fix: refresh from latest project snapshot

Sync GitHub commit ad338a5; ship the validated 148-project snapshot/index so deployed refresh reuses existing embeddings.

hackathon_advisor/data.py

@@ -32,7 +32,6 @@ DEFAULT_EMBEDDING_MODEL_REPO = "ggml-org/embeddinggemma-300m-qat-q8_0-GGUF"
 DEFAULT_EMBEDDING_MODEL_FILE = "embeddinggemma-300m-qat-Q8_0.gguf"
 DEFAULT_EMBEDDING_RUNTIME = "llama.cpp via llama-cpp-python"
 APP_FILE_EMBEDDING_CHAR_LIMIT = 2000
-README_EMBEDDING_CHAR_LIMIT = 1200
 
 
 EmbeddingFunction = Callable[[str], Sequence[float]]
@@ -92,9 +91,6 @@ class Project:
                 f"title: {self.title}",
                 f"slug: {self.slug.replace('-', ' ').replace('_', ' ')}",
                 f"summary: {self.summary}",
-                f"readme:\n{bounded_embedding_text(self.readme_body, README_EMBEDDING_CHAR_LIMIT)}"
-                if self.readme_body
-                else "",
                 f"tags: {' '.join(self.tags)}",
                 f"models: {' '.join(self.models)}",
                 f"datasets: {' '.join(self.datasets)}",

tests/test_data.py

@@ -85,7 +85,7 @@ def test_searchable_text_includes_main_app_file_signals() -> None:
     assert "Project idea" in searchable
 
 
-def test_searchable_text_bounds_readme_body_for_embedding() -> None:
+def test_searchable_text_excludes_refresh_readme_body_for_stable_reuse() -> None:
     project = Project(
         id="build-small-hackathon/long-readme",
         title="Long README",
@@ -105,8 +105,8 @@ def test_searchable_text_bounds_readme_body_for_embedding() -> None:
 
     searchable = project.searchable_text
 
+    assert "readme:" not in searchable
     assert "middle should not be embedded" not in searchable
-    assert len(searchable) < 1600
 
 
 def test_project_index_rejects_mismatched_snapshot(tmp_path: Path) -> None: