Upload 4 files

app.py

@@ -0,0 +1,59 @@
+# app.py — Streamlit UI for AutoReview AI
+
+import streamlit as st
+from main import review_pr_simple
+
+# ── PAGE CONFIG ───────────────────────────
+st.set_page_config(
+    page_title="AutoReview AI",
+    page_icon="🤖",
+    layout="centered"
+)
+
+# ── TITLE ─────────────────────────────────
+st.title("🤖 AutoReview AI")
+st.markdown("*Autonomous GitHub PR Reviewer powered by LLaMA + Groq*")
+st.divider()
+
+# ── INPUT ─────────────────────────────────
+pr_url = st.text_input(
+    "Enter GitHub PR URL",
+    placeholder="https://github.com/owner/repo/pull/123"
+)
+
+# ── BUTTON ────────────────────────────────
+if st.button("🔍 Review PR", type="primary"):
+
+    # check if URL is empty
+    if not pr_url:
+        st.warning("Please enter a PR URL first!")
+
+    else:
+        # show spinner while reviewing
+        with st.spinner("🤖 Reviewing PR... this may take 30 seconds"):
+            result = review_pr_simple(pr_url)
+
+        # if success
+        if result["success"]:
+            st.success("✅ Review Complete!")
+            st.divider()
+
+            # show PR data
+            st.subheader("📋 PR Summary")
+            st.code(result["pr_data"][:500])
+
+            # show analysis
+            st.subheader("🔍 Code Analysis")
+            st.write(result["analysis"])
+
+            # show verdict
+            st.subheader("✅ Final Verdict")
+            st.write(result["verdict"])
+
+        # if failed
+        else:
+            st.error(f"❌ Failed: {result['error']}")
+
+# ── FOOTER ────────────────────────────────
+st.divider()
+st.caption("Built with LangChain + Groq + Streamlit")

