> ## Documentation Index
> Fetch the complete documentation index at: https://gitquarry.micr.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Search Behavior

> Reference for the native-versus-discover split, host support, bounded discovery, and ranking semantics.

# Search Behavior

This page describes the behavioral contract behind gitquarry search.

## Native Path

No enhancement flags means:

* one GitHub repository search
* native ordering
* no hidden local reranking
* no README fetches

That is the boring path by design.

## Discover Path

`--mode discover` enables:

* staged candidate collection
* dedupe by `full_name`
* metadata enrichment
* optional README window enrichment
* explicit rank modes

The discover engine is bounded. There is no persistent index and no background cache.

## Ranking Semantics

Rank modes:

* `native`
* `query`
* `activity`
* `quality`
* `blended`

`blended` uses a normalized weighted average of query, activity, and quality components.

## Host Support

GitHub.com is the primary target.

Custom hosts are best effort and normalized into:

* canonical web host
* canonical API base URL

## Rate Limits and Concurrency

Default concurrency is `1`.

That is deliberate because GitHub secondary rate limits make aggressive fan-out a bad default.

When GitHub sends retry guidance through `retry-after` or `x-ratelimit-reset`, gitquarry respects that guidance before retrying.