feat(changelog): add a gen ai summary of the diff

changelog_generator/ai_generator.py

@@ -0,0 +1,70 @@
+import os
+from google import genai
+from google.genai import types
+
+PROMPT = (
+    "Here's a git diff between two versions (represented as git tags). "
+    "Write a summary of this diff by following this structure. "
+    "A TL;DR section with very few lines expressing the salient points of your analysis "
+    "A Functional Changes section that should be understandable by readers that have no technical knowledge "
+    "A section about possible regressions that could appear : this section should aim to be brief as it might be used during a production incident, but it musn't be too generic either"
+    "A section about key kong/log metrics that should be observed during the production release process, these must be very precise"
+    "The summary must be written with markdown. It should be pleasing and give priority to salient points, must use colors and must use tasteful fonts, and follow best practices. You can use UML schemas and diagrams."
+)
+
+
+def generate_ai_summary(git_diff: str | None) -> str | None:
+    if not git_diff:
+        return None
+
+    project = os.getenv("VERTEX_PROJECT")
+    location = os.getenv("VERTEX_LOCATION")
+    model = os.getenv("VERTEX_MODEL")
+    credentials = os.getenv("VERTEX_CREDENTIALS")
+
+    if not all([project, location, model, credentials]):
+        return None
+
+    client = genai.Client(
+        vertexai=True,
+        project=project,
+        location=location,
+        credentials=credentials,
+    )
+
+    contents = [
+        types.Content(
+            role="user",
+            parts=[
+                types.Part.from_text(
+                    f"{PROMPT} {git_diff}"
+                )
+            ],
+        )
+    ]
+    generate_content_config = types.GenerateContentConfig(
+        temperature=1,
+        top_p=0.95,
+        max_output_tokens=8192,
+        response_modalities=["TEXT"],
+        safety_settings=[
+            types.SafetySetting(category="HARM_CATEGORY_HATE_SPEECH", threshold="OFF"),
+            types.SafetySetting(
+                category="HARM_CATEGORY_DANGEROUS_CONTENT", threshold="OFF"
+            ),
+            types.SafetySetting(
+                category="HARM_CATEGORY_SEXUALLY_EXPLICIT", threshold="OFF"
+            ),
+            types.SafetySetting(category="HARM_CATEGORY_HARASSMENT", threshold="OFF"),
+        ],
+    )
+
+    res = []
+    for chunk in client.models.generate_content_stream(
+        model=model,
+        contents=contents,
+        config=generate_content_config,
+    ):
+        res.append(chunk.text)
+
+    return "".join(res)

changelog_generator/changelog_template.jinja

@@ -1,5 +1,10 @@
 # [{{current_tag}}](https://github.com/{{organization}}/{{repository}}/compare/{{previous_tag}}...{{current_tag}})
 
+{% if ai_summary %}
+## AI generated summary
+{{ ai_summary }}
+{% endif -%}
+
 {% for type_node in commit_trees %}
 ## {{type_node.commit_type}}
 

changelog_generator/generator.py

@@ -5,6 +5,7 @@
 
 from .commit import Commit
 from .repository_manager import RepositoryManager
+from .ai_generator import generate_ai_summary
 
 
 class CommitTree(NamedTuple):
@@ -18,6 +19,7 @@ def render_changelog(
     previous_tag: str,
     current_tag: str,
     commit_trees: Sequence[CommitTree],
+    ai_summary: str | None,
 ) -> str:
     template_loader = FileSystemLoader(searchpath=os.path.dirname(__file__))
     template_environment = Environment(loader=template_loader)
@@ -29,6 +31,7 @@ def render_changelog(
         previous_tag=previous_tag,
         current_tag=current_tag,
         commit_trees=commit_trees,
+        ai_summary=ai_summary,
     )
 
 
@@ -67,15 +70,18 @@ def generate(
     if target:
         commits = repository.from_target(target)
         previous_tag, current_tag = target.split("..")
+        diff = None
     else:
         commits = repository.commits_since_last_tag
         previous_tag, current_tag = repository.previous_tag, repository.current_tag
+        diff = repository.get_diff_since_last_tag
 
     changelog = render_changelog(
         organization=repository.organization,
         repository=repository.name,
         previous_tag=previous_tag,
         current_tag=current_tag,
         commit_trees=get_commit_trees(commits),
+        ai_summary=generate_ai_summary(diff),
     )
     return changelog

changelog_generator/repository_manager.py

@@ -62,6 +62,14 @@ def commits_since_last_tag(self) -> Sequence[Commit]:
             revision = self.current_tag
         return self._get_commits(revision)
 
+    @property
+    @lru_cache()
+    def get_diff_since_last_tag(self) -> str | None:
+        if not self.previous_tag:
+            return None
+
+        return self.repository.git.diff(f"{self.previous_tag}..{self.current_tag}")
+
     def _get_commits(self, revision: str) -> Sequence[Commit]:
         options = {"no_merges": True}
         if self.filter_paths:

requirements.in

@@ -1,2 +1,3 @@
 gitpython
 jinja2
+google-genai