crackerjack 0.18.2__py3-none-any.whl → 0.45.2__py3-none-any.whl

crackerjack/agents/semantic_agent.py

@@ -0,0 +1,479 @@
+"""Semantic search and context analysis agent for code pattern discovery and semantic improvements."""
+
+import typing as t
+from pathlib import Path
+
+from ..models.semantic_models import SearchQuery, SemanticConfig
+from ..services.vector_store import VectorStore
+from .base import (
+    AgentContext,
+    FixResult,
+    Issue,
+    IssueType,
+    SubAgent,
+    agent_registry,
+)
+
+
+class SemanticAgent(SubAgent):
+    """AI agent specialized in semantic search and code context analysis.
+
+    This agent enhances code understanding by providing semantic context,
+    finding similar code patterns, and suggesting improvements based on
+    codebase-wide analysis using vector embeddings.
+    """
+
+    def __init__(self, context: AgentContext) -> None:
+        super().__init__(context)
+        self.semantic_insights: dict[str, t.Any] = {}
+        self.pattern_stats: dict[str, int] = {
+            "patterns_discovered": 0,
+            "context_enhancements": 0,
+            "semantic_suggestions": 0,
+            "similar_patterns_found": 0,
+        }
+
+    def get_supported_types(self) -> set[IssueType]:
+        return {IssueType.SEMANTIC_CONTEXT}
+
+    async def can_handle(self, issue: Issue) -> float:
+        """Determine confidence level for handling semantic context issues."""
+        if issue.type != IssueType.SEMANTIC_CONTEXT:
+            return 0.0
+
+        confidence = 0.8
+        message_lower = issue.message.lower()
+
+        # Higher confidence for semantic-specific terms
+        if any(
+            pattern in message_lower
+            for pattern in (
+                "semantic",
+                "context",
+                "pattern",
+                "similarity",
+                "related code",
+                "code understanding",
+                "similar implementation",
+            )
+        ):
+            confidence = 0.85
+
+        return confidence
+
+    async def analyze_and_fix(self, issue: Issue) -> FixResult:
+        """Analyze code using semantic search and provide contextual insights."""
+        self.log(f"Analyzing semantic context issue: {issue.message}")
+
+        validation_result = self._validate_semantic_issue(issue)
+        if validation_result:
+            return validation_result
+
+        if issue.file_path is None:
+            return FixResult(
+                success=False,
+                confidence=0.0,
+                remaining_issues=["No file path provided for semantic analysis"],
+            )
+
+        file_path = Path(issue.file_path)
+
+        try:
+            # Initialize semantic services
+            config = self._create_semantic_config()
+            vector_store = self._get_vector_store(config)
+
+            # Perform semantic analysis
+            result = await self._perform_semantic_analysis(
+                file_path, vector_store, issue
+            )
+
+            # Update stats
+            self._update_pattern_stats(result)
+
+            return result
+
+        except Exception as e:
+            return self._create_semantic_error_result(e)
+
+    @staticmethod
+    def _validate_semantic_issue(issue: Issue) -> FixResult | None:
+        """Validate that the semantic issue can be processed."""
+        if not issue.file_path:
+            return FixResult(
+                success=False,
+                confidence=0.0,
+                remaining_issues=["No file path specified for semantic analysis"],
+            )
+
+        file_path = Path(issue.file_path)
+        if not file_path.exists():
+            return FixResult(
+                success=False,
+                confidence=0.0,
+                remaining_issues=[f"File not found: {file_path}"],
+            )
+
+        return None
+
+    @staticmethod
+    def _create_semantic_config() -> SemanticConfig:
+        """Create semantic search configuration."""
+        return SemanticConfig(
+            embedding_model="sentence-transformers/all-MiniLM-L6-v2",
+            chunk_size=512,
+            chunk_overlap=50,
+            max_search_results=10,
+            similarity_threshold=0.7,
+            embedding_dimension=384,
+        )
+
+    def _get_vector_store(self, config: SemanticConfig) -> VectorStore:
+        """Get vector store instance with persistent database."""
+        db_path = self._get_persistent_db_path()
+        return VectorStore(config, db_path=db_path)
+
+    def _get_persistent_db_path(self) -> Path:
+        """Get the path to the persistent semantic search database."""
+        db_path = self.context.project_path / ".crackerjack" / "semantic_index.db"
+        db_path.parent.mkdir(exist_ok=True)
+        return db_path
+
+    async def _perform_semantic_analysis(
+        self, file_path: Path, vector_store: VectorStore, issue: Issue
+    ) -> FixResult:
+        """Perform comprehensive semantic analysis of the code file."""
+        # Read file content
+        content = self.context.get_file_content(file_path)
+        if not content:
+            return FixResult(
+                success=False,
+                confidence=0.0,
+                remaining_issues=[f"Could not read file: {file_path}"],
+            )
+
+        # Index the file if not already indexed
+        try:
+            embeddings = vector_store.index_file(file_path)
+            self.log(f"Indexed {len(embeddings)} chunks from {file_path.name}")
+        except Exception as e:
+            self.log(f"Warning: Could not index file {file_path}: {e}")
+
+        # Perform semantic search for related patterns
+        semantic_insights = await self._discover_semantic_patterns(
+            vector_store, file_path, content, issue
+        )
+
+        # Generate recommendations based on semantic analysis
+        recommendations = self._generate_semantic_recommendations(semantic_insights)
+
+        return FixResult(
+            success=True,
+            confidence=0.8,
+            fixes_applied=[
+                f"Semantic analysis completed for {file_path.name}",
+                f"Discovered {len(semantic_insights.get('related_patterns', []))} related patterns",
+                f"Generated {len(recommendations)} semantic recommendations",
+            ],
+            recommendations=recommendations,
+            files_modified=[],  # Semantic agent provides insights, doesn't modify files
+        )
+
+    async def _discover_semantic_patterns(
+        self,
+        vector_store: VectorStore,
+        file_path: Path,
+        content: str,
+        issue: Issue,
+    ) -> dict[str, t.Any]:
+        """Discover semantic patterns and related code through vector search."""
+        insights: dict[str, t.Any] = {
+            "related_patterns": [],
+            "similar_functions": [],
+            "context_suggestions": [],
+            "pattern_clusters": [],
+        }
+
+        # Extract key functions and classes for semantic analysis
+        code_elements = self._extract_code_elements(content)
+
+        for element in code_elements:
+            # Search for similar patterns
+            search_query = SearchQuery(
+                query=element["signature"],
+                max_results=5,
+                min_similarity=0.6,
+                file_types=["py"],
+            )
+
+            try:
+                results = vector_store.search(search_query)
+                if results:
+                    # Filter out results from the same file
+                    related_results = [
+                        result for result in results if result.file_path != file_path
+                    ]
+
+                    if related_results:
+                        insights["related_patterns"].append(
+                            {
+                                "element": element,
+                                "related_code": [
+                                    {
+                                        "file_path": str(result.file_path),
+                                        "content": result.content[
+                                            :200
+                                        ],  # Truncate for readability
+                                        "similarity_score": result.similarity_score,
+                                        "lines": f"{result.start_line}-{result.end_line}",
+                                    }
+                                    for result in related_results[:3]  # Top 3 matches
+                                ],
+                            }
+                        )
+
+            except Exception as e:
+                self.log(f"Warning: Semantic search failed for {element['name']}: {e}")
+
+        # Analyze issue-specific context
+        if issue.message:
+            issue_insights = await self._analyze_issue_context(vector_store, issue)
+            insights["context_suggestions"].extend(issue_insights)
+
+        return insights
+
+    def _extract_docstring_from_node(self, node: t.Any) -> str:
+        """Extract docstring from AST node, handling both old and new formats."""
+        import ast
+
+        if not node.body or not isinstance(node.body[0], ast.Expr):
+            return ""
+
+        value = node.body[0].value
+        if hasattr(value, "s"):  # Old ast.Str format
+            return str(value.s)[:100]
+        elif hasattr(value, "value") and isinstance(
+            value.value, str
+        ):  # New ast.Constant format
+            return str(value.value)[:100]
+        return ""
+
+    def _build_function_signature(self, node: t.Any) -> str:
+        """Build function signature from AST FunctionDef node."""
+        signature = f"def {node.name}("
+        if node.args.args:
+            args = [arg.arg for arg in node.args.args[:3]]  # First 3 args
+            signature += ", ".join(args)
+        signature += ")"
+        return signature
+
+    def _build_class_signature(self, node: t.Any) -> str:
+        """Build class signature from AST ClassDef node."""
+        signature = f"class {node.name}"
+        if node.bases:
+            bases = [self._get_ast_name(base) for base in node.bases[:2]]
+            signature += f"({', '.join(bases)})"
+        return signature
+
+    def _get_ast_name(self, node: t.Any) -> str:
+        """Get name from AST node."""
+        import ast
+
+        if isinstance(node, ast.Name):
+            return node.id
+        elif isinstance(node, ast.Attribute):
+            return f"{self._get_ast_name(node.value)}.{node.attr}"
+        return "Unknown"
+
+    def _extract_ast_elements(self, content: str) -> list[dict[str, t.Any]]:
+        """Extract code elements using AST parsing."""
+        import ast
+
+        class CodeElementExtractor(ast.NodeVisitor):
+            def __init__(self, parent: "SemanticAgent") -> None:
+                self.elements: list[dict[str, t.Any]] = []
+                self.parent = parent
+
+            def visit_FunctionDef(self, node: ast.FunctionDef) -> None:
+                self.elements.append(
+                    {
+                        "type": "function",
+                        "name": node.name,
+                        "signature": self.parent._build_function_signature(node),
+                        "docstring": self.parent._extract_docstring_from_node(node),
+                        "line_number": node.lineno,
+                    }
+                )
+                self.generic_visit(node)
+
+            def visit_ClassDef(self, node: ast.ClassDef) -> None:
+                self.elements.append(
+                    {
+                        "type": "class",
+                        "name": node.name,
+                        "signature": self.parent._build_class_signature(node),
+                        "line_number": node.lineno,
+                    }
+                )
+                self.generic_visit(node)
+
+        tree = ast.parse(content)
+        extractor = CodeElementExtractor(self)
+        extractor.visit(tree)
+        return extractor.elements[:10]  # Limit to top 10 elements
+
+    def _extract_text_elements(self, content: str) -> list[dict[str, t.Any]]:
+        """Extract code elements using simple text patterns."""
