From 17b7e4674cb6e138d5e38a5d696b177a0e272999 Mon Sep 17 00:00:00 2001
From: Hrishikesh Uchake <uchakeh@gmail.com>
Date: Sat, 24 Jan 2026 18:18:49 -0600
Subject: [PATCH] Implement semantic tech-stack matching for GitHub projects
 selection

---
 app.py                     | 10 ++++++-
 app_production.py          | 22 +++++++++++++-
 scrapers/github_scraper.py | 60 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 90 insertions(+), 2 deletions(-)

diff --git a/app.py b/app.py
index 4c683db..c01f783 100644
--- a/app.py
+++ b/app.py
@@ -319,7 +319,15 @@ def generate_resume():
         # Step 2: Analyze job description
         job_requirements = job_analyzer.analyze(job_description)
 
-        # Step 3: Generate tailored resume content
+        # Step 3: Refine GitHub projects based on job requirements
+        if "github" in profile_data and "repositories" in profile_data["github"]:
+            job_skills = job_requirements.get("skills", {})
+            relevant_projects = github_scraper.select_relevant_projects(
+                profile_data["github"]["repositories"], job_skills
+            )
+            profile_data["github"]["top_projects"] = relevant_projects
+
+        # Step 4: Generate tailored resume content
         resume_content = resume_generator.generate(profile_data, job_requirements)
 
         # Step 4: Compile to PDF
diff --git a/app_production.py b/app_production.py
index cbfb654..d2d4e39 100644
--- a/app_production.py
+++ b/app_production.py
@@ -492,7 +492,27 @@ def generate_resume():
             logger.error(f"[{request.request_id}] Job analysis failed: {e}")
             raise ResumeGenerationError("Failed to analyze job description")
 
-        # Step 3: Generate tailored resume content
+        # Step 3: Refine GitHub projects based on job requirements
+        if "github" in profile_data and "repositories" in profile_data["github"]:
+            try:
+                logger.debug(
+                    f"[{request.request_id}] Refining GitHub projects by relevance"
+                )
+                job_skills = job_requirements.get("skills", {})
+                relevant_projects = github_scraper.select_relevant_projects(
+                    profile_data["github"]["repositories"], job_skills
+                )
+                profile_data["github"]["top_projects"] = relevant_projects
+                logger.debug(
+                    f"[{request.request_id}] Refined to {len(relevant_projects)} relevant projects"
+                )
+            except Exception as e:
+                logger.warning(
+                    f"[{request.request_id}] GitHub project refinement failed: {e}"
+                )
+                # Continue with default projects if refinement fails
+
+        # Step 4: Generate tailored resume content
         try:
             logger.debug(f"[{request.request_id}] Generating resume content")
             resume_content = resume_generator.generate(profile_data, job_requirements)
diff --git a/scrapers/github_scraper.py b/scrapers/github_scraper.py
index dcf5d15..acda815 100644
--- a/scrapers/github_scraper.py
+++ b/scrapers/github_scraper.py
@@ -144,3 +144,63 @@ def scrape_profile(self, username: str) -> Dict:
                 exc_info=True,
             )
             raise GitHubAPIError(f"Failed to scrape GitHub profile: {str(e)}")
+
+    def select_relevant_projects(
+        self, repos: List[Dict], job_skills: Dict[str, List[str]]
+    ) -> List[Dict]:
+        """
+        Select projects that best match the job requirements
+
+        Args:
+            repos: List of repositories
+            job_skills: Dictionary of required skills by category
+
+        Returns:
+            List of relevant repositories sorted by relevance
+        """
+        if not repos:
+            return []
+
+        # Flatten all job skills into a single set for easy matching
+        all_required_skills = set()
+        for skills in job_skills.values():
+            if isinstance(skills, list):
+                all_required_skills.update([s.lower() for s in skills])
+
+        if not all_required_skills:
+            # Fallback to stars if no skills identified
+            return sorted(repos, key=lambda x: x.get("stars", 0), reverse=True)[
+                : config.GITHUB_MAX_TOP_PROJECTS
+            ]
+
+        scored_repos = []
+        for repo in repos:
+            score = 0
+            repo_name = repo.get("name", "").lower()
+            repo_desc = (repo.get("description") or "").lower()
+            repo_lang = (repo.get("language") or "").lower()
+            repo_topics = [t.lower() for t in repo.get("topics", [])]
+
+            # Match language (High weight)
+            if repo_lang in all_required_skills:
+                score += 5
+
+            # Match topics (Medium weight)
+            for topic in repo_topics:
+                if topic in all_required_skills:
+                    score += 3
+
+            # Match keywords in name/description (Low weight)
+            for skill in all_required_skills:
+                if skill in repo_name or skill in repo_desc:
+                    score += 1
+
+            scored_repos.append((score, repo))
+
+        # Sort by score (desc), then stars (desc) as tie-breaker
+        sorted_repos = sorted(
+            scored_repos, key=lambda x: (x[0], x[1].get("stars", 0)), reverse=True
+        )
+
+        # Return only the repository dictionaries
+        return [repo for score, repo in sorted_repos[: config.GITHUB_MAX_TOP_PROJECTS]]