#!/usr/bin/env python3
"""Demo of enhanced semantic indexer capabilities."""
from pathlib import Path
from mcp_server.utils.semantic_indexer import SemanticIndexer
def demo_enhanced_features():
"""Demonstrate the enhanced indexer features."""
print("=" * 60)
print("ENHANCED SEMANTIC INDEXER DEMONSTRATION")
print("=" * 60)
print("\n1. DOCUMENT TYPE WEIGHTS")
print("-" * 40)
indexer = SemanticIndexer.__new__(SemanticIndexer) # Create without API initialization
for doc_type, weight in indexer.DOCUMENT_TYPE_WEIGHTS.items():
print(f" {doc_type:12} -> {weight:4.2f}x weight")
print("\n2. DOCUMENT SECTION STRUCTURE")
print("-" * 40)
# Demo markdown content
demo_content = """# API Documentation
This is the main API documentation.
## Authentication
All requests require authentication.
### JWT Tokens
Use JWT tokens for authentication.
#### Token Format
Tokens should be in the format: Bearer <token>
### API Keys
Alternative authentication using API keys.
## Endpoints
Available API endpoints.
### Search
Search for code and documentation.
#### Basic Search
Simple text search functionality.
#### Advanced Search
Complex search with filters.
### Index
Index new content.
"""
sections = indexer._parse_markdown_sections(demo_content, "api.md")
def print_section_tree(sections, level=0):
"""Print sections in a tree structure."""
for section in sections:
if section.parent_section is None and level == 0:
indent = " " * level
print(f"{indent}📄 {section.title} (Level {section.level})")
print_subsections(section, sections, level + 1)
elif level > 0 and section.parent_section and level == 1:
# This is handled by print_subsections
pass
def print_subsections(parent, all_sections, level):
"""Print subsections recursively."""
for section in all_sections:
if section.parent_section == parent.title:
indent = " " * level
icon = "📁" if section.subsections else "📄"
print(f"{indent}{icon} {section.title} (Level {section.level})")
if section.subsections:
print_subsections(section, all_sections, level + 1)
# Build hierarchy display
print("Document hierarchy:")
root_sections = [s for s in sections if s.parent_section is None]
def print_hierarchy(sections, parent_title=None, level=0):
"""Print document hierarchy."""
for section in sections:
if section.parent_section == parent_title:
indent = " " * level
icon = "📁" if section.subsections else "📄"
print(f"{indent}{icon} {section.title}")
print_hierarchy(sections, section.title, level + 1)
print_hierarchy(sections)
print("\n3. CONTEXT-AWARE EMBEDDING PREPARATION")
print("-" * 40)
# Show how embeddings would be enriched
sample_section = sections[3] if len(sections) > 3 else sections[0] # JWT Tokens section
print(f"Original section content:")
print(f" Title: {sample_section.title}")
print(f" Content: {sample_section.content.strip()}")
print(f"\nContext-enriched embedding text:")
# Build context path
context_path = []
current_section = sample_section
# Find parent chain
parent_chain = []
temp_section = sample_section
while temp_section.parent_section:
parent_chain.append(temp_section.parent_section)
# Find the parent section object
for s in sections:
if s.title == temp_section.parent_section:
temp_section = s
break
else:
break
# Reverse to get root-to-leaf order
parent_chain.reverse()
context_str = " > ".join(parent_chain + [sample_section.title])
print(f" Document: API Documentation")
print(f" Section: {context_str}")
print(f" Content: {sample_section.content.strip()}")
print(f"\nFull embedding text would be:")
embedding_text = f"""Document: API Documentation
Section: {context_str}
{sample_section.content.strip()}"""
print(" " + embedding_text.replace("\n", "\n "))
print("\n4. NATURAL LANGUAGE QUERY TYPES")
print("-" * 40)
query_examples = [
("How do I authenticate API requests?", ["markdown", "api"], "Authentication guidance"),
("What are the available search endpoints?", ["api", "markdown"], "API endpoint documentation"),
("Show me JWT token examples", ["markdown", "tutorial"], "Code examples and tutorials"),
("Getting started with the API", ["readme", "tutorial"], "Introductory content"),
]
for query, doc_types, purpose in query_examples:
print(f"\nQuery: '{query}'")
print(f" Target doc types: {doc_types}")
print(f" Purpose: {purpose}")
# Show which weights would apply
weights = [indexer.DOCUMENT_TYPE_WEIGHTS.get(dt, 1.0) for dt in doc_types]
print(f" Weights: {[f'{dt}={w:.2f}' for dt, w in zip(doc_types, weights)]}")
print("\n5. SECTION-AWARE SEARCH BENEFITS")
print("-" * 40)
benefits = [
"🎯 Context preservation: Section titles and hierarchy inform embeddings",
"📚 Document type weighting: Documentation gets priority for doc searches",
"🔍 Natural language queries: Optimized for human-readable questions",
"🌳 Hierarchical structure: Parent-child relationships maintained",
"🏷️ Metadata enrichment: Tags, summaries, and context included",
"📊 Weighted scoring: Different content types have appropriate relevance",
]
for benefit in benefits:
print(f" {benefit}")
print("\n6. USAGE PATTERNS")
print("-" * 40)
usage_patterns = [
("Documentation Search", "query_natural_language()", ["markdown", "readme"]),
("API Reference Lookup", "query_natural_language()", ["api"]),
("Tutorial Content", "query_natural_language()", ["tutorial", "guide"]),
("Code + Docs Combined", "query_natural_language()", None),
("Pure Code Search", "search() or query()", ["code"]),
]
for pattern, method, doc_types in usage_patterns:
print(f"\n {pattern}:")
print(f" Method: {method}")
if doc_types:
print(f" Doc types: {doc_types}")
else:
print(f" Doc types: All types included")
print("\n" + "=" * 60)
print("The enhanced indexer provides sophisticated document")
print("understanding while maintaining efficient code search.")
print("=" * 60)
if __name__ == "__main__":
demo_enhanced_features()
