tapps-agents 3.6.0__py3-none-any.whl → 3.6.1__py3-none-any.whl

tapps_agents/simple_mode/intent_parser.py

@@ -1,510 +1,510 @@
-"""
-Intent Parser - Parse natural language input into structured intents.
-
-Maps user commands to agent sequences using keyword matching and
-simple NLP patterns.
-"""
-
-import re
-from dataclasses import dataclass
-from enum import Enum
-from typing import Any
-
-
-class IntentType(Enum):
-    """Types of user intents."""
-
-    BUILD = "build"
-    VALIDATE = "validate"  # Validation workflow for comparing implementations
-    REVIEW = "review"
-    FIX = "fix"
-    TEST = "test"
-    EPIC = "epic"
-    EXPLORE = "explore"
-    REFACTOR = "refactor"
-    PLAN_ANALYSIS = "plan-analysis"
-    PR = "pr"
-    REQUIREMENTS = "requirements"
-    BROWNFIELD = "brownfield"
-    ENHANCE = "enhance"  # §3.4: first-class *enhance
-    BREAKDOWN = "breakdown"  # §3.4: first-class *breakdown
-    TODO = "todo"  # §3.5: *todo (Beads-backed)
-    UNKNOWN = "unknown"
-
-
-@dataclass
-class Intent:
-    """Structured intent parsed from user input."""
-
-    type: IntentType
-    confidence: float
-    parameters: dict[str, Any]
-    original_input: str
-    compare_to_codebase: bool = False  # Flag for "compare to codebase" intent
-
-    def get_agent_sequence(self) -> list[str]:
-        """Get the sequence of agents for this intent."""
-        if self.type == IntentType.BUILD:
-            return ["planner", "architect", "designer", "implementer"]
-        elif self.type == IntentType.REVIEW:
-            return ["reviewer", "improver"]
-        elif self.type == IntentType.FIX:
-            return ["debugger", "implementer", "tester"]
-        elif self.type == IntentType.TEST:
-            return ["tester"]
-        elif self.type == IntentType.EPIC:
-            return ["epic-orchestrator"]
-        elif self.type == IntentType.EXPLORE:
-            return ["analyst", "reviewer"]
-        elif self.type == IntentType.REFACTOR:
-            return ["reviewer", "architect", "implementer", "tester"]
-        elif self.type == IntentType.PLAN_ANALYSIS:
-            return ["analyst", "architect", "reviewer"]
-        elif self.type == IntentType.PR:
-            return ["reviewer", "documenter"]
-        elif self.type == IntentType.REQUIREMENTS:
-            return ["analyst", "planner", "documenter"]
-        elif self.type == IntentType.BROWNFIELD:
-            return ["brownfield-review"]
-        elif self.type == IntentType.ENHANCE:
-            return ["enhancer"]
-        elif self.type == IntentType.BREAKDOWN:
-            return ["planner"]
-        elif self.type == IntentType.TODO:
-            return ["beads"]
-        else:
-            return []
-
-
-class IntentParser:
-    """Parse natural language input into structured intents."""
-
-    def __init__(self):
-        """Initialize the intent parser with keyword mappings."""
