ai experiments

.github/scripts/ai_review.py

@@ -0,0 +1,173 @@
+import os
+import requests
+import openai
+import logging
+
+# Configure logging
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
+)
+logger = logging.getLogger('pr_review_bot')
+
+# Define model arrays
+REGULAR_MODELS = ["gpt-3.5-turbo", "gpt-4o"]
+REASONING_MODELS = ["o1-mini", "o1-preview"]
+
+# Extract PR number from github environment
+def get_pr_number(github_ref):
+    parts = github_ref.split("/")
+    if len(parts) >= 3 and parts[1] == "pull":
+        pr_number = parts[2]
+    else:
+        raise ValueError(f"Unexpected GITHUB_REF format: {github_ref}")
+    return pr_number
+
+# Fetch PR code diff
+def fetch_diff(pr_url, github_token):
+    headers = {
+        "Authorization": f"Bearer {github_token}",
+        "Accept": "application/vnd.github.v3.diff"
+    }
+    diff_response = requests.get(pr_url, headers=headers)
+    if diff_response.status_code != 200:
+        raise RuntimeError(f"Failed to fetch PR diff: {diff_response.status_code} {diff_response.text}")
+    return diff_response.text
+
+# Fetch PR description
+def fetch_pr_description(pr_url, github_token):
+    json_headers = {
+        "Authorization": f"Bearer {github_token}",
+        "Accept": "application/vnd.github.v3+json"
+    }
+    pr_details_response = requests.get(pr_url, headers=json_headers)
+    pr_data = pr_details_response.json()
+    pr_title = pr_data.get("title", "(no title)")
+    pr_body  = pr_data.get("body", "(no body)")
+    return pr_title, pr_body
+
+def generate_review_regular(diff_text, pr_title, pr_body, model_name):
+    logger.info(f"Generating code review with regular model: {model_name}")
+    prompt = (
+        "You are acting as an advanced code review assistant. Below is a diff from a Pull Request. "
+        "Please analyze these changes in detail and provide a constructive critique. Focus on:\n"
+        "- Potential bugs and security vulnerabilities\n"
+        "- Conformance to coding style and best practices\n"
+        "- Opportunities for performance or maintainability improvements\n"
+        "\n"
+        f"Diff:\n{diff_text}"
+    )
+
+    dev_prompt = (
+        "You are a highly experienced senior software engineer and code reviewer with deep "
+        "expertise across various programming languages and frameworks "
+        "(including Solidity, JavaScript/TypeScript, and Rust). "
+        "Your role is to thoroughly analyze code changes, focusing on correctness, security, "
+        "maintainability, and adherence to best practices. Provide clear, actionable, and concise "
+        "feedback with concrete suggestions for improvement where necessary. Avoid unnecessary elaboration, "
+        "but ensure that critical details are clearly explained."
+    )
+
+    try:
+        completion = openai.chat.completions.create(
+            model=model_name,
+            messages=[
+                {"role": "developer", "content": dev_prompt},
+                {"role": "user", "content": prompt}
+            ],
+            temperature=0.3,
+            max_completion_tokens=1000
+        )
+    except Exception as e:
+        logger.error(f"Failed to generate review: {e}")
+        raise RuntimeError(f"Failed to generate review: {e}")
+
+    return completion.choices[0].message.content
+
+def generate_review_reasoning(diff_text, pr_title, pr_body, model_name):
+    logger.info(f"Generating code review with reasoning model: {model_name}")
+    prompt = (
+        "Do code review and analyze code changes. "
+        "Focus on:\n"
+        "- Potential bugs and security vulnerabilities\n"
+        "- Conformance to coding style and best practices\n"
+        "- Opportunities for performance or maintainability improvements\n"
+        "\n"
+        "Give summary of what the code changes are doing. "
+        "Then identify existing issues and provide specific, actionable suggestions "
+        "how to fix the identified issues if there are any. "
+        "Be brief, avoid giving general recommendations not related to code fixes. "
+        "Do not recommend changes outside of the code changes in the diff."
+        "\n\n"
+        f"PR Title:\n{pr_title}\n"
+        f"PR Description:\n{pr_body}\n"
+        f"Diff:\n{diff_text}"
+    )
+
+    try:
+        completion = openai.chat.completions.create(
+            model=model_name,
+            messages=[
+                {"role": "user", "content": prompt}
+            ]
+        )
+    except Exception as e:
+        logger.error(f"Failed to generate review: {e}")
+        raise RuntimeError(f"Failed to generate review: {e}")
+
+    return completion.choices[0].message.content
+
+def generate_review(diff_text, pr_title, pr_body, model_name="o1-mini"):
+    if model_name in REGULAR_MODELS:
+        return generate_review_regular(diff_text, pr_title, pr_body, model_name)
+    elif model_name in REASONING_MODELS:
+        return generate_review_reasoning(diff_text, pr_title, pr_body, model_name)
+    else:
+        logger.warning(f"Unknown model type: {model_name}, trying regular review")
+        return generate_review_regular(diff_text, pr_title, pr_body, model_name)
+
+# Post review as a comment
+def post_review_comment(repo, pr_number, review_text, github_token):
+    logger.info(f"Posting review comment for PR #{pr_number}")
+    comment_url = f"https://api.github.com/repos/{repo}/issues/{pr_number}/comments"
+    comment_headers = {"Authorization": f"Bearer {github_token}"} 
+    response = requests.post(comment_url, headers=comment_headers, json={"body": review_text})
+    if response.status_code != 201:
+        logger.error(f"Failed to post comment: {response.status_code} {response.text}")
+        return False
+    else:
+        logger.info("Successfully posted review comment")
+        return True
+
+# Action start
+logger.info("Starting PR review workflow")
+
+# Get environment variables
+repo = os.environ.get("GITHUB_REPOSITORY")
+github_ref = os.environ.get("GITHUB_REF")
+github_token = os.environ.get("GITHUB_TOKEN")
+openai.api_key = os.environ["OPENAI_API_KEY"]
+model_name = os.environ.get("OPENAI_MODEL", "o1-mini")
+
+# Check if required environment variables exist
+if not repo or not github_ref or not github_token:
+    logger.critical("Missing required environment variables")
+    raise ValueError("Missing required environment variables: GITHUB_REPOSITORY, GITHUB_REF, or GITHUB_TOKEN")
+
+# Get PR ids
+pr_number = get_pr_number(github_ref)
+logger.info(f"Processing PR #{pr_number} with model {model_name}")
+pr_url = f"https://api.github.com/repos/{repo}/pulls/{pr_number}"
+
+# Fetch PR details
+logger.info("Fetching PR diff")
+diff_text = fetch_diff(pr_url, github_token)
+logger.info("Fetching PR description")
+pr_title, pr_body = fetch_pr_description(pr_url, github_token)
+
+# Generate review with AI
+review_text = generate_review(diff_text, pr_title, pr_body, model_name)
+
+# Post the review
+post_review_comment(repo, pr_number, review_text, github_token)
+logger.info("PR review workflow completed")

.github/workflows/ai-review.yml

@@ -0,0 +1,26 @@
+name: AI Code Review
+
+permissions:
+  contents: read
+  pull-requests: write
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+
+jobs:
+  code_review:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v3
+
+      - name: Install dependencies
+        run: pip install openai requests
+
+      - name: Run AI code review script
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          OPENAI_MODEL: o1-preview
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: python3 .github/scripts/ai_review.py