Learn how to use DeepReview

Everything you need to integrate AI-powered code reviews into your workflow.

Quick Start

Get your team up and running with DeepReview in under 10 minutes. This guide walks you through integrating AI-powered code reviews into your Git workflow.

Prerequisites

  • Admin access to your Git repository (GitHub, GitLab, or Gitee)
  • A DeepReview enterprise account (contact us to get started)
  • SSH key pair for repository access, or a deploy token (for token-based authentication)

Step 1: Get Your Enterprise Account

Contact our team to create your enterprise account. Once created, you'll receive:

  • Enterprise Name: Your unique identifier for login
  • Initial Password: Your temporary password (change it after first login)
Contact Us

Step 2: Login to Enterprise Portal

Access the enterprise portal at:

Portal URL
https://www.deepcodereviewer.com/portal

Enter your enterprise name and password to log in. After your first login, we strongly recommend changing your password for security.

Step 3: Add Your Repository

Once logged in, navigate to the Repositories tab and click the "+ Add Repository" button.

Repository Management Page

Enterprise Portal - Repositories Page

Fill in the repository details in the form. DeepReview supports two authentication methods:

Option A: SSH Key Authentication

  1. Name: Enter a display name for your repository (e.g., "my-project")
  2. Authentication Method: Select "SSH Key"
  3. Repository URL: Enter the SSH URL (e.g., git@github.com:your-org/your-repo.git)
  4. SSH Private Key: Paste your SSH private key (must have read access to the repository)
  5. Enable code reviews: Check this box to activate automatic reviews
  6. Click "Add Repository" to save

Option B: Token Authentication

  1. Name: Enter a display name for your repository (e.g., "my-project")
  2. Authentication Method: Select "Token"
  3. Git Clone Command: Paste the full clone command with embedded credentials. The required format is:
    git clone https://<username>:<deploy_token>@gitlab.example.com/group/project.git
    Important: The input must follow the exact format above — starting with git clone, followed by an HTTPS URL containing credentials (username:token@host). Token credentials will be encrypted and stored securely. The repository URL will be automatically extracted from the clone command.
  4. Enable code reviews: Check this box to activate automatic reviews
  5. Click "Add Repository" to save
Add Repository Form

Add Repository Form

SSH Key Setup: Generate a deploy key for your repository:
ssh-keygen -t ed25519 -C "deepreview@your-company.com" -f deploy_key
Add the public key (deploy_key.pub) to your Git platform as a Deploy Key with read permissions, and paste the private key (deploy_key) content in the portal form.
Token Setup: For platforms that support deploy tokens (GitLab, JihuLab, etc.):
  1. Go to your repository or group → SettingsRepositoryDeploy tokens
  2. Create a token with read_repository scope
  3. Use the generated username and token in the clone command format above
All credentials are encrypted and stored securely.

Step 4: Configure Webhook

Set up a webhook in your Git platform to notify DeepReview when code is pushed:

Webhook URL
https://www.deepcodereviewer.com/api/webhook/<your-enterprise-name>

GitHub Setup

  1. Go to your repository → SettingsWebhooksAdd webhook
  2. Set Payload URL to the webhook URL above
  3. Set Content type to application/json
  4. Select "Just the push event"
  5. Click Add webhook

GitLab Setup

  1. Go to your project → SettingsWebhooks
  2. Enter the webhook URL above
  3. Check Push events
  4. Click Add webhook

Gitee Setup

  1. Go to your repository → ManagementWebHooksAdd WebHook
  2. Enter the webhook URL above
  3. Select Push event
  4. Click Add

Step 5: Push Code and Get Reviews

That's it! Now when you push code to your repository, DeepReview will automatically:

  1. Receive the webhook notification from your Git platform
  2. Clone your repository and analyze the code changes
  3. Use the ReAct AI engine to perform deep code review
  4. Generate a detailed review report within minutes

View all your review tasks in the enterprise portal under the Review Tasks tab:

Review Tasks Page

Code Review Tasks - Monitor review status and access reports

Click "View Report" on any completed task to see the detailed AI code review:

Code Review Report

Detailed Code Review Report - Including change summary, file analysis, and AI recommendations

Key Advantage: DeepReview analyzes your entire codebase for context, not just the diff. This enables detection of cross-file issues, architectural problems, and potential bugs that traditional code review tools might miss.