-        # Build intent keywords
-        self.build_keywords = [
-            "build",
-            "create",
-            "make",
-            "generate",
-            "add",
-            "implement",
-            "develop",
-            "write",
-            "new",
-            "feature",
-        ]
-
-        # Review intent keywords
-        self.review_keywords = [
-            "review",
-            "check",
-            "analyze",
-            "inspect",
-            "examine",
-            "score",
-            "quality",
-            "audit",
-            "assess",
-            "evaluate",
-            "compare",
-            "compare to",
-            "match",
-            "align with",
-            "follow patterns",
-        ]
-
-        # Fix intent keywords
-        self.fix_keywords = [
-            "fix",
-            "repair",
-            "resolve",
-            "debug",
-            "error",
-            "bug",
-            "issue",
-            "problem",
-            "broken",
-            "correct",
-        ]
-
-        # Test intent keywords
-        self.test_keywords = [
-            "test",
-            "verify",
-            "validate",
-            "coverage",
-            "testing",
-            "tests",
-            "unit test",
-            "integration test",
-        ]
-
-        # Epic intent keywords
-        self.epic_keywords = [
-            "epic",
-            "implement epic",
-            "execute epic",
-            "run epic",
-            "story",
-            "stories",
-        ]
-
-        # Explore intent keywords
-        self.explore_keywords = [
-            "explore",
-            "understand",
-            "navigate",
-            "find",
-            "discover",
-            "overview",
-            "codebase",
-            "trace",
-            "search",
-            "locate",
-        ]
-
-        # Refactor intent keywords
-        self.refactor_keywords = [
-            "refactor",
-            "modernize",
-            "update",
-            "improve code",
-            "modernize code",
-            "legacy",
-            "deprecated",
-            "upgrade code",
-        ]
-
-        # Plan Analysis intent keywords
-        self.plan_analysis_keywords = [
-            "plan",
-            "planning",
-            "analyze",
-            "analysis",
-            "design",
-            "proposal",
-            "strategy",
-            "roadmap",
-        ]
-
-        # PR intent keywords
-        self.pr_keywords = [
-            "pr",
-            "pull request",
-            "create pr",
-            "open pr",
-            "merge request",
-            "mr",
-        ]
-
-        # Requirements intent keywords
-        self.requirements_keywords = [
-            "requirements",
-            "gather requirements",
-            "extract requirements",
-            "document requirements",
-            "analyze requirements",
-            "requirements document",
-            "requirements gathering",
-            "requirements analysis",
-        ]
-
-        # Brownfield intent keywords
-        self.brownfield_keywords = [
-            "brownfield",
-            "brownfield review",
-            "review brownfield",
-            "analyze brownfield",
-            "brownfield system",
-            "review system",
-            "analyze project",
-            "create experts",
-            "auto experts",
-            "expert creation",
-            "populate rag",
-            "fill rag",
-        ]
-
-        # Simple Mode intent keywords
-        self.simple_mode_keywords = [
-            "@simple-mode",
-            "simple mode",
-            "use simple mode",
-            "simple-mode",
-            "@simple_mode",
-            "simple_mode",
-        ]
-
-    def detect_simple_mode_intent(self, text: str) -> bool:
-        """
-        Detect if user wants Simple Mode.
-
-        Args:
-            text: User input text
-
-        Returns:
-            True if Simple Mode intent detected, False otherwise
-
-        Examples:
-            >>> parser = IntentParser()
-            >>> parser.detect_simple_mode_intent("@simple-mode build feature")
-            True
-            >>> parser.detect_simple_mode_intent("use simple mode to create api")
-            True
-            >>> parser.detect_simple_mode_intent("build feature")
-            False
-        """
-        text_lower = text.lower()
-        return any(keyword in text_lower for keyword in self.simple_mode_keywords)
-
-    def parse(self, input_text: str) -> Intent:
-        """
-        Parse natural language input into a structured intent.
