diff --git a/README.md b/README.md
index 4927b3e..e7d3189 100644
--- a/README.md
+++ b/README.md
@@ -2,6 +2,9 @@
 
 ![](https://github.com/ringsaturn/tzf/blob/gh-pages/docs/tzf-social-media.png?raw=true)
 
+**NOTE**: this package use a simplified shape data so not so accurate around
+border.
+
 ## Build options
 
 By default, the binary is built as well. If you don't want/need it, then build
@@ -56,8 +59,8 @@ Here is what has been done to improve performance:
 1. Using pre-indexing to handle most queries takes approximately 1000
    nanoseconds.
 2. Using a finely-tuned Ray Casting algorithm package
-   <https://github.com/ringsaturn/geometry-rs> to verify whether a polygon
-   contains a point.
+   [`ringsaturn/geometry-rs`](https://github.com/ringsaturn/geometry-rs) to
+   verify whether a polygon contains a point.
 
 That's all. There are no black magic tricks inside the tzf-rs.
 
@@ -68,10 +71,10 @@ ns per query:
 test benches_default::bench_default_finder_random_city ... bench:       2,870 ns/iter (+/- 182)
 ```
 
-| Criterion result | Pic                              |
-| ---------------- | -------------------------------- |
-| PDF              | ![](assets/pdf_small.svg)        |
-| Regression       | ![](assets/regression_small.svg) |
+| Criterion result | Pic                                                                                       |
+| ---------------- | ----------------------------------------------------------------------------------------- |
+| PDF              | ![](https://raw.githubusercontent.com/ringsaturn/tzf-rs/main/assets/pdf_small.svg)        |
+| Regression       | ![](https://raw.githubusercontent.com/ringsaturn/tzf-rs/main/assets/regression_small.svg) |
 
 You can view more details from latest benchmark from
 [GitHub Actions logs](https://github.com/ringsaturn/tzf-rs/actions/workflows/rust.yml).
diff --git a/src/lib.rs b/src/lib.rs
index 3ad8bfb..013b1ed 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -1,27 +1,5 @@
-//! Fast timezone finder for Rust.
-//!
-//! It's designed for high performance geo queries related services like
-//! weather forecast API. And most queries could return in very limited time,
-//! averagely like 2000 nanoseconds.
-//!
-//! Please note that this package use a simplified shape data so not so accurate
-//! around border.
-//!
-//! There are there finders implements:
-//! - [Finder]: works anywhere.
-//! - [FuzzyFinder]: blazing fast for most places on earth, use a preindex data.
-