enhance PR content extraction and issue linking (#26)

* arragne logger.py

* Add comments to llm_integration/initial_load.py

* refactor: Remove torch dependency for device detection

- Replace PyTorch-based device detection with a lightweight custom implementation
- Add support for CUDA, MPS (Apple Silicon), and CPU detection using standard libraries
- Improve portability and reduce overall package size

* feat(pr-analysis): enhance PR content extraction and issue linking

- Improve issue number extraction with comprehensive regex patterns
- Add support for GitHub's official autolink reference formats
- Implement commit details extraction in PR content
- Add validation for issue numbers to prevent invalid references
- Support multiple issue reference formats (URL, #, keywords)

This commit enhances the PR analysis capabilities by implementing
a more robust issue linking system and detailed commit information
extraction, following GitHub's official documentation and best practices.

* remove torch

* Update extract_related_issues function

---------

Co-authored-by: Jet Xu <[email protected]>

Author: JetXu-LLM

CHANGELOG.md

@@ -5,6 +5,18 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.1.6] - 2024-10-30
+
+### New Features
+- Enhanced PR content analysis with detailed commit information extraction
+- Improved issue linking detection with support for multiple reference formats
+  - Full GitHub URLs, #references, and keyword-based references
+  - Added validation for issue numbers
+
+### Improvements
+- Added detailed commit metadata extraction including stats and file changes
+- Enhanced error handling for commit fetching
+
 ## [0.1.5] - 2024-10-14
 
 ### Optimized

llama_github/data_retrieval/github_entities.py

@@ -241,25 +241,79 @@ def get_issue_content(self, number, issue=None) -> str:
 
     def extract_related_issues(self, pr_data: Dict[str, Any]) -> List[int]:
         """
-        Extracts related issue numbers from the PR description and other fields.
-
-        :param pr_data: The pull request data dictionary.
-        :return: A list of related issue numbers.
+        Extracts related issue numbers from all PR data following GitHub's reference syntax.
+        
+        This function implements GitHub's official autolink reference formats to find:
+        1. Full GitHub issue/PR URLs
+        2. Numeric references (#123)
+        3. Keywords + issue references (fixes #123)
+        4. Repository cross-references (owner/repo#123)
+        
+        See: https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/autolinked-references-and-urls
+        
+        Args:
+            pr_data: Dict[str, Any] - The complete pull request data dictionary
+            
+        Returns:
+            List[int] - A sorted list of unique issue numbers found in the PR data
         """
+        # GitHub's official closing keywords
+        closing_keywords = (
+            'close', 'closes', 'closed',
+            'fix', 'fixes', 'fixed',
+            'resolve', 'resolves', 'resolved'
+        )
+
+        # Regex patterns for GitHub issue references
         patterns = [
-            rf'https://github\.com/{re.escape(self.full_name)}/issues/(\d+)',
-            r'(?:^|\s)#(\d+)',
-            r'(?:^|\s)(\d+)(?:\s|$)',
+            # Full GitHub issue/PR URL pattern
+            rf'(?:https?://)?github\.com/{re.escape(self.full_name)}/(?:issues|pull)/(\d+)',
+            
+            # Standard #123 reference with proper boundaries
+            r'(?:^|[^\w/])#(\d+)(?=[^\w/]|$)',
+            
+            # Closing keywords (fixes #123)
+            fr'(?:^|[^\w/])(?:{"|".join(closing_keywords)}):?\s+#(\d+)(?=[^\w/]|$)',
+            
+            # Cross-repo reference (owner/repo#123)
+            rf'{re.escape(self.full_name)}#(\d+)',
+            
+            # Issue keyword reference (issue #123 or issue: #123)
+            r'(?:^|[^\w/])(?:issue|bug|ticket|todo|task)s?:?\s+#?(\d+)(?=[^\w/]|$)'
         ]
+
         issues = set()
-        # Convert PR data to JSON string for pattern matching
-        pr_description = json.dumps(pr_data, default=str)
 
-        for pattern in patterns:
-            matches = re.findall(pattern, pr_description)
-            issues.update(int(match) for match in matches)
+        def extract_from_text(text: str) -> None:
+            """Helper function to extract issue numbers from text"""
+            if not isinstance(text, str):
+                return
+                
+            for pattern in patterns:
+                matches = re.findall(pattern, text, re.IGNORECASE | re.MULTILINE)
+                # Validate issue numbers (reasonable length and positive values)
+                valid_matches = [
+                    int(match) for match in matches 
+                    if match.isdigit() and len(match) <= 7 and int(match) > 0
+                ]
+                issues.update(valid_matches)
+
+        def process_value(value: Any) -> None:
+            """Recursively process dictionary values and extract issue numbers"""
+            if isinstance(value, dict):
+                for v in value.values():
+                    process_value(v)
+            elif isinstance(value, (list, tuple)):
+                for item in value:
+                    process_value(item)
+            elif isinstance(value, str):
+                extract_from_text(value)
+
+        # Process all data in pr_data recursively
+        process_value(pr_data)
 
-        return list(issues)
+        return sorted(list(issues))
+
 
     def get_issue_contents(self, issue_numbers: List[int], pr_number: int) -> List[Dict[str, Any]]:
         """
@@ -324,11 +378,34 @@ def get_pr_content(self, number, pr=None, context_lines=10) -> Dict[str, Any]:
                                 "head_branch": pr.head.ref,
                             },
                             "related_issues": [],
+                            "commits": [],
                             "file_changes": [],
                             "ci_cd_results": [],
                             "interactions": []
                         }
 
+                        # Fetch and process commits
+                        try:
+                            commits = pr.get_commits()
+                            for commit in commits:
+                                commit_data = {
+                                    "sha": commit.sha,
+                                    "message": commit.commit.message,
+                                    "author": commit.commit.author.name,
+                                    "date": self.to_isoformat(commit.commit.author.date),
+                                    "stats": {
+                                        "additions": commit.stats.additions,
+                                        "deletions": commit.stats.deletions,
+                                        "total": commit.stats.total
+                                    },
+                                    "files": [f.filename for f in commit.files]  # Just keep changed file names
+                                }
+                                pr_data["commits"].append(commit_data)
+                        except GithubException as e:
+                            logger.exception(f"Error fetching commits for PR #{number}")
+                            pr_data["commits"] = []
+                            pr_data["commit_stats"] = {}
+
                         # Fetch CI/CD results
                         try:
                             last_commit = pr.get_commits().reversed[0]

llama_github/version.py

@@ -1 +1 @@
-__version__ = '0.1.5'
+__version__ = '0.1.6'

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = llama-github
-version = 0.1.5
+version = 0.1.6
 author = Jet Xu
 author_email = [email protected]
 description = Llama-github is an open-source Python library that empowers LLM Chatbots, AI Agents, and Auto-dev Agents to conduct Retrieval from actively selected GitHub public projects. It Augments through LLMs and Generates context for any coding question, in order to streamline the development of sophisticated AI-driven applications.