-
-        Args:
-            input_text: User's natural language command
-
-        Returns:
-            Intent object with type, confidence, and parameters
-        """
-        input_lower = input_text.lower().strip()
-
-        # Extract parameters
-        parameters = self._extract_parameters(input_text)
-
-        # Check for Simple Mode intent first
-        if self.detect_simple_mode_intent(input_text):
-            # Force Simple Mode workflow
-            parameters["force_simple_mode"] = True
-
-        # Check for explicit command patterns
-        if input_text.strip().startswith("*epic") or input_text.strip().startswith("@simple-mode *epic"):
-            # Extract epic path
-            epic_path = input_text.replace("*epic", "").replace("@simple-mode", "").strip()
-            if epic_path:
-                parameters["epic_path"] = epic_path
-            return Intent(
-                type=IntentType.EPIC,
-                confidence=1.0,
-                parameters=parameters,
-                original_input=input_text,
-            )
-
-        if input_text.strip().startswith("*explore") or input_text.strip().startswith("@simple-mode *explore"):
-            return Intent(
-                type=IntentType.EXPLORE,
-                confidence=1.0,
-                parameters=parameters,
-                original_input=input_text,
-            )
-
-        if input_text.strip().startswith("*validate") or input_text.strip().startswith("@simple-mode *validate"):
-            return Intent(
-                type=IntentType.VALIDATE,
-                confidence=1.0,
-                parameters=parameters,
-                original_input=input_text,
-            )
-
-        if input_text.strip().startswith("*refactor") or input_text.strip().startswith("@simple-mode *refactor"):
-            return Intent(
-                type=IntentType.REFACTOR,
-                confidence=1.0,
-                parameters=parameters,
-                original_input=input_text,
-            )
-
-        if input_text.strip().startswith("*plan") or input_text.strip().startswith("@simple-mode *plan") or "plan-analysis" in input_lower:
-            return Intent(
-                type=IntentType.PLAN_ANALYSIS,
-                confidence=1.0,
-                parameters=parameters,
-                original_input=input_text,
-            )
-
-        if input_text.strip().startswith("*pr") or input_text.strip().startswith("@simple-mode *pr") or "pull request" in input_lower:
-            return Intent(
-                type=IntentType.PR,
-                confidence=1.0,
-                parameters=parameters,
-                original_input=input_text,
-            )
-
-        if input_text.strip().startswith("*brownfield") or input_text.strip().startswith("*brownfield-review") or input_text.strip().startswith("@simple-mode *brownfield"):
-            return Intent(
-                type=IntentType.BROWNFIELD,
-                confidence=1.0,
-                parameters=parameters,
-                original_input=input_text,
-            )
-
-        # *enhance and *breakdown (§3.4)
-        if input_text.strip().startswith("*enhance") or input_text.strip().startswith("@simple-mode *enhance"):
-            prompt = (parameters.get("description") or re.sub(r"^(@simple-mode\s*)?\*enhance\s*", "", input_text, flags=re.I).strip() or input_text)
-            parameters["prompt"] = prompt
-            return Intent(type=IntentType.ENHANCE, confidence=1.0, parameters=parameters, original_input=input_text)
-        if input_text.strip().startswith("*breakdown") or input_text.strip().startswith("@simple-mode *breakdown"):
-            prompt = (parameters.get("description") or re.sub(r"^(@simple-mode\s*)?\*breakdown\s*", "", input_text, flags=re.I).strip() or input_text)
-            parameters["prompt"] = prompt
-            return Intent(type=IntentType.BREAKDOWN, confidence=1.0, parameters=parameters, original_input=input_text)
-
-        # *todo (§3.5) – remainder forwarded to bd
-        if input_text.strip().startswith("*todo") or input_text.strip().startswith("@simple-mode *todo"):
-            rest = re.sub(r"^(@simple-mode\s*)?\*todo\s*", "", input_text.strip(), flags=re.I).strip()
-            parameters["todo_rest"] = rest
-            return Intent(type=IntentType.TODO, confidence=1.0, parameters=parameters, original_input=input_text)
-
-        # Score each intent type
-        scores = {
-            IntentType.BUILD: self._score_intent(input_lower, self.build_keywords),
-            IntentType.REVIEW: self._score_intent(input_lower, self.review_keywords),
-            IntentType.FIX: self._score_intent(input_lower, self.fix_keywords),
-            IntentType.TEST: self._score_intent(input_lower, self.test_keywords),
-            IntentType.EPIC: self._score_intent(input_lower, self.epic_keywords),
-            IntentType.EXPLORE: self._score_intent(input_lower, self.explore_keywords),
-            IntentType.REFACTOR: self._score_intent(input_lower, self.refactor_keywords),
-            IntentType.PLAN_ANALYSIS: self._score_intent(input_lower, self.plan_analysis_keywords),
-            IntentType.PR: self._score_intent(input_lower, self.pr_keywords),
-            IntentType.REQUIREMENTS: self._score_intent(input_lower, self.requirements_keywords),
-            IntentType.BROWNFIELD: self._score_intent(input_lower, self.brownfield_keywords),
-        }
-
-        # Find best match
-        best_intent = max(scores.items(), key=lambda x: x[1])
-        intent_type, confidence = best_intent
-
-        # If confidence is too low, mark as unknown
-        if confidence < 0.3:
-            intent_type = IntentType.UNKNOWN
-            confidence = 0.0
-
-        # Detect "compare to codebase" intent with enhanced semantics
-        compare_to_codebase = False
-        compare_phrases = [
-            # Direct compare phrases
-            "compare to",
-            "compare with",
-            "compare against",
-            "compare this to",
-            "compare this with",
-            "compare to codebase",
-            "compare to our",
-            "compare to project",
-            # Match/align phrases
-            "match our",
-            "match the",
-            "match patterns",
-            "match codebase",
-            "align with",
-            "align to",
-            "follow patterns",
-            "follow our patterns",
-            "follow project patterns",
-            # Make match phrases (implicit compare)
-            "make match",
-            "make this match",
-            "make it match",
-            # Consistency phrases
-            "consistent with",
-            "consistency with",
-            "conform to",
-            "conform with",
-            # Standard phrases
-            "match standards",
-            "follow standards",
-            "meet standards",
-            "adhere to",
-        ]
-        if any(phrase in input_lower for phrase in compare_phrases):
-            compare_to_codebase = True
-
-        return Intent(
-            type=intent_type,
-            confidence=confidence,
-            parameters=parameters,
-            original_input=input_text,
-            compare_to_codebase=compare_to_codebase,
-        )
-
-    def _score_intent(self, text: str, keywords: list[str]) -> float:
-        """
-        Score how well text matches an intent based on keywords.
-
-        Args:
-            text: Lowercase input text
-            keywords: List of keywords for this intent
-
-        Returns:
-            Confidence score between 0.0 and 1.0
-        """
-        matches = sum(1 for keyword in keywords if keyword in text)
-        if not matches:
-            return 0.0
-
-        # Normalize by number of keywords (max score is 1.0)
-        score = min(matches / len(keywords) * 2.0, 1.0)
-
-        # Boost score if keyword appears at start of sentence
-        for keyword in keywords:
-            if text.startswith(keyword):
-                score = min(score + 0.2, 1.0)
-                break
-
-        return score
-
-    def _extract_parameters(self, text: str) -> dict[str, Any]:
-        """
-        Extract parameters from user input.
-
-        Args:
-            text: User input text
-
-        Returns:
-            Dictionary of extracted parameters
-        """
-        parameters: dict[str, Any] = {}
-
-        # Extract file paths (quoted strings or common file extensions)
-        file_pattern = r'["\']?([^\s"\']+\.(py|ts|js|tsx|jsx|java|go|rs|rb|php|cs|sql|yaml|yml|json|md|txt))["\']?'
-        file_matches = re.findall(file_pattern, text, re.IGNORECASE)
-        if file_matches:
-            parameters["files"] = [match[0] for match in file_matches]
-
-        # Extract quoted descriptions
-        quoted_pattern = r'["\']([^"\']+)["\']'
-        quoted_matches = re.findall(quoted_pattern, text)
-        if quoted_matches:
-            parameters["description"] = quoted_matches[0]
-
-        # Extract Epic document paths (epic-*.md or docs/prd/epic-*.md)
-        epic_pattern = r'(?:epic-[\w-]+\.md|docs/prd/epic-[\w-]+\.md|[\w/]+epic[\w-]*\.md)'
-        epic_matches = re.findall(epic_pattern, text, re.IGNORECASE)
-        if epic_matches:
-            parameters["epic_path"] = epic_matches[0]
-
-        # Extract feature/functionality mentions
-        feature_patterns = [
-            r"(?:add|create|build|implement)\s+(?:a\s+)?([a-z\s]+?)(?:\s+for|\s+in|\s+with|$)",
-            r"feature[:\s]+([a-z\s]+?)(?:\s+for|\s+in|\s+with|$)",
-        ]
-        for pattern in feature_patterns:
-            match = re.search(pattern, text.lower())
-            if match:
-                parameters["feature"] = match.group(1).strip()
-                break
-
-        return parameters
-
-    def get_agent_sequence(self, intent: Intent) -> list[str]:
-        """
-        Get the sequence of agents for an intent.
-
-        Args:
-            intent: Parsed intent
-
-        Returns:
-            List of agent names in execution order
-        """
-        return intent.get_agent_sequence()
-
+"""
+Intent Parser - Parse natural language input into structured intents.
+
+Maps user commands to agent sequences using keyword matching and
+simple NLP patterns.
+"""
+
+import re
+from dataclasses import dataclass
+from enum import Enum
+from typing import Any
+
+
+class IntentType(Enum):
+    """Types of user intents."""
+
+    BUILD = "build"
+    VALIDATE = "validate"  # Validation workflow for comparing implementations
+    REVIEW = "review"
+    FIX = "fix"
+    TEST = "test"
+    EPIC = "epic"
+    EXPLORE = "explore"
+    REFACTOR = "refactor"
+    PLAN_ANALYSIS = "plan-analysis"
+    PR = "pr"
+    REQUIREMENTS = "requirements"
+    BROWNFIELD = "brownfield"
+    ENHANCE = "enhance"  # §3.4: first-class *enhance
+    BREAKDOWN = "breakdown"  # §3.4: first-class *breakdown
+    TODO = "todo"  # §3.5: *todo (Beads-backed)
+    UNKNOWN = "unknown"
+
+
+@dataclass
+class Intent:
+    """Structured intent parsed from user input."""
+
+    type: IntentType
+    confidence: float
+    parameters: dict[str, Any]
+    original_input: str
+    compare_to_codebase: bool = False  # Flag for "compare to codebase" intent
+
+    def get_agent_sequence(self) -> list[str]:
+        """Get the sequence of agents for this intent."""
+        if self.type == IntentType.BUILD:
+            return ["planner", "architect", "designer", "implementer"]
+        elif self.type == IntentType.REVIEW:
+            return ["reviewer", "improver"]
+        elif self.type == IntentType.FIX:
+            return ["debugger", "implementer", "tester"]
+        elif self.type == IntentType.TEST:
+            return ["tester"]
+        elif self.type == IntentType.EPIC:
+            return ["epic-orchestrator"]
+        elif self.type == IntentType.EXPLORE:
+            return ["analyst", "reviewer"]
+        elif self.type == IntentType.REFACTOR:
+            return ["reviewer", "architect", "implementer", "tester"]
+        elif self.type == IntentType.PLAN_ANALYSIS:
+            return ["analyst", "architect", "reviewer"]
+        elif self.type == IntentType.PR:
+            return ["reviewer", "documenter"]
+        elif self.type == IntentType.REQUIREMENTS:
+            return ["analyst", "planner", "documenter"]
+        elif self.type == IntentType.BROWNFIELD:
+            return ["brownfield-review"]
+        elif self.type == IntentType.ENHANCE:
+            return ["enhancer"]
+        elif self.type == IntentType.BREAKDOWN:
+            return ["planner"]
+        elif self.type == IntentType.TODO:
+            return ["beads"]
+        else:
+            return []
+
+
+class IntentParser:
+    """Parse natural language input into structured intents."""
+
+    def __init__(self):
+        """Initialize the intent parser with keyword mappings."""