Installation

DeepReview is available as a cloud-based SaaS platform for seamless integration with your development workflow.

Cloud SaaS Platform

The fastest and easiest way to get started. No infrastructure management required.

  1. Contact our sales team to create your enterprise account
  2. Receive your enterprise credentials (name and initial password)
  3. Login to the enterprise portal at www.deepcodereviewer.com/portal
  4. Add your repositories with SSH keys or deploy tokens
  5. Configure webhooks in your Git platform
  6. Start receiving AI-powered code reviews automatically
Benefits of Cloud SaaS:
  • Zero infrastructure management
  • Automatic updates and new features
  • High availability and scalability
  • Enterprise-grade security
  • 24/7 technical support

Self-Hosted Deployment (Enterprise)

For organizations with specific compliance or data sovereignty requirements, private deployment options are available.

📞 Contact Sales Team

Self-hosted deployment requires an Enterprise license. Please contact our sales team to discuss your requirements and get deployment assistance.

Contact Sales

ReAct Engine

The ReAct (Reasoning and Action) engine is the core of DeepReview's intelligent code review capability. Unlike traditional static analyzers that apply fixed rules, ReAct enables the AI to think, explore, and iterate until it fully understands the code changes.

How It Works

The ReAct engine operates in iterative cycles, similar to how a human reviewer would approach code:

1
Think

Analyze the diff and reason about what context is needed

2
Act

Use tools to gather information (read files, search code, fetch docs)

3
Observe

Integrate findings and decide: loop back or generate report

Example ReAct Flow

ReAct Execution Trace
# Thought 1
The PR modifies the authentication function. I need to understand 
how the password hashing is implemented.

# Action 1
read_file("src/auth/password.ts")

# Observation 1
File contains bcrypt implementation with salt rounds = 10

# Thought 2
I see a potential security issue - the salt rounds are hardcoded.
Let me check if there's a config file for security settings.

# Action 2
search_code("salt.*rounds", "src/", "*.ts")

# Observation 2
Found: config/security.ts defines SALT_ROUNDS constant

# Thought 3
The PR should reference the config constant instead of hardcoding.
I have enough context to generate the review.

Available Tools

The AI agent has access to the following tools during review:

Tool Description Example Use
read_file Read the contents of any file Understanding full function implementation
search_code Search for patterns in codebase Finding all usages of a function
list_directory Explore project structure Understanding module organization
think Internal reasoning scratchpad Complex logical reasoning
Key Advantage: Unlike diff-only tools, DeepReview can trace function calls, understand architectural patterns, and detect issues that require cross-file context.

Review Workflow

Understanding the complete flow from PR creation to review completion.

Synchronous vs Asynchronous Mode

Mode Best For Response Time
Synchronous Small PRs (<500 lines) 30-60 seconds
Asynchronous Large PRs, deep analysis 2-10 minutes (webhook callback)

Review Output Structure

Each review generates multiple outputs:

  • Detailed Report: Full analysis with all findings
  • Summary: Executive summary for quick overview
  • Suggestions List: Actionable items with severity levels
  • Inline Comments: Posted directly on PR lines

Severity Levels

Level Icon Description
Critical 🔒 Security vulnerabilities, data loss risks
Warning Performance issues, potential bugs
Info 💡 Code style, refactoring suggestions

Retry & Fault Tolerance

DeepReview implements multiple layers of fault tolerance to ensure reliable reviews.

Retry Strategy

Error Type Strategy Max Retries
Network timeout Exponential backoff 5
Rate limit Wait + retry 5
Context too long Switch to fallback model 1
Model unavailable Exponential backoff 5