AI Research MCP Server

"""GitHub client for searching AI/ML repositories.""" import os from datetime import datetime, timedelta, timezone from typing import List, Dict, Optional from github import Github, GithubException class GithubClient: """Client for searching GitHub repositories.""" # Keywords by AI research area AI_KEYWORDS = { "core_ai": ["LLM", "transformer", "diffusion", "GPT", "neural network", "deep learning"], "multimodal": ["CLIP", "stable diffusion", "text-to-image", "vision-language"], "applications": ["RAG", "AI agent", "langchain", "prompt engineering", "RLHF"], "infrastructure": ["pytorch", "tensorflow", "vLLM", "model optimization"], "robotics": ["robotics", "robot learning", "embodied AI", "manipulation", "navigation"], "bioinfo": ["bioinformatics", "protein folding", "alphafold", "drug discovery", "genomics"], "science": ["AI4Science", "scientific machine learning", "physics-informed neural networks"], "rl": ["reinforcement learning", "multi-agent", "game AI", "AlphaGo"], "graph": ["graph neural network", "GNN", "molecular modeling"], "recsys": ["recommender systems", "personalization"], "timeseries": ["time series forecasting", "anomaly detection"], "emerging": ["federated learning", "neuromorphic computing", "quantum machine learning"], } # Common topics for AI repositories AI_TOPICS = [ "llm", "gpt", "transformer", "deep-learning", "machine-learning", "computer-vision", "nlp", "reinforcement-learning", "generative-ai", "diffusion", "pytorch", "tensorflow", "huggingface", "ai-agent", "robotics", "bioinformatics", "graph-neural-network", ] def __init__(self, token: Optional[str] = None): """Initialize GitHub client. Args: token: GitHub personal access token (optional but recommended) """ token = token or os.getenv("GITHUB_TOKEN") self.github = Github(token) if token else Github() def search_repositories( self, keywords: Optional[List[str]] = None, topics: Optional[List[str]] = None, min_stars: int = 50, language: Optional[str] = None, created_since: Optional[str] = None, pushed_since: Optional[str] = None, sort_by: str = "stars", max_results: int = 50, ) -> List[Dict]: """Search for repositories. Args: keywords: Keywords to search in name/description/README topics: Repository topics to filter by min_stars: Minimum number of stars language: Programming language filter created_since: Created since (e.g., '7d', '1w', '1m') pushed_since: Last pushed since (e.g., '7d', '1w', '1m') sort_by: Sort by 'stars', 'forks', 'updated', or 'best-match' max_results: Maximum number of results Returns: List of repository dictionaries """ # Build query query_parts = [] # Add keywords if keywords: keyword_query = " ".join(keywords) query_parts.append(keyword_query) # Add topics if topics: for topic in topics: query_parts.append(f"topic:{topic}") # Add stars filter query_parts.append(f"stars:>={min_stars}") # Add language filter if language: query_parts.append(f"language:{language}") # Add date filters if created_since: date = self._parse_relative_date(created_since) query_parts.append(f"created:>={date}") if pushed_since: date = self._parse_relative_date(pushed_since) query_parts.append(f"pushed:>={date}") query = " ".join(query_parts) # Search repositories try: repositories = self.github.search_repositories( query=query, sort=sort_by, order="desc", ) results = [] count = 0 for repo in repositories: if count >= max_results: break results.append({ "name": repo.name, "full_name": repo.full_name, "description": repo.description, "url": repo.html_url, "stars": repo.stargazers_count, "forks": repo.forks_count, "language": repo.language, "topics": repo.topics if hasattr(repo, "topics") else [], "created_at": repo.created_at.isoformat(), "updated_at": repo.updated_at.isoformat(), "pushed_at": repo.pushed_at.isoformat() if repo.pushed_at else None, "homepage": repo.homepage, "license": repo.license.name if repo.license else None, "source": "github", }) count += 1 return results except GithubException as e: print(f"GitHub API error: {e}") return [] def get_trending_repositories( self, period: str = "daily", language: Optional[str] = None, max_results: int = 25, ) -> List[Dict]: """Get trending repositories. Note: GitHub API doesn't have official trending endpoint, so we approximate by searching for recently created/updated repos with high stars. Args: period: 'daily', 'weekly', or 'monthly' language: Programming language filter max_results: Maximum number of results Returns: List of repository dictionaries """ # Map period to days period_days = { "daily": 1, "weekly": 7, "monthly": 30, } days = period_days.get(period, 7) # Search for recently starred repos with AI topics return self.search_repositories( topics=self.AI_TOPICS[:5], # Use top 5 most common topics min_stars=100, language=language, pushed_since=f"{days}d", sort_by="stars", max_results=max_results, ) def search_by_area( self, area: str, min_stars: int = 50, days: int = 30, max_results: int = 25, ) -> List[Dict]: """Search repositories by research area. Args: area: Research area (e.g., 'llm', 'robotics', 'bioinfo') min_stars: Minimum number of stars days: Look back this many days max_results: Maximum number of results Returns: List of repository dictionaries """ keywords = self.AI_KEYWORDS.get(area.lower()) if not keywords: raise ValueError(f"Unknown area: {area}. Valid areas: {list(self.AI_KEYWORDS.keys())}") return self.search_repositories( keywords=keywords[:3], # Use top 3 keywords to avoid too restrictive search min_stars=min_stars, pushed_since=f"{days}d", sort_by="stars", max_results=max_results, ) def _parse_relative_date(self, date_str: str) -> str: """Parse relative date string to ISO date. Args: date_str: Relative date (e.g., '7d', '1w', '1m') Returns: ISO format date string """ unit = date_str[-1] value = int(date_str[:-1]) if unit == "d": delta = timedelta(days=value) elif unit == "w": delta = timedelta(weeks=value) elif unit == "m": delta = timedelta(days=value * 30) else: raise ValueError(f"Invalid date format: {date_str}. Use format like '7d', '1w', '1m'") date = datetime.now(timezone.utc) - delta return date.strftime("%Y-%m-%d")