+        # Build intent keywords
+        self.build_keywords = [
+            "build",
+            "create",
+            "make",
+            "generate",
+            "add",
+            "implement",
+            "develop",
+            "write",
+            "new",
+            "feature",
+        ]
+
+        # Review intent keywords
+        self.review_keywords = [
+            "review",
+            "check",
+            "analyze",
+            "inspect",
+            "examine",
+            "score",
+            "quality",
+            "audit",
+            "assess",
+            "evaluate",
+            "compare",
+            "compare to",
+            "match",
+            "align with",
+            "follow patterns",
+        ]
+
+        # Fix intent keywords
+        self.fix_keywords = [
+            "fix",
+            "repair",
+            "resolve",
+            "debug",
+            "error",
+            "bug",
+            "issue",
+            "problem",
+            "broken",
+            "correct",
+        ]
+
+        # Test intent keywords
+        self.test_keywords = [
+            "test",
+            "verify",
+            "validate",
+            "coverage",
+            "testing",
+            "tests",
+            "unit test",
+            "integration test",
+        ]
+
+        # Epic intent keywords
+        self.epic_keywords = [
+            "epic",
+            "implement epic",
+            "execute epic",
+            "run epic",
+            "story",
+            "stories",
+        ]
+
+        # Explore intent keywords
+        self.explore_keywords = [
+            "explore",
+            "understand",
+            "navigate",
+            "find",
+            "discover",
+            "overview",
+            "codebase",
+            "trace",
+            "search",
+            "locate",
+        ]
+
+        # Refactor intent keywords
+        self.refactor_keywords = [
+            "refactor",
+            "modernize",
+            "update",
+            "improve code",
+            "modernize code",
+            "legacy",
+            "deprecated",
+            "upgrade code",
+        ]
+
+        # Plan Analysis intent keywords
+        self.plan_analysis_keywords = [
+            "plan",
+            "planning",
+            "analyze",
+            "analysis",
+            "design",
+            "proposal",
+            "strategy",
+            "roadmap",
+        ]
+
+        # PR intent keywords
+        self.pr_keywords = [
+            "pr",
+            "pull request",
+            "create pr",
+            "open pr",
+            "merge request",
+            "mr",
+        ]
+
+        # Requirements intent keywords
+        self.requirements_keywords = [
+            "requirements",
+            "gather requirements",
+            "extract requirements",
+            "document requirements",
+            "analyze requirements",
+            "requirements document",
+            "requirements gathering",
+            "requirements analysis",
+        ]
+
+        # Brownfield intent keywords
+        self.brownfield_keywords = [
+            "brownfield",
+            "brownfield review",
+            "review brownfield",
+            "analyze brownfield",
+            "brownfield system",
+            "review system",
+            "analyze project",
+            "create experts",
+            "auto experts",
+            "expert creation",
+            "populate rag",
+            "fill rag",
+        ]
+
+        # Simple Mode intent keywords
+        self.simple_mode_keywords = [
+            "@simple-mode",
+            "simple mode",
+            "use simple mode",
+            "simple-mode",
+            "@simple_mode",
+            "simple_mode",
+        ]
+
+    def detect_simple_mode_intent(self, text: str) -> bool:
+        """
+        Detect if user wants Simple Mode.
+
+        Args:
+            text: User input text
+
+        Returns:
+            True if Simple Mode intent detected, False otherwise
+
+        Examples:
+            >>> parser = IntentParser()
+            >>> parser.detect_simple_mode_intent("@simple-mode build feature")
+            True
+            >>> parser.detect_simple_mode_intent("use simple mode to create api")
+            True
+            >>> parser.detect_simple_mode_intent("build feature")
+            False
+        """
+        text_lower = text.lower()
+        return any(keyword in text_lower for keyword in self.simple_mode_keywords)
+
+    def parse(self, input_text: str) -> Intent:
+        """
+        Parse natural language input into a structured intent.
+
+        Args:
+            input_text: User's natural language command
+
+        Returns:
+            Intent object with type, confidence, and parameters
+        """
+        input_lower = input_text.lower().strip()
+
+        # Extract parameters
