Quick Start Guide

Welcome to Unoplat Code Confluence - Your Gateway to Code Understanding!

Prerequisites

Before you begin, ensure you have the following tool installed:

Setting Up Unoplat Code Confluence

Follow these steps to set up Unoplat Code Confluence:

1. Create Project Directory and Download Configuration:

   mkdir -p code-confluence && cd code-confluence
   
   # Download Docker Compose file
   curl -O https://raw.githubusercontent.com/unoplat/unoplat-code-confluence/refs/heads/main/prod-docker-compose.yml

2. Launch Services:

   docker compose -f prod-docker-compose.yml up -d

3. Verify Deployment:

   docker compose -f prod-docker-compose.yml ps

Running the Application

1. Access the Web Interface

• Open your browser and go to http://localhost:3000

2. Connect a Repository Provider

• When prompted, select your repository provider and enter a Personal Access Token (PAT) with read access to the repositories you want to ingest.
• This is required for Unoplat Code Confluence to clone and analyze your repositories.

Unoplat Code Confluence supports two provider options:

GitHub

For GitHub.com and standard GitHub Enterprise Cloud users:

1. Select GitHub from the Provider dropdown
2. Enter your Personal Access Token
3. Click Connect

GitHub Enterprise

For Enterprise Server (self-hosted) or Enterprise Cloud with data residency:

1. Select GitHub Enterprise from the Provider dropdown
2. Enter your Base URL (e.g., https://yourcompany.ghe.com or https://github.mycompany.com)
3. Enter your Personal Access Token
4. Click Connect

3. Search for GitHub Repositories

• Press s or S (Shift + s) to open the search dialog.
• You can search through your personal, public, and open source repositories.

4. Ingest Repository

• In the Actions column, click on Ingest Repo for the repository you want to analyze.
• Unoplat Code Confluence will automatically detect Python and TypeScript codebases as part of repository along with corresponding package manager metadata.

5. Configure Model Provider

After ingestion completes, configure your AI model provider to enable code intelligence features.

• Navigate to Settings > Model Providers in the application
• Unoplat Code Confluence supports multiple AI model providers for code analysis and understanding

Recommended Models

We recommend the following models for optimal performance and code understanding:

Setup Steps:

1. Select your preferred Provider from the dropdown
2. Enter your API key in the API key field
3. Set the Model name to your chosen model
4. Click Save Configuration

6. Web Search for Documentation Lookup

Unoplat Code Confluence AI agents use web search to find official documentation, library references, and best practices during code analysis. The search provider is resolved automatically in the following priority order:

1. Exa — Used when an Exa API key is configured (recommended for best results)
2. Built-in Web Search — Used when the configured model provider supports native web search (Anthropic, OpenAI Codex, Grok, Groq)
3. DuckDuckGo — Automatic fallback when neither Exa nor built-in web search is available

Optional: Configure Exa for Enhanced Search

For the best documentation search results, configure Exa Search:

1. Navigate to Settings > Tool Configuration in the application
2. Click Configure next to Exa Search
3. Enter your Exa API key
4. Click Save

7. Repository Operations

After successfully ingesting your repository, you can perform various operations to manage your codebase. Navigate to Ingestion Management to access these operations.

Unoplat Code Confluence provides three key repository operations:

• Generate Agents.md — Automatically generates required metadata for AI agents to understand your codebase and operate on it effectively.
• Refresh Repository — Updates your ingested codebase to reflect the latest changes from the source repository
• Delete Repository — Removes the ingested repository and all associated data from Unoplat Code Confluence

AGENTS.md

When you trigger the Generate Agents.md operation:

1. Navigate to Workspace > Operations Management in the sidebar
2. Find your Agents Generation job in the operations table
3. Click the Actions menu (...) for your job and select View Details

You can monitor the progress in real-time. The workflow shows progress for each codebase individually, as well as overall progress when your repository contains multiple codebases. You can view agent events in real-time as actions are performed.

You can continue ingesting and generating AGENTS.md for multiple repositories in parallel until you reach the maximum tokens per second limit of your model provider.

The generated AGENTS.md provides structured documentation covering:

• Project Structure — Overview of your codebase organization
• Development Workflow — Development practices and patterns
• Business Logic — Core functionality and domain logic