main.py

@@ -0,0 +1,179 @@
+# main.py — AutoReview AI
+# LangChain-based autonomous GitHub PR reviewer
+
+import os
+import re
+from dotenv import load_dotenv
+
+from langchain_groq import ChatGroq
+from langchain.agents import AgentExecutor, create_tool_calling_agent
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.messages import HumanMessage, SystemMessage
+
+from tools import fetch_pr_diff, analyze_diff, post_inline_comment, post_overall_review
+
+load_dotenv()
+
+
+# ── LLM ──────────────────────────────────
+llm = ChatGroq(model="llama-3.3-70b-versatile", temperature=0)
+
+# ── TOOLS ─────────────────────────────────
+tools = [fetch_pr_diff, analyze_diff, post_inline_comment, post_overall_review]
+
+# ── PROMPT ────────────────────────────────
+prompt = ChatPromptTemplate.from_messages([
+    ("system", """You are AutoReview AI — an expert autonomous GitHub PR reviewer.
+
+Your job when given a PR URL:
+1. Use fetch_pr_diff to get the PR details and code changes
+2. Use analyze_diff for each changed file to find bugs, security issues, style problems
+3. Use post_inline_comment to post specific comments on problematic lines
+4. Use post_overall_review to post a final APPROVE or REQUEST_CHANGES verdict
+
+Be thorough but concise. Focus on real issues that matter.
+Always complete all 4 steps — don't stop after fetching.
+
+If GitHub token is not available, still analyze and return the review as text.
+"""),
+    ("human", "{input}"),
+    ("placeholder", "{agent_scratchpad}")
+])
+
+# ── AGENT ─────────────────────────────────
+agent          = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(
+    agent=agent,
+    tools=tools,
+    verbose=True,
+    max_iterations=10,
+    handle_parsing_errors=True
+)
+
+# ─────────────────────────────────────────
+# MAIN REVIEW FUNCTION
+# ─────────────────────────────────────────
+def review_pr(pr_url: str) -> dict:
+    """
+    Main function to review a GitHub PR.
+    Returns structured review result.
+    """
+    print(f"\n{'='*60}")
+    print(f"🔍 AutoReview AI starting review...")
+    print(f"📎 PR: {pr_url}")
+    print("="*60)
+
+    try:
+        result = agent_executor.invoke({
+            "input": f"""
+Please review this GitHub Pull Request: {pr_url}
+
+Steps to follow:
+1. Fetch the PR diff using fetch_pr_diff
+2. Analyze each changed file using analyze_diff
+3. Post inline comments for specific issues using post_inline_comment
+4. Post overall verdict using post_overall_review
+
+Provide a thorough code review focusing on:
+- Bugs and logic errors
+- Security vulnerabilities
+- Code style and best practices
+- Performance issues
+"""
+        })
+
+        output = result.get("output", "")
+        print(f"\n✅ Review complete!")
+        print(f"\nOutput:\n{output}")
+
+        return {
+            "success": True,
+            "pr_url":  pr_url,
+            "review":  output,
+            "error":   None
+        }
+
+    except Exception as e:
+        error_msg = str(e)
+        print(f"\n❌ Error: {error_msg}")
+        return {
+            "success": False,
+            "pr_url":  pr_url,
+            "review":  None,
+            "error":   error_msg
+        }
+
+# ─────────────────────────────────────────
+# SIMPLE REVIEW (no GitHub token needed)
+# For demo purposes
+# ─────────────────────────────────────────
+def review_pr_simple(pr_url: str) -> dict:
+    """
+    Reviews PR without posting comments to GitHub.
+    Returns full analysis as text — good for demo.
+    """
+    print(f"\n🔍 Fetching PR data...")
+
+    # Step 1: Fetch diff
+    diff_result = fetch_pr_diff.invoke(pr_url)
+
+    if diff_result.startswith("ERROR"):
+        return {"success": False, "error": diff_result, "review": None}
+
+    print("✅ PR fetched!")
+    print(f"\n🤖 Analyzing code...")
+
+    # Step 2: Analyze
+    analysis = analyze_diff.invoke(diff_result[:4000])
+
+    print("✅ Analysis complete!")
+
+    # Step 3: Generate final verdict
+    verdict_response = llm.invoke([
+        SystemMessage("You are a senior engineer. Give a final PR review verdict."),
+        HumanMessage(f"""
+Based on this PR analysis, write a complete review report:
+
+PR DATA:
+{diff_result[:2000]}
+
+ANALYSIS:
+{analysis}
+
+Write a professional review with:
+1. Overall verdict (APPROVE / REQUEST CHANGES)
+2. Key issues found
+3. Positive aspects
+4. Specific suggestions
+""")
+    ])
+
+    return {
+        "success":  True,
+        "pr_url":   pr_url,
+        "pr_data":  diff_result,
+        "analysis": analysis,
+        "verdict":  verdict_response.content,
+        "error":    None
+    }
+
+# ─────────────────────────────────────────
+# RUN
+# ─────────────────────────────────────────
+if __name__ == "__main__":
+    # Test with a real public PR
+    test_pr = "https://github.com/langchain-ai/langchain/pull/1"
+
+    print("Mode: Simple review (no GitHub token needed for demo)")
+    result = review_pr_simple(test_pr)
+
+    if result["success"]:
+        print(f"\n{'='*60}")
+        print("📋 PR DATA SUMMARY:")
+        print(result["pr_data"][:500])
+        print(f"\n🔍 ANALYSIS:")
+        print(result["analysis"])
+        print(f"\n✅ FINAL VERDICT:")
+        print(result["verdict"])
+    else:
+        print(f"❌ Failed: {result['error']}")

requirements.txt

@@ -1,3 +1,6 @@
-altair
-pandas
-streamlit
+langchain==0.2.0
+langchain-groq==0.1.6
+# langchain-core==0.2.0
+PyGithub==2.3.0
+python-dotenv==1.0.0
+streamlit==1.35.0

tools.py

@@ -0,0 +1,218 @@
+ # tools.py — LangChain tools for AutoReview AI
+
+import os
+import re
+from github import Github
+from dotenv import load_dotenv
+from langchain.tools import tool
+from langchain_groq import ChatGroq
+from langchain_core.messages import HumanMessage, SystemMessage
+load_dotenv()
+print("Tools key:",os.getenv("GROQ_API_KEY"))
+# ── LLM ──────────────────────────────────
+llm = ChatGroq(model="llama-3.3-70b-versatile", temperature=0)
+
+# ── GITHUB CLIENT ─────────────────────────
+def get_github_client():
+    token = os.environ.get("GITHUB_TOKEN", "")
+    return Github(token) if token else Github()
+
+# ─────────────────────────────────────────
+# TOOL 1 — FETCH PR DIFF
+# ─────────────────────────────────────────
+@tool
+def fetch_pr_diff(pr_url: str) -> str:
+    """
+    Fetches the diff and metadata from a GitHub Pull Request URL.
+    Returns structured info: PR title, description, changed files and their diffs.
+    Input: GitHub PR URL like https://github.com/owner/repo/pull/123
+    """
+    try:
+        # Parse owner, repo, PR number from URL
+        pattern = r"github\.com/([^/]+)/([^/]+)/pull/(\d+)"
+        match   = re.search(pattern, pr_url)
+        if not match:
+            return "ERROR: Invalid GitHub PR URL. Format: https://github.com/owner/repo/pull/123"
+
+        owner   = match.group(1)
+        repo    = match.group(2)
+        pr_num  = int(match.group(3))
+
+        g       = get_github_client()
+        repo_obj= g.get_repo(f"{owner}/{repo}")
+        pr      = repo_obj.get_pull(pr_num)
+
+        # Get changed files
+        files   = pr.get_files()
+        file_diffs = []
+
+        for f in files:
+            if f.patch:  # only files with actual changes
+                file_diffs.append({
+                    "filename": f.filename,
+                    "status":   f.status,      # added/modified/deleted
+                    "additions":f.additions,
+                    "deletions":f.deletions,
+                    "patch":    f.patch[:3000]  # limit patch size
+                })
+
+        result = {
+            "pr_title":       pr.title,
+            "pr_description": pr.body or "No description",
+            "pr_number":      pr_num,
+            "owner":          owner,
+            "repo":           repo,
+            "base_branch":    pr.base.ref,
+            "head_branch":    pr.head.ref,
+            "total_files":    pr.changed_files,
+            "additions":      pr.additions,
+            "deletions":      pr.deletions,
+            "files":          file_diffs
+        }
+
+        # Format as readable string for LLM
+        output = f"""
+PR TITLE: {result['pr_title']}
+PR NUMBER: #{result['pr_number']}
+DESCRIPTION: {result['pr_description'][:500]}
+BRANCH: {result['head_branch']} → {result['base_branch']}
+STATS: +{result['additions']} additions, -{result['deletions']} deletions, {result['total_files']} files changed
+
+FILES CHANGED:
+"""
+        for f in file_diffs[:10]:  # max 10 files
+            output += f"""
+--- {f['filename']} ({f['status']}) +{f['additions']}/-{f['deletions']} ---
+{f['patch'][:2000]}
+"""
+        return output.strip()
+
+    except Exception as e:
+        return f"ERROR fetching PR: {str(e)}"
+
+# ─────────────────────────────────────────
+# TOOL 2 — ANALYZE DIFF
+# ─────────────────────────────────────────
+@tool
+def analyze_diff(file_info: str) -> str:
+    """
+    Analyzes a code diff and returns review comments.
+    Input: string containing filename and its diff/patch
+    Returns: structured review with bugs, security issues, suggestions
+    """
+    response = llm.invoke([
