fix bug

sweepai/agents/assistant_function_modify.py

@@ -208,7 +208,7 @@ def save_ticket_progress(assistant_id: str, thread_id: str, run_id: str):
                     if error_message:
                         logger.error(error_message)
                         tool_name, tool_call = assistant_generator.send(
-                            f"ERROR\nNo changes we're made due to the following error:\n\n{error_message}"
+                            f"ERROR\nNo changes were made due to the following error:\n\n{error_message}"
                         )
                     else:
                         logger.info(success_message)

sweepai/core/lexical_search.py

@@ -253,7 +253,7 @@ def prepare_index_from_docs(docs):
     index = CustomIndex()
     try:
         for doc in tqdm(all_docs, total=len(all_docs)):
-            index.add_document(title=f"{doc.url}", content=doc.content)
+            index.add_document(title=f"{doc.url}", tokens=compute_document_tokens(doc.content))
     except FileNotFoundError as e:
         logger.error(e)
     return index

sweepai/core/vector_db.py

@@ -189,7 +189,7 @@ def get_deeplake_vs_from_repo(
     )
     # scoring for vector search
     files_to_scores = compute_vector_search_scores(
-        file_list, cloned_repo, repo_full_name
+        file_list, cloned_repo
     )
 
     collection_name, documents, ids, metadatas = prepare_documents_metadata_ids(
@@ -226,7 +226,7 @@ def prepare_documents_metadata_ids(
     return collection_name, documents, ids, metadatas
 
 
-def compute_vector_search_scores(file_list, cloned_repo, repo_full_name):
+def compute_vector_search_scores(file_list, cloned_repo):
     files_to_scores = {}
     score_factors = []
     for file_path in tqdm(file_list):
@@ -254,9 +254,8 @@ def compute_vector_search_scores(file_list, cloned_repo, repo_full_name):
     # compute all scores
     all_scores = get_scores(score_factors)
     files_to_scores = {
-        file_path: score for file_path, score in zip(file_list, all_scores)
+        file_path[len(cloned_repo.cached_dir) + 1 :]: score for file_path, score in zip(file_list, all_scores)
     }
-    logger.info(f"Found {len(file_list)} files in repository {repo_full_name}")
     return files_to_scores
 
 

sweepai/utils/ticket_utils.py

@@ -8,7 +8,7 @@
 from sweepai.core.context_pruning import RepoContextManager, get_relevant_context
 from sweepai.core.entities import Snippet
 from sweepai.core.lexical_search import search_index
-from sweepai.core.vector_db import prepare_lexical_search_index
+from sweepai.core.vector_db import compute_vector_search_scores, prepare_lexical_search_index
 from sweepai.logn.cache import file_cache
 from sweepai.utils.chat_logger import discord_log_error
 from sweepai.utils.event_logger import posthog
@@ -42,12 +42,14 @@ def prep_snippets(
         lambda snippet: f"{snippet.file_path}:{snippet.start}:{snippet.end}"
     )
 
+    files_to_scores = compute_vector_search_scores(file_list, cloned_repo)
     for snippet in snippets:
+        codebase_score = files_to_scores.get(snippet.file_path, 0.08)
         snippet_score = 0.1
         if snippet_to_key(snippet) in content_to_lexical_score:
-            snippet_score = content_to_lexical_score[snippet_to_key(snippet)]
+            snippet_score = content_to_lexical_score[snippet_to_key(snippet)] * codebase_score
         else:
-            content_to_lexical_score[snippet_to_key(snippet)] = snippet_score
+            content_to_lexical_score[snippet_to_key(snippet)] = snippet_score * codebase_score
 
     ranked_snippets = sorted(
         snippets,