"""Command: recommend - Get skill recommendations for current project."""
from __future__ import annotations
from pathlib import Path
import click
from rich.table import Table
from mcp_skills.cli.shared.console import console
from mcp_skills.models.config import MCPSkillsConfig
from mcp_skills.services.indexing.engine import IndexingEngine
from mcp_skills.services.skill_manager import SkillManager
from mcp_skills.services.toolchain_detector import ToolchainDetector
@click.command()
@click.option(
"--search-mode",
type=click.Choice(
["semantic_focused", "graph_focused", "balanced", "current"],
case_sensitive=False,
),
help="Hybrid search weighting preset (overrides config file)",
)
def recommend(search_mode: str | None) -> None:
"""Get skill recommendations for current project.
Analyzes the current directory's toolchain (language, frameworks, tools)
and recommends relevant skills from the skill repository.
Search Modes:
- semantic_focused: Optimize for semantic similarity (90% vector, 10% graph)
- graph_focused: Optimize for relationships (30% vector, 70% graph)
- balanced: Equal weighting (50% vector, 50% graph)
- current: Default optimized preset (70% vector, 30% graph)
If --search-mode is not specified, loads from config.yaml or uses default.
Examples:
mcp-skillset recommend
mcp-skillset recommend --search-mode semantic_focused
"""
console.print("💡 [bold]Skill Recommendations[/bold]")
if search_mode:
console.print(f"⚖️ [dim]Search mode: {search_mode}[/dim]")
console.print()
try:
# Detect current directory toolchain
detector = ToolchainDetector()
current_dir = Path.cwd()
toolchain = detector.detect(current_dir)
# Display detected toolchain
console.print("[bold cyan]Detected Toolchain:[/bold cyan]")
console.print(f" • Language: {toolchain.primary_language}")
if toolchain.frameworks:
console.print(f" • Frameworks: {', '.join(toolchain.frameworks)}")
if toolchain.test_frameworks:
console.print(f" • Testing: {', '.join(toolchain.test_frameworks)}")
console.print(f" • Confidence: {toolchain.confidence:.0%}\n")
# Get recommendations with config
skill_manager = SkillManager()
# Load config and optionally override with CLI flag
config = MCPSkillsConfig()
if search_mode:
# Override config with CLI flag
config.hybrid_search = config._get_preset(search_mode)
console.print(
f"[dim]Using {search_mode} preset: "
f"vector={config.hybrid_search.vector_weight:.1f}, "
f"graph={config.hybrid_search.graph_weight:.1f}[/dim]\n"
)
indexing_engine = IndexingEngine(skill_manager=skill_manager, config=config)
# Build query from toolchain
query_parts = [toolchain.primary_language]
query_parts.extend(toolchain.frameworks[:2]) # Add top 2 frameworks
query = " ".join(query_parts)
# Search for relevant skills
results = indexing_engine.search(query, top_k=10)
if not results:
console.print("[yellow]No recommendations available[/yellow]")
console.print("\nTry:")
console.print(" • Running: mcp-skillset setup")
console.print(" • Adding repositories: mcp-skillset repo add <url>")
console.print(" • Rebuilding index: mcp-skillset index --force")
return
# Display recommendations
table = Table(title=f"Recommended Skills ({len(results)} found)")
table.add_column("Name", style="cyan", no_wrap=True)
table.add_column("Category", style="magenta")
table.add_column("Relevance", justify="right", style="green")
table.add_column("Description", style="dim")
for result in results:
skill = result.skill
relevance_str = f"{result.score:.2f}"
# Truncate description
desc = skill.description[:50]
if len(skill.description) > 50:
desc += "..."
table.add_row(
skill.name,
skill.category,
relevance_str,
desc,
)
console.print(table)
console.print("\n[dim]Use 'mcp-skillset info <skill-id>' for details[/dim]")
except Exception as e:
console.print(f"[red]Recommendations failed: {e}[/red]")
import logging
logger = logging.getLogger(__name__)
logger.exception("Recommendations failed")
raise SystemExit(1)