+        parameters = self._extract_parameters(input_text)
+
+        # Check for Simple Mode intent first
+        if self.detect_simple_mode_intent(input_text):
+            # Force Simple Mode workflow
+            parameters["force_simple_mode"] = True
+
+        # Check for explicit command patterns
+        if input_text.strip().startswith("*epic") or input_text.strip().startswith("@simple-mode *epic"):
+            # Extract epic path
+            epic_path = input_text.replace("*epic", "").replace("@simple-mode", "").strip()
+            if epic_path:
+                parameters["epic_path"] = epic_path
+            return Intent(
+                type=IntentType.EPIC,
+                confidence=1.0,
+                parameters=parameters,
+                original_input=input_text,
+            )
+
+        if input_text.strip().startswith("*explore") or input_text.strip().startswith("@simple-mode *explore"):
+            return Intent(
+                type=IntentType.EXPLORE,
+                confidence=1.0,
+                parameters=parameters,
+                original_input=input_text,
+            )
+
+        if input_text.strip().startswith("*validate") or input_text.strip().startswith("@simple-mode *validate"):
+            return Intent(
+                type=IntentType.VALIDATE,
+                confidence=1.0,
+                parameters=parameters,
+                original_input=input_text,
+            )
+
+        if input_text.strip().startswith("*refactor") or input_text.strip().startswith("@simple-mode *refactor"):
+            return Intent(
+                type=IntentType.REFACTOR,
+                confidence=1.0,
+                parameters=parameters,
+                original_input=input_text,
+            )
+
+        if input_text.strip().startswith("*plan") or input_text.strip().startswith("@simple-mode *plan") or "plan-analysis" in input_lower:
+            return Intent(
+                type=IntentType.PLAN_ANALYSIS,
+                confidence=1.0,
+                parameters=parameters,
+                original_input=input_text,
+            )
+
+        if input_text.strip().startswith("*pr") or input_text.strip().startswith("@simple-mode *pr") or "pull request" in input_lower:
+            return Intent(
+                type=IntentType.PR,
+                confidence=1.0,
+                parameters=parameters,
+                original_input=input_text,
+            )
+
+        if input_text.strip().startswith("*brownfield") or input_text.strip().startswith("*brownfield-review") or input_text.strip().startswith("@simple-mode *brownfield"):
+            return Intent(
+                type=IntentType.BROWNFIELD,
+                confidence=1.0,
+                parameters=parameters,
+                original_input=input_text,
+            )
+
+        # *enhance and *breakdown (§3.4)
+        if input_text.strip().startswith("*enhance") or input_text.strip().startswith("@simple-mode *enhance"):
+            prompt = (parameters.get("description") or re.sub(r"^(@simple-mode\s*)?\*enhance\s*", "", input_text, flags=re.I).strip() or input_text)
+            parameters["prompt"] = prompt
+            return Intent(type=IntentType.ENHANCE, confidence=1.0, parameters=parameters, original_input=input_text)
+        if input_text.strip().startswith("*breakdown") or input_text.strip().startswith("@simple-mode *breakdown"):
+            prompt = (parameters.get("description") or re.sub(r"^(@simple-mode\s*)?\*breakdown\s*", "", input_text, flags=re.I).strip() or input_text)
+            parameters["prompt"] = prompt
+            return Intent(type=IntentType.BREAKDOWN, confidence=1.0, parameters=parameters, original_input=input_text)
+
+        # *todo (§3.5) – remainder forwarded to bd
+        if input_text.strip().startswith("*todo") or input_text.strip().startswith("@simple-mode *todo"):
+            rest = re.sub(r"^(@simple-mode\s*)?\*todo\s*", "", input_text.strip(), flags=re.I).strip()
+            parameters["todo_rest"] = rest
+            return Intent(type=IntentType.TODO, confidence=1.0, parameters=parameters, original_input=input_text)
+
+        # Score each intent type
+        scores = {
+            IntentType.BUILD: self._score_intent(input_lower, self.build_keywords),
+            IntentType.REVIEW: self._score_intent(input_lower, self.review_keywords),
+            IntentType.FIX: self._score_intent(input_lower, self.fix_keywords),
+            IntentType.TEST: self._score_intent(input_lower, self.test_keywords),
+            IntentType.EPIC: self._score_intent(input_lower, self.epic_keywords),
+            IntentType.EXPLORE: self._score_intent(input_lower, self.explore_keywords),
+            IntentType.REFACTOR: self._score_intent(input_lower, self.refactor_keywords),
+            IntentType.PLAN_ANALYSIS: self._score_intent(input_lower, self.plan_analysis_keywords),
