Merge pull request #352 from Codium-ai/tr/prompts

Refactoring Patch Extra Lines and Updating Prompts

Usage.md

@@ -261,6 +261,30 @@ All PR-Agent tools have a parameter called `extra_instructions`, that enables to
 /update_changelog --pr_update_changelog.extra_instructions="Make sure to update also the version ..."
 ```
 
+#### Patch Extra Lines
+By default, around any change in your PR, git patch provides 3 lines of context above and below the change.
+```
+@@ -12,5 +12,5 @@ def func1():
+ code line that already existed in the file...
+ code line that already existed in the file...
+ code line that already existed in the file....
+-code line that was removed in the PR
++new code line added in the PR
+ code line that already existed in the file...
+ code line that already existed in the file...
+ code line that already existed in the file...
+```
+
+For the `review`, `describe`, `ask` and `add_docs` tools, if the token budget allows, PR-Agent tries to increase the number of lines of context, via the parameter:
+```
+[config]
+patch_extra_lines=3
+```
+
+Increasing this number provides more context to the model, but will also increase the token budget.
+If the PR is too large (see [PR Compression strategy](./PR_COMPRESSION.md)), PR-Agent automatically sets this number to 0, using the original git patch.
+
+
 #### Azure DevOps provider
 To use Azure DevOps provider use the following settings in configuration.toml:
 ```

pr_agent/algo/pr_processing.py

@@ -21,7 +21,6 @@ MORE_MODIFIED_FILES_ = "More modified files:\n"
 
 OUTPUT_BUFFER_TOKENS_SOFT_THRESHOLD = 1000
 OUTPUT_BUFFER_TOKENS_HARD_THRESHOLD = 600
-PATCH_EXTRA_LINES = 3
 
 def get_pr_diff(git_provider: GitProvider, token_handler: TokenHandler, model: str,
                 add_line_numbers_to_hunks: bool = False, disable_extra_lines: bool = False) -> str:
@@ -44,8 +43,9 @@ def get_pr_diff(git_provider: GitProvider, token_handler: TokenHandler, model: s
     """
 
     if disable_extra_lines:
-        global PATCH_EXTRA_LINES
         PATCH_EXTRA_LINES = 0
+    else:
+        PATCH_EXTRA_LINES = get_settings().config.patch_extra_lines
 
     try:
         diff_files = git_provider.get_diff_files()
@@ -57,8 +57,8 @@ def get_pr_diff(git_provider: GitProvider, token_handler: TokenHandler, model: s
     pr_languages = sort_files_by_main_languages(git_provider.get_languages(), diff_files)
 
     # generate a standard diff string, with patch extension
-    patches_extended, total_tokens, patches_extended_tokens = pr_generate_extended_diff(pr_languages, token_handler,
-                                                               add_line_numbers_to_hunks)
+    patches_extended, total_tokens, patches_extended_tokens = pr_generate_extended_diff(
+        pr_languages, token_handler, add_line_numbers_to_hunks, patch_extra_lines=PATCH_EXTRA_LINES)
 
     # if we are under the limit, return the full diff
     if total_tokens + OUTPUT_BUFFER_TOKENS_SOFT_THRESHOLD < MAX_TOKENS[model]:
@@ -80,7 +80,8 @@ def get_pr_diff(git_provider: GitProvider, token_handler: TokenHandler, model: s
 
 def pr_generate_extended_diff(pr_languages: list,
                               token_handler: TokenHandler,
-                              add_line_numbers_to_hunks: bool) -> Tuple[list, int, list]:
+                              add_line_numbers_to_hunks: bool,
+                              patch_extra_lines: int = 0) -> Tuple[list, int, list]:
     """
     Generate a standard diff string with patch extension, while counting the number of tokens used and applying diff
     minimization techniques if needed.
@@ -102,7 +103,7 @@ def pr_generate_extended_diff(pr_languages: list,
                 continue
 
             # extend each patch with extra lines of context
-            extended_patch = extend_patch(original_file_content_str, patch, num_lines=PATCH_EXTRA_LINES)
+            extended_patch = extend_patch(original_file_content_str, patch, num_lines=patch_extra_lines)
             full_extended_patch = f"\n\n## {file.filename}\n\n{extended_patch}\n"
 
             if add_line_numbers_to_hunks:

pr_agent/settings/configuration.toml

@@ -10,6 +10,7 @@ use_repo_settings_file=true
 ai_timeout=180
 max_description_tokens = 500
 max_commits_tokens = 500
+patch_extra_lines = 3
 secret_provider="google_cloud_storage"
 cli_mode=false
 

