feat(#96): use code suggestions for GitLab

Author: scastlara

assets/review-comment.png

@@ -14,6 +14,7 @@
 )
 from lgtm_ai.ai.schemas import AgentSettings, CommentCategory, SupportedAIModels, SupportedAIModelsList
 from lgtm_ai.base.schemas import OutputFormat, PRUrl
+from lgtm_ai.base.utils import git_source_supports_suggestions
 from lgtm_ai.config.handler import ConfigHandler, PartialConfig
 from lgtm_ai.formatters.base import Formatter
 from lgtm_ai.formatters.json import JsonFormatter
@@ -138,7 +139,11 @@ def review(
         config_file=config,
     ).resolve_config()
     agent_extra_settings = AgentSettings(retries=resolved_config.ai_retries)
-    git_client = get_git_client(pr_url=pr_url, config=resolved_config, formatter=MarkDownFormatter())
+    git_client = get_git_client(
+        pr_url=pr_url,
+        config=resolved_config,
+        formatter=MarkDownFormatter(use_suggestions=git_source_supports_suggestions(pr_url.source)),
+    )
     code_reviewer = CodeReviewer(
         reviewer_agent=get_reviewer_agent_with_settings(agent_extra_settings),
         summarizing_agent=get_summarizing_agent_with_settings(agent_extra_settings),

src/lgtm_ai/__main__.py

@@ -101,9 +101,12 @@ def _get_full_category_explanation() -> str:
     - Evaluate whether some comments are more likely to simply be incorrect. If they are likely to be incorrect, remove them.
     - Merge duplicate comments. If there are two comments that refer to the same issue, merge them into one.
     - Comments have a code snippet that they refer to. Consider whether the snippet needs a bit more code context, and if so, expand the snippet. Otherwise don't touch them.
-    - If you can add a suggestion code snippet to the comment text, do it. Use markdown code blocks with 5 backticks. Do it only when you are very sure about the suggestion with the context you have.
     - Check that categories of each comment are correct. Re-categorize them if needed.
     - Check the summary. Feel free to rephrase it, add more information, or generally improve it. The summary comment must be a general comment informing the PR author about the overall quality of the PR, the weakpoints it has, and which general issues need to be addressed.
+    - If you can add a suggestion code snippet to the comment text, do it. Do it only when you are very sure about the suggestion with the context you have.
+    - Suggestions must be passed separately (not as part of the comment content), and they must include how many lines above and below the comment to include in the suggestion.
+    - The offsets of suggestions must encompass all the code that needs to be changed. e.g., if you intend to change a whole function, the suggestion must include the full function. If you intend to change a single line, then the offsets will be 0.
+    - If a suggestion is given, a flag indicating whether the suggestion is ready to be applied directly by the author must be given. That is, if the suggestion includes comments to be filled by the author, or skips parts and is intended for clarification, the flag `ready_for_replacement` must be set to `false`.
 
     The review will have a score for the PR (1-5, with 5 being the best). It is your job to evaluate whether this score holds after removing the comments.
     You must evaluate the score, and change it if necessary. Here is some guidance:

src/lgtm_ai/ai/prompts.py

@@ -2,12 +2,12 @@
 import zoneinfo
 from dataclasses import dataclass
 from functools import cached_property
-from typing import Annotated, Final, Literal, get_args
+from typing import Annotated, Final, Literal, Self, get_args
 from uuid import uuid4
 
 from lgtm_ai.git_client.schemas import PRDiff
 from openai.types import ChatModel
-from pydantic import AfterValidator, BaseModel, Field, computed_field
+from pydantic import AfterValidator, BaseModel, Field, computed_field, model_validator
 from pydantic_ai.models.anthropic import LatestAnthropicModelNames
 from pydantic_ai.models.mistral import LatestMistralModelNames
 from pydantic_ai.usage import Usage
@@ -81,6 +81,35 @@
 }
 
 
+class CodeSuggestionOffset(BaseModel):
+    offset: Annotated[int, Field(description="Offset relative to the comment line number")]
+    direction: Annotated[
+        Literal["+", "-"],
+        Field(description="Direction of the offset. + means below, - means above"),
+    ]
+
+    @model_validator(mode="after")
+    def change_direction_of_zero(self) -> Self:
+        # GitLab freaks out if the offset is 0 and the direction is +. We change it to -.
+        if self.offset == 0:
+            self.direction = "-"
+        return self
+
+
+class CodeSuggestion(BaseModel):
+    start_offset: Annotated[
+        CodeSuggestionOffset, Field(description="Offset (from comment line number) to start the suggestion")
+    ]
+    end_offset: Annotated[
+        CodeSuggestionOffset, Field(description="Offset (from comment line number) to end the suggestion")
+    ]
+    snippet: Annotated[str, Field(description="Suggested code snippet to replace the commented code")]
+    programming_language: Annotated[str, Field(description="Programming language of the code snippet")]
+    ready_for_replacement: Annotated[
+        bool, Field(description="Whether the suggestion is totally ready to be applied directly")
+    ] = False
+
+
 class ReviewComment(BaseModel):
     """Individual comment representation in a PR code review."""
 
@@ -94,6 +123,7 @@ class ReviewComment(BaseModel):
     is_comment_on_new_path: Annotated[bool, Field(description="Whether the comment is on a new path")]
     programming_language: Annotated[str, Field(description="Programming language of the file")]
     quote_snippet: Annotated[str | None, Field(description="Quoted code snippet")] = None
+    suggestion: Annotated[CodeSuggestion | None, Field(description="Suggested code change")] = None
 
 
 class ReviewResponse(BaseModel):

src/lgtm_ai/ai/schemas.py

@@ -1,14 +1,18 @@
 from dataclasses import dataclass
 from enum import StrEnum
-from typing import Literal
+
+
+class PRSource(StrEnum):
+    github = "github"
+    gitlab = "gitlab"
 
 
 @dataclass(frozen=True, slots=True)
 class PRUrl:
     full_url: str
     repo_path: str
     pr_number: int
-    source: Literal["github", "gitlab"]
+    source: PRSource
 
 
 class OutputFormat(StrEnum):

src/lgtm_ai/base/schemas.py

@@ -1,6 +1,8 @@
 import fnmatch
 import pathlib
 
+from lgtm_ai.base.schemas import PRSource
+
 
 def file_matches_any_pattern(file_name: str, patterns: tuple[str, ...]) -> bool:
     for pattern in patterns:
@@ -10,3 +12,11 @@ def file_matches_any_pattern(file_name: str, patterns: tuple[str, ...]) -> bool:
         if matches:
             return True
     return False
+
+
+def git_source_supports_suggestions(source: PRSource) -> bool:
+    """For now, we only support suggestions in GitLab.
+
+    TODO: https://github.com/elementsinteractive/lgtm-ai/issues/96
+    """
+    return source == PRSource.gitlab

src/lgtm_ai/base/utils.py

@@ -17,7 +17,8 @@ class MarkDownFormatter(Formatter[str]):
     USAGES_SUMMARY_TEMPLATE: ClassVar[str] = "usages_summary.md.j2"
     METADATA_TEMPLATE: ClassVar[str] = "metadata.md.j2"
 
-    def __init__(self) -> None:
+    def __init__(self, use_suggestions: bool = False) -> None:
+        self.use_suggestions = use_suggestions
         template_dir = pathlib.Path(__file__).parent / "templates"
         self._template_env = Environment(loader=FileSystemLoader(template_dir), autoescape=True)
 
@@ -52,6 +53,8 @@ def format_review_comment(self, comment: ReviewComment, *, with_footer: bool = T
             severity_icon=severity_icon,
             snippet=snippet,
             comment=comment.comment,
+            suggestion=comment.suggestion,
+            use_suggestions=self.use_suggestions,
             with_footer=with_footer,
             new_path=comment.new_path,
             line_number=comment.line_number,

src/lgtm_ai/formatters/markdown.py

@@ -43,6 +43,7 @@ def format_review_comment(self, comment: ReviewComment, *, with_footer: bool = T
                 padding=(1, 1),
             )
             content = Group(snippet_panel, Text(comment.comment))
+            # TODO: add code suggestions
         else:
             content = Text(comment.comment)
 

src/lgtm_ai/formatters/pretty.py

@@ -8,6 +8,18 @@
 {% endif %}
 {{ comment | safe }}
 
+{% if suggestion %}
+{% if use_suggestions and suggestion.ready_for_replacement %}
+`````suggestion:{{ suggestion.start_offset.direction}}{{ suggestion.start_offset.offset }}{{ suggestion.end_offset.direction }}{{ suggestion.end_offset.offset }}
+{{ suggestion.snippet | safe }}
+`````
+{% else %}
+`````{{ suggestion.programming_language }}
+{{ suggestion.snippet | safe }}
+`````
+{% endif %}
+{% endif %}
+
 {% if with_footer %}
 <details><summary>More information about this comment</summary>
 

src/lgtm_ai/formatters/templates/review_comment.md.j2

@@ -1,9 +1,8 @@
 from enum import StrEnum
-from typing import Literal
 from urllib.parse import ParseResult, urlparse
 
 import click
-from lgtm_ai.base.schemas import PRUrl
+from lgtm_ai.base.schemas import PRSource, PRUrl
 
 
 class AllowedLocations(StrEnum):
@@ -35,7 +34,7 @@ def parse_pr_url(ctx: click.Context, param: str, value: object) -> PRUrl:
             return _parse_pr_url(
                 parsed,
                 split_str="/-/merge_requests/",
-                source="gitlab",
+                source=PRSource.gitlab,
                 error_url_msg="The PR URL must be a merge request URL.",
                 error_num_msg="The PR URL must contain a valid MR number.",
             )
@@ -44,7 +43,7 @@ def parse_pr_url(ctx: click.Context, param: str, value: object) -> PRUrl:
             return _parse_pr_url(
                 parsed,
                 split_str="/pull/",
-                source="github",
+                source=PRSource.github,
                 error_url_msg="The PR URL must be a pull request URL.",
                 error_num_msg="The PR URL must contain a valid PR number.",
             )
@@ -94,7 +93,7 @@ def validate_model_url(ctx: click.Context, param: click.Parameter, value: str |
 
 
 def _parse_pr_url(
-    parsed: ParseResult, *, split_str: str, source: Literal["github", "gitlab"], error_url_msg: str, error_num_msg: str
+    parsed: ParseResult, *, split_str: str, source: PRSource, error_url_msg: str, error_num_msg: str
 ) -> PRUrl:
     full_project_path = parsed.path
     try: