# Design a code search engine like GitHub search: system design interview question

> Code search with literal matches, symbol definitions, current repository snapshots, and strict access checks.

Source: https://sysround.com/questions/design-code-search-engine-like-github-search
Updated: 2026-11-05


Difficulty: hard. Levels: Senior, Staff+.

## How the interviewer opens it

"Design a service that searches source code across many repositories. Users need text search, symbol search, and repository filters, with current results that respect repository permissions."

## What to clarify first

- Corpus size
- Search behavior
- Traffic and read:write mix
- Latency and expensive queries
- Freshness and consistency
- Repository changes
- Repository permissions
- Durability and availability
- Result details
- Scope limits

## Scope of a strong answer

Core requirements:

- Search literal text and symbol definitions across repositories.
- Filter results by repository, path prefix, and language.
- Keep search results current as code changes.
- Return results only from repositories that the user can access.

Usually out of scope:

- Repository hosting and code edits
- Code execution and build services
- Semantic search with embeddings
- Historical search across all commits
- File-level permissions

## What interviewers probe

### Text and symbol results

Code contains punctuation and partial names. Short queries match many files, while case rules can change which matches the service returns.

### Code changes

Events can repeat or disappear. Search must publish complete repository states while keeping branch-head lag within the target.

### Repository access

Search can reveal private code through snippets, names, counts, or cached responses. Access revocation must take effect independently of code changes.

### Search across many repositories

Broad queries can contact many shards. One slow shard can dominate latency, while common terms can consume most search capacity.

## Common mistakes

- Use a text index without clear rules for punctuation, short queries, and case matches.
- Treat symbol search as plain text search.
- Update files separately without complete repository snapshots or deletion rules.
- Define freshness per commit, then skip commits without a branch-head lag rule.
- Tie access revocation to code freshness or share cached results across different access scopes.
- Ignore query fan-out or hide partial results after a timeout.

## The bar by level

- **Senior:** Handles short and case-insensitive queries. Publishes complete repository snapshots despite late events and force pushes. Prevents cache leaks and bounds query cost.
- **Staff+:** Defines access and snapshot rules across failures. Quantifies branch-head lag, recovery capacity, and query fan-out. Adapts repository placement without weakening those rules.

Practice this question with an AI interviewer: https://sysround.com/practice/new?problem=code-search


---
SysRound (https://sysround.com) is an AI practice tool for system design interviews. It has a voice or typed AI interviewer and a shared whiteboard. At the end, you get feedback on each skill and a model answer. The AI feedback can be wrong.