Since ingestion runs within Alita, only model names are needed.
+    The Alita client handles provider details, API keys, etc.
+    """
+    
+    # Model names (Alita provides the actual LLM/embedding instances)
+    llm_model: str = Field(
+        default="gpt-4o-mini", 
+        description="LLM model name (e.g., gpt-4o-mini, claude-3-sonnet)"
+    )
+    embedding_model: Optional[str] = Field(
+        default=None, 
+        description="Embedding model name (optional, for semantic search)"
+    )
+    
+    # Model parameters
+    temperature: float = Field(default=0.0, description="LLM temperature")
+    
+    # Guardrails configuration
+    guardrails: GuardrailsConfig = Field(default_factory=GuardrailsConfig)
+    
+    # Graph configuration
+    graph_path: str = Field(default="./knowledge_graph.json", description="Path to persist graph")
+    auto_save: bool = Field(default=True, description="Auto-save after mutations")
+    
+    # Extraction settings
+    extract_relations: bool = Field(default=True, description="Extract relations between entities")
+    chunk_size: int = Field(default=4000, description="Document chunk size for processing")
+    chunk_overlap: int = Field(default=200, description="Overlap between chunks")
+    
+    # Concurrency
+    max_concurrent_extractions: int = Field(
+        default=1, 
+        description="Max parallel extraction tasks (1 = sequential)"
+    )
+    
+    @classmethod
+    def from_yaml(cls, path: str) -> "IngestionConfig":
+        """Load configuration from YAML file."""
+        import yaml
+        
+        with open(path, 'r') as f:
+            data = yaml.safe_load(f)
+        
+        return cls(**data)
+    
+    @classmethod
+    def from_json(cls, path: str) -> "IngestionConfig":
+        """Load configuration from JSON file."""
+        import json
+        
+        with open(path, 'r') as f:
+            data = json.load(f)
+        
+        return cls(**data)
+    
+    @classmethod
+    def from_env(cls) -> "IngestionConfig":
+        """
+        Create configuration from environment variables.
+        
+        Environment variables:
+            LLM_MODEL: Model name (default: gpt-4o-mini)
+            EMBEDDING_MODEL: Embedding model name (optional)
+            LLM_TEMPERATURE: Temperature (default: 0.0)
+            GRAPH_PATH: Path to save graph (default: ./knowledge_graph.json)
+            MAX_TOKENS_PER_DOC: Max tokens per doc (default: 8000)
+            MAX_ENTITIES_PER_DOC: Max entities per doc (default: 50)
+            CONTENT_FILTER_ENABLED: true/false (default: true)
+            EXTRACT_RELATIONS: true/false (default: true)
+        """
+        guardrails = GuardrailsConfig(
+            max_tokens_per_doc=int(os.environ.get('MAX_TOKENS_PER_DOC', '8000')),
+            max_entities_per_doc=int(os.environ.get('MAX_ENTITIES_PER_DOC', '50')),
+            content_filter_enabled=os.environ.get('CONTENT_FILTER_ENABLED', 'true').lower() == 'true',
+            max_retries=int(os.environ.get('MAX_RETRIES', '3')),
+        )
+        
+        return cls(
+            llm_model=os.environ.get('LLM_MODEL', 'gpt-4o-mini'),
+            embedding_model=os.environ.get('EMBEDDING_MODEL'),
+            temperature=float(os.environ.get('LLM_TEMPERATURE', '0.0')),
+            guardrails=guardrails,
+            graph_path=os.environ.get('GRAPH_PATH', './knowledge_graph.json'),
+            extract_relations=os.environ.get('EXTRACT_RELATIONS', 'true').lower() == 'true',
+        )
+    
+    def to_yaml(self, path: str) -> None:
+        """Save configuration to YAML file."""
+        import yaml
+        
+        with open(path, 'w') as f:
+            yaml.safe_dump(self.model_dump(), f, default_flow_style=False)
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert to dictionary."""
+        return self.model_dump()
+
+
+# Example YAML configuration template
+EXAMPLE_CONFIG_YAML = """# Inventory Ingestion Configuration
+# Model names only - Alita provides the actual LLM/embedding instances
+
+# LLM model name (required)
+llm_model: gpt-4o-mini
+temperature: 0.0
+
+# Embedding model (optional, for semantic search)
+embedding_model: text-embedding-3-small
+
+# Guardrails - safety and control
+guardrails:
+  max_tokens_per_doc: 8000
+  max_entities_per_doc: 50
+  max_relations_per_doc: 100
+  content_filter_enabled: true
+  max_retries: 3
+  retry_delay_seconds: 1.0
+  skip_on_error: true
+  entity_confidence_threshold: 0.5
+  relation_confidence_threshold: 0.5
+  deduplicate_entities: true
+  # rate_limit_requests_per_minute: 60  # Uncomment to rate limit
+
+# Graph persistence
+graph_path: ./knowledge_graph.json
+auto_save: true
+
+# Extraction settings
+extract_relations: true
+chunk_size: 4000
+chunk_overlap: 200
+max_concurrent_extractions: 1
+"""
+
+
+def generate_config_template(output_path: str = "./ingestion-config.yml") -> str:
+    """Generate a configuration template file."""
+    with open(output_path, 'w') as f:
+        f.write(EXAMPLE_CONFIG_YAML)
+    return output_path