pr_agent/settings/pr_add_docs.toml

@@ -42,7 +42,9 @@ Specific instructions:
 {%- if extra_instructions %}
 
 Extra instructions from the user:
+'
 {{ extra_instructions }}
+'
 {%- endif %}
 
 You must use the following YAML schema to format your answer:

pr_agent/settings/pr_code_suggestions_prompts.toml

@@ -1,6 +1,6 @@
 [pr_code_suggestions_prompt]
 system="""You are a language model called PR-Code-Reviewer, that specializes in suggesting code improvements for Pull Request (PR).
-Your task is to provide meaningful and actionable code suggestions, to improve the new code presented in a PR.
+Your task is to provide meaningful and actionable code suggestions, to improve the new code presented in a PR (the '+' lines in the diff).
 
 Example for a PR Diff input:
 '
@@ -31,14 +31,13 @@ __old hunk__
 '
 
 Specific instructions:
-- Provide up to {{ num_code_suggestions }} code suggestions.
+- Provide up to {{ num_code_suggestions }} code suggestions. Try to provide diverse and insightful suggestions.
 - Prioritize suggestions that address major problems, issues and bugs in the code.
   As a second priority, suggestions should focus on best practices, code readability, maintainability, enhancments, performance, and other aspects.
-  Don't suggest to add docstring, type hints, or comments.
-  Try to provide diverse and insightful suggestions.
+- Don't suggest to add docstring, type hints, or comments.
 - Suggestions should refer only to code from the '__new hunk__' sections, and focus on new lines of code (lines starting with '+').
-  Avoid making suggestions that have already been implemented in the PR code. For example, if you want to add logs, or change a variable to const, or anything else, make sure it isn't already in the '__new hunk__' code.
-  For each suggestion, make sure to take into consideration also the context, meaning the lines before and after the relevant code.
+- Avoid making suggestions that have already been implemented in the PR code. For example, if you want to add logs, or change a variable to const, or anything else, make sure it isn't already in the '__new hunk__' code.
+- For each suggestion, make sure to take into consideration also the context, meaning the lines before and after the relevant code.
 - Provide the exact line numbers range (inclusive) for each issue.
 - Assume there is additional relevant code, that is not included in the diff.
 
@@ -46,7 +45,9 @@ Specific instructions:
 {%- if extra_instructions %}
 
 Extra instructions from the user:
+'
 {{ extra_instructions }}
+'
 {%- endif %}
 
 You must use the following YAML schema to format your answer:

pr_agent/settings/pr_description_prompts.toml

@@ -7,7 +7,9 @@ Your task is to provide full description of the PR content.
 {%- if extra_instructions %}
 
 Extra instructions from the user:
+'
 {{ extra_instructions }}
+'
 {% endif %}
 
 You must use the following YAML schema to format your answer:

pr_agent/settings/pr_reviewer_prompts.toml

@@ -35,7 +35,9 @@ The review should focus on new code added in the PR (lines starting with '+'), a
 {%- if extra_instructions %}
 
 Extra instructions from the user:
+'
 {{ extra_instructions }}
+'
 {% endif %}
 
 You must use the following YAML schema to format your answer:
@@ -129,8 +131,7 @@ PR Feedback:
   Security concerns:
     type: string
     description: >-
-      yes\\no question: does this PR code introduce possible security concerns or
-      issues, like SQL injection, XSS, CSRF, and others ? If answered 'yes',explain your answer shortly
+      yes\\no question: does this PR code introduce possible vulnerabilities such as exposure of sensitive information (e.g., API keys, secrets, passwords), or security concerns like SQL injection, XSS, CSRF, and others ? If answered 'yes', explain your answer briefly.
 {%- endif %}
 ```
 
@@ -196,7 +197,9 @@ Here are questions to better understand the PR. Use the answers to provide bette
 {{question_str|trim}}
 
 User answers:
+'
 {{answer_str|trim}}
+'
 ######
 {%- endif %}
 

pr_agent/settings/pr_update_changelog_prompts.toml

@@ -8,7 +8,9 @@ Your task is to update the CHANGELOG.md file of the project, to shortly summariz
 {%- if extra_instructions %}
 
 Extra instructions from the user:
+'
 {{ extra_instructions }}
+'
 {%- endif %}
 """
 

pr_agent/tools/pr_add_docs.py

@@ -68,7 +68,7 @@ class PRAddDocs:
                                         self.token_handler,
                                         model,
                                         add_line_numbers_to_hunks=True,
-                                        disable_extra_lines=True)
+                                        disable_extra_lines=False)
 
         logging.info('Getting AI prediction...')
         self.prediction = await self._get_prediction(model)