+        SystemMessage(content="""You are a senior software engineer doing a code review.
+Analyze the code diff carefully and provide specific, actionable feedback.
+
+Return your review in EXACTLY this format:
+BUGS:
+- <line number if known>: <specific bug description> or NONE
+
+SECURITY:
+- <line number if known>: <security issue> or NONE
+
+STYLE:
+- <specific style issue> or NONE
+
+SUGGESTIONS:
+- <improvement suggestion> or NONE
+
+VERDICT: APPROVE or REQUEST_CHANGES
+SUMMARY: <one sentence overall assessment>"""),
+        HumanMessage(content=f"""
+Review this code change:
+
+{file_info}
+
+Be specific — mention exact line numbers when possible.
+Focus on real issues, not nitpicks.
+""")
+    ])
+
+    return response.content
+
+# ─────────────────────────────────────────
+# TOOL 3 — POST INLINE COMMENTS
+# ───────────────────────────���─────────────
+@tool
+def post_inline_comment(comment_data: str) -> str:
+    """
+    Posts an inline review comment on a GitHub PR.
+    Input format: 'owner|repo|pr_number|commit_sha|filename|line_number|comment_body'
+    Returns: success or error message
+    """
+    try:
+        parts = comment_data.split("|")
+        if len(parts) < 7:
+            return "ERROR: Need format: owner|repo|pr_number|commit_sha|filename|line_number|comment"
+
+        owner      = parts[0].strip()
+        repo       = parts[1].strip()
+        pr_number  = int(parts[2].strip())
+        commit_sha = parts[3].strip()
+        filename   = parts[4].strip()
+        line       = int(parts[5].strip())
+        body       = "|".join(parts[6:]).strip()
+
+        token = os.environ.get("GITHUB_TOKEN", "")
+        if not token:
+            return "ERROR: GITHUB_TOKEN not set — cannot post comments"
+
+        g        = Github(token)
+        repo_obj = g.get_repo(f"{owner}/{repo}")
+        pr       = repo_obj.get_pull(pr_number)
+        commit   = repo_obj.get_commit(commit_sha)
+
+        pr.create_review_comment(
+            body=f"🤖 **AutoReview AI:** {body}",
+            commit=commit,
+            path=filename,
+            line=line
+        )
+
+        return f"✅ Comment posted on {filename} line {line}"
+
+    except Exception as e:
+        return f"ERROR posting comment: {str(e)}"
+
+# ─────────────────────────────────────────
+# TOOL 4 — POST OVERALL REVIEW
+# ─────────────────────────────────────────
+@tool
+def post_overall_review(review_data: str) -> str:
+    """
+    Posts an overall PR review (approve or request changes) on GitHub.
+    Input format: 'owner|repo|pr_number|APPROVE or REQUEST_CHANGES|summary_body'
+    Returns: success or error message
+    """
+    try:
+        parts  = review_data.split("|")
+        owner  = parts[0].strip()
+        repo   = parts[1].strip()
+        pr_num = int(parts[2].strip())
+        event  = parts[3].strip()  # APPROVE or REQUEST_CHANGES
+        body   = "|".join(parts[4:]).strip()
+
+        if event not in ("APPROVE", "REQUEST_CHANGES", "COMMENT"):
+            event = "COMMENT"
+
+        token = os.environ.get("GITHUB_TOKEN", "")
+        if not token:
+            return "Review NOT posted to GitHub (no token) — here is the review:\n" + body
+
+        g        = Github(token)
+        repo_obj = g.get_repo(f"{owner}/{repo}")
+        pr       = repo_obj.get_pull(pr_num)
+
+        review_body = f"""🤖 **AutoReview AI Report**
+
+{body}
+
+---
+*Reviewed by AutoReview AI — LangChain + Groq*
+"""
+        pr.create_review(body=review_body, event=event)
+        return f"✅ Overall review posted: {event}"
+
+    except Exception as e:
+        return f"ERROR posting review: {str(e)}"