Practical overview
When it fits
Use Codegraph when code discovery requires both semantic matching and relationships between code entities, rather than exact-name search alone.
Who it is for
- Developers exploring an unfamiliar codebase
- Teams that need to trace how components, classes or functions connect
- Maintainers investigating implementation flows across files and layers
Documented uses
- Ask plain-English questions about a repository and retrieve relevant code with structural context.
- Explore class and function relationships and understand how components connect.
- Trace an end-to-end flow, such as a request moving through service and repository layers.
Requirements & limits
The project is open source under the MIT license. Its documented setup requires an OpenAI API key and Neo4j credentials.
- The stack includes Python, Neo4j, OpenAI embeddings and LLM calls, Streamlit, and a Docker Compose setup.
- The documented inputs are a public GitHub repository URL or an absolute local folder path.
- Chunking, overlap, file extensions, result count and graph context are configurable in the interface or configuration.
- Code and graph data are stored in Neo4j, with database statistics and guarded cleanup available in the UI.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up Codegraph with Docker Compose
Prepare the documented Docker setup and open the Streamlit interface.
Before you start
- Docker and Docker Compose
- An OpenAI API key
- Neo4j connection settings
Steps
Clone the Codegraph repository and change into its directory.
Create a project-root .env file with Neo4j URI, username and password, OPENAI_API_KEY, and the documented chunking and model settings.
Run ./docker/docker-prerequisite-check.sh, then start the services with docker compose -f docker/docker-compose.yml up -d.
Open the Codegraph UI at http://localhost:8501.
Expected outcome
The documented setup brings up the application and Neo4j, with the UI at localhost:8501.
Guide limitations
- An OpenAI API key and Neo4j credentials are required.
- These are documented setup steps, not an independent execution test.
Index a repository and ask a code question
Use the interface workflow to ingest a codebase, build its indexes and submit a natural-language query.
Before you start
- A running Codegraph instance
- A configured OpenAI API key and Neo4j connection
- A public GitHub repository URL or absolute local folder path
Steps
Enter a GitHub URL or local folder path in the Codebase input tab.
Configure chunk size, overlap and file extensions in Ingestion, then run parsing.
Run entity extraction in Knowledge graph and generate embeddings in Vector index.
Use the Query tab to ask a question about the codebase and review the response with code context and similarity scores.
Expected outcome
The interface returns a natural-language response grounded in semantic matches and related graph structure.
Guide limitations
- The workflow uses OpenAI for embeddings and LLM tasks.
- This describes the repository's documented workflow, not a tested result.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 16 October 2026
- Original website title
- Abhishek-Aditya-bs/CodeGraph · Official homepage title · retained in its source language