Allow reading of comments for github issue

app/main.py

@@ -106,7 +106,12 @@ def main(args, subparser_dest_attr_name: str = "command"):
             setup_dir = abspath(setup_dir)
 
         task = RawGithubTask(
-            args.task_id, args.clone_link, args.commit_hash, args.issue_link, setup_dir
+            args.task_id,
+            args.clone_link,
+            args.commit_hash,
+            args.issue_link,
+            setup_dir,
+            args.use_comments,
         )
         groups = {"github": [task]}
         run_task_groups(groups, num_processes)
@@ -160,6 +165,12 @@ def set_github_parser_args(parser: ArgumentParser) -> None:
         type=str,
         help="The commit hash to checkout. If not specified, the latest commit on default branch will be used.",
     )
+    parser.add_argument(
+        "--use-comments",
+        action="store_true",
+        default=False,
+        help="Include the comments of the issue.",
+    )
     parser.add_argument("--issue-link", type=str, help="The link to the issue.")
     parser.add_argument(
         "--setup-dir",

app/model/common.py

@@ -206,7 +206,7 @@ def set_model(model_name: str):
                 completion_tokens=completion_tokens,
             )
         )
-        litellm.set_verbose = True
+        # litellm.set_verbose = True
         SELECTED_MODEL = LiteLLMGeneric(
             real_model_name,
             prompt_tokens_cost_usd_dollar,

app/raw_tasks.py

@@ -95,13 +95,15 @@ def __init__(
         commit_hash: str | None,
         issue_link: str,
         setup_dir: str,
+        use_comments: bool = False,
     ):
         self._task_id = task_id
         self.clone_link = clone_link
         # if commit_hash is None, assume using the HEAD of default branch
         self.commit_hash = commit_hash
         self.issue_link = issue_link
         self.setup_dir = setup_dir
+        self.use_comments = use_comments
         self.clone_path = pjoin(self.setup_dir, self.task_id)
         self.problem_statement, self.created_at = self.fetch_issue()
         self.clone_repo()
@@ -144,7 +146,7 @@ def fetch_issue(self):
         if "github.com" not in self.issue_link:
             raise NotImplementedError("Only GitHub issues are supported for now.")
 
-        retrieved_issue = self.fetch_github_issue(self.issue_link)
+        retrieved_issue = self.fetch_github_issue(self.issue_link, self.use_comments)
 
         if retrieved_issue is None:
             raise RuntimeError(
@@ -185,25 +187,50 @@ def process_links(cls, body: str):
         return body
 
     @classmethod
-    def fetch_github_issue(cls, issue_url: str) -> tuple[str, str, str]:
+    def fetch_github_issue(
+        cls, issue_url: str, use_comments: bool = False
+    ) -> tuple[str, str, str]:
         """Extract owner, repo, and issue number from the URL"""
 
         # Example issue URL: https://github.com/owner/repo/issues/123
 
         _, owner, repo, _, issue_number = issue_url.rsplit("/", 4)
 
         api_url = f"https://api.github.com/repos/{owner}/{repo}/issues/{issue_number}"
-        response = httpx.get(api_url)
+        comments_url = f"https://api.github.com/repos/{owner}/{repo}/issues/{issue_number}/comments"
 
-        if response.status_code != 200:
+        issue_response = httpx.get(api_url)
+
+        if issue_response.status_code != 200:
             raise RuntimeError(
-                f"Failed to fetch issue information: {response.status_code}"
+                f"Failed to fetch issue information: {issue_response.status_code}"
             )
 
-        issue_info = response.json()
+        issue_info = issue_response.json()
 
         title = issue_info["title"]
         body = issue_info["body"]
+
+        if use_comments:
+            comments_response = httpx.get(comments_url)
+            if comments_response.status_code != 200:
+                raise RuntimeError(
+                    f"Failed to fetch comments information: {comments_response.status_code}"
+                )
+
+            comments_info = comments_response.json()
+            for comment in comments_info:
+                if (
+                    "user" not in comment
+                    or comment["user"]["type"] == "Bot"
+                    or comment["user"]["login"] == "acr-bot"
+                ):
+                    continue
+
+                body += (
+                    f"\nUser: {comment['user']['login']}\nComment: {comment['body']}"
+                )
+
         created_at = issue_info["created_at"]
 
         return title, body, created_at