+            IntentType.PR: self._score_intent(input_lower, self.pr_keywords),
+            IntentType.REQUIREMENTS: self._score_intent(input_lower, self.requirements_keywords),
+            IntentType.BROWNFIELD: self._score_intent(input_lower, self.brownfield_keywords),
+        }
+
+        # Find best match
+        best_intent = max(scores.items(), key=lambda x: x[1])
+        intent_type, confidence = best_intent
+
+        # If confidence is too low, mark as unknown
+        if confidence < 0.3:
+            intent_type = IntentType.UNKNOWN
+            confidence = 0.0
+
+        # Detect "compare to codebase" intent with enhanced semantics
+        compare_to_codebase = False
+        compare_phrases = [
+            # Direct compare phrases
+            "compare to",
+            "compare with",
+            "compare against",
+            "compare this to",
+            "compare this with",
+            "compare to codebase",
+            "compare to our",
+            "compare to project",
+            # Match/align phrases
+            "match our",
+            "match the",
+            "match patterns",
+            "match codebase",
+            "align with",
+            "align to",
+            "follow patterns",
+            "follow our patterns",
+            "follow project patterns",
+            # Make match phrases (implicit compare)
+            "make match",
+            "make this match",
+            "make it match",
+            # Consistency phrases
+            "consistent with",
+            "consistency with",
+            "conform to",
+            "conform with",
+            # Standard phrases
+            "match standards",
+            "follow standards",
+            "meet standards",
+            "adhere to",
+        ]
+        if any(phrase in input_lower for phrase in compare_phrases):
+            compare_to_codebase = True
+
+        return Intent(
+            type=intent_type,
+            confidence=confidence,
+            parameters=parameters,
+            original_input=input_text,
+            compare_to_codebase=compare_to_codebase,
+        )
+
+    def _score_intent(self, text: str, keywords: list[str]) -> float:
+        """
+        Score how well text matches an intent based on keywords.
+
+        Args:
+            text: Lowercase input text
+            keywords: List of keywords for this intent
+
+        Returns:
+            Confidence score between 0.0 and 1.0
+        """
+        matches = sum(1 for keyword in keywords if keyword in text)
+        if not matches:
+            return 0.0
+
+        # Normalize by number of keywords (max score is 1.0)
+        score = min(matches / len(keywords) * 2.0, 1.0)
+
+        # Boost score if keyword appears at start of sentence
+        for keyword in keywords:
+            if text.startswith(keyword):
+                score = min(score + 0.2, 1.0)
+                break
+
+        return score
+
+    def _extract_parameters(self, text: str) -> dict[str, Any]:
+        """
+        Extract parameters from user input.
+
+        Args:
+            text: User input text
+
+        Returns:
+            Dictionary of extracted parameters
+        """
+        parameters: dict[str, Any] = {}
+
+        # Extract file paths (quoted strings or common file extensions)
+        file_pattern = r'["\']?([^\s"\']+\.(py|ts|js|tsx|jsx|java|go|rs|rb|php|cs|sql|yaml|yml|json|md|txt))["\']?'
+        file_matches = re.findall(file_pattern, text, re.IGNORECASE)
+        if file_matches:
+            parameters["files"] = [match[0] for match in file_matches]
+
+        # Extract quoted descriptions
+        quoted_pattern = r'["\']([^"\']+)["\']'
+        quoted_matches = re.findall(quoted_pattern, text)
+        if quoted_matches:
+            parameters["description"] = quoted_matches[0]
+
+        # Extract Epic document paths (epic-*.md or docs/prd/epic-*.md)
+        epic_pattern = r'(?:epic-[\w-]+\.md|docs/prd/epic-[\w-]+\.md|[\w/]+epic[\w-]*\.md)'
+        epic_matches = re.findall(epic_pattern, text, re.IGNORECASE)
+        if epic_matches:
+            parameters["epic_path"] = epic_matches[0]
+
+        # Extract feature/functionality mentions
+        feature_patterns = [
+            r"(?:add|create|build|implement)\s+(?:a\s+)?([a-z\s]+?)(?:\s+for|\s+in|\s+with|$)",
+            r"feature[:\s]+([a-z\s]+?)(?:\s+for|\s+in|\s+with|$)",
+        ]
+        for pattern in feature_patterns:
+            match = re.search(pattern, text.lower())
+            if match:
+                parameters["feature"] = match.group(1).strip()
+                break
+
+        return parameters
+
+    def get_agent_sequence(self, intent: Intent) -> list[str]:
+        """
+        Get the sequence of agents for an intent.
+
+        Args:
+            intent: Parsed intent
+
+        Returns:
+            List of agent names in execution order
+        """
+        return intent.get_agent_sequence()
+