+        elements = []
+        lines = content.split("\n")
+        for i, line in enumerate(lines[:50]):  # Check first 50 lines
+            stripped = line.strip()
+            if stripped.startswith("def ") and "(" in stripped:
+                func_name = stripped.split("(")[0].replace("def ", "").strip()
+                elements.append(
+                    {
+                        "type": "function",
+                        "name": func_name,
+                        "signature": stripped,
+                        "line_number": i + 1,
+                    }
+                )
+            elif stripped.startswith("class ") and ":" in stripped:
+                class_name = stripped.split(":")[0].replace("class ", "").strip()
+                elements.append(
+                    {
+                        "type": "class",
+                        "name": class_name,
+                        "signature": stripped,
+                        "line_number": i + 1,
+                    }
+                )
+        return elements
+
+    def _extract_code_elements(self, content: str) -> list[dict[str, t.Any]]:
+        """Extract key code elements for semantic analysis."""
+        try:
+            return self._extract_ast_elements(content)
+        except SyntaxError:
+            return self._extract_text_elements(content)
+
+    async def _analyze_issue_context(
+        self, vector_store: VectorStore, issue: Issue
+    ) -> list[dict[str, t.Any]]:
+        """Analyze the specific issue context using semantic search."""
+        suggestions = []
+
+        # Search for similar issues or patterns
+        search_query = SearchQuery(
+            query=issue.message,
+            max_results=5,
+            min_similarity=0.5,
+        )
+
+        try:
+            results = vector_store.search(search_query)
+            if results:
+                suggestions.append(
+                    {
+                        "type": "similar_issues",
+                        "description": f"Found {len(results)} similar patterns in codebase",
+                        "examples": [
+                            {
+                                "file": str(result.file_path.name),
+                                "content": result.content[:150],
+                                "similarity": result.similarity_score,
+                            }
+                            for result in results[:3]
+                        ],
+                    }
+                )
+        except Exception as e:
+            self.log(f"Warning: Issue context analysis failed: {e}")
+
+        return suggestions
+
+    def _generate_semantic_recommendations(
+        self, insights: dict[str, t.Any]
+    ) -> list[str]:
+        """Generate actionable recommendations based on semantic analysis."""
+        recommendations = []
+
+        related_patterns = insights.get("related_patterns", [])
+        context_suggestions = insights.get("context_suggestions", [])
+
+        if related_patterns:
+            recommendations.append(
+                f"Found {len(related_patterns)} similar code patterns across the codebase"
+            )
+
+            # Analyze pattern consistency
+            high_similarity_count = sum(
+                1
+                for pattern in related_patterns
+                for code in pattern["related_code"]
+                if code["similarity_score"] > 0.8
+            )
+
+            if high_similarity_count > 0:
+                recommendations.append(
+                    f"Detected {high_similarity_count} highly similar implementations - "
+                    "consider refactoring for DRY principle compliance"
+                )
+
+        if context_suggestions:
+            recommendations.append(
+                "Semantic analysis revealed contextual insights for code understanding"
+            )
+
+        # General semantic recommendations
+        recommendations.extend(
+            [
+                "Consider semantic indexing of related modules for better code discovery",
+                "Review similar patterns for consistency in naming and implementation",
+                "Use semantic search to discover reusable components before implementing new ones",
+            ]
+        )
+
+        return recommendations
+
+    def _update_pattern_stats(self, result: FixResult) -> None:
+        """Update pattern discovery statistics."""
+        if result.success:
+            self.pattern_stats["patterns_discovered"] += len(result.fixes_applied)
+            self.pattern_stats["semantic_suggestions"] += len(result.recommendations)
+
+    @staticmethod
+    def _create_semantic_error_result(error: Exception) -> FixResult:
+        """Create error result for semantic analysis failures."""
+        return FixResult(
+            success=False,
+            confidence=0.0,
+            remaining_issues=[f"Semantic analysis failed: {error}"],
+            recommendations=[
+                "Ensure semantic search index is properly initialized",
+                "Check if file contains valid code for analysis",
+                "Verify semantic search configuration is correct",
+            ],
+        )
+
+    async def plan_before_action(self, issue: Issue) -> dict[str, t.Any]:
+        """Plan semantic analysis strategy before execution."""
+        return {
+            "strategy": "semantic_context_analysis",
+            "confidence": 0.8,
+            "approach": [
+                "Index file content for semantic search",
+                "Discover related code patterns using vector embeddings",
+                "Analyze semantic similarity across codebase",
+                "Generate contextual recommendations",
+            ],
+            "expected_insights": [
+                "Similar code patterns and implementations",
+                "Opportunities for code reuse and refactoring",
+                "Contextual understanding of code relationships",
+            ],
+        }
+
+
+# Register the agent with the agent registry
+agent_registry.register(SemanticAgent)