# Indx — llms.txt

## What Indx Is

Indx is a high-performance search engine designed for developers.
It operates directly on structured and unstructured text (typically JSON) and
focuses on fast, reliable retrieval with minimal configuration.

Indx is built to work well for:
- Exact term queries
- Partial and imperfect input
- Large datasets with predictable performance
- Both human-facing search and agent/tool usage

Indx can be embedded directly into applications or accessed via hosted services.

---

## Developer Experience

Indx eliminates the complexity typically associated with search engines.

You won't need to deal with:
- **Stop words** — all words are treated equally
- **Stemming or lemmatization** — no linguistic processing configuration
- **Analyzer chains** — no tokenizer or filter tuning
- **Query parsing edge cases** — queries work as written
- **Language-specific configuration** — works consistently across languages

This means:
- Predictable behavior from the start
- Minimal configuration overhead
- Results that match developer intuition
- Easier debugging and troubleshooting

Indx is designed to work well by default, without requiring deep search engine expertise.

---

## Core Search Model

Indx uses a **two-step search model**. This model is central to how relevance,
ranking, and performance are achieved.

### Step 1: Pattern Matching (Always Executed)

- Runs across **all documents** in the dataset
- Detects textual and structural patterns in the data
- Produces a candidate result set
- Extremely fast and tolerant of variation in input

This step provides:
- Strong recall
- Robust handling of typos and formatting differences
- Reliable behavior without query preprocessing or tuning

Exact queries are handled naturally in this phase, just as efficiently as partial
or imperfect queries.

---

### Step 2: Coverage (Optional)

Coverage is a refinement phase applied to the top-K results from pattern matching
(default: 500).

Coverage evaluates **how completely the query is represented in each document**.
It can:

- Lift exact and near-exact matches
- Improve ranking quality
- Truncate low-quality results at a defined threshold

Coverage focuses on result quality rather than raw matching speed.

---

## Human vs Agent Usage

Coverage behavior can be tuned depending on how search results are consumed:

- **Human search**
  - Pattern matches + coverage matches
  - Supports exploration and discovery

- **Agent or tool usage**
  - Coverage-focused results
  - Reduces noise and false positives

This distinction is intentional and allows the same index to support different
consumers effectively.

---

## Field Configuration Model

Indx uses explicit field configuration rather than implicit inference.

Fields can be marked as:
- **Searchable** — participate in matching and scoring
- **Filterable** — used in filters and boosts
- **Facetable** — used for aggregations and counts
- **Sortable** — used for ordering results

At least one field must be searchable.
Fields used in filters or boosts must be marked filterable.

This model keeps behavior predictable and avoids hidden configuration.

---

## Performance Characteristics

- Search is fast and consistent across dataset sizes
- Coverage scales linearly with configured depth
- No per-query tuning or analyzer configuration is required
- Indexing and querying are deterministic and explainable

---

## Typical Use Cases

Indx is commonly used for:
- Application-embedded search
- Product and content catalogs
- Internal tools and dashboards
- Developer platforms
- Agent-assisted retrieval workflows

It is designed to be simple to integrate while remaining powerful enough for
production systems.

---

## Summary

Indx is a developer-focused search engine built around a **two-step retrieval
model**:
1. Fast, broad pattern matching
2. Optional coverage-based refinement

This approach provides strong performance, high result quality, and reliable
behavior across exact, partial, and imperfect queries, without requiring complex
configuration or tuning.

---

## Resources

- **NuGet Package**: https://www.nuget.org/packages/IndxSearchLib/
- **Ready-to-Deploy Server**: https://github.com/indxSearch/IndxCloudApi
- **React UI Components**: https://github.com/indxSearch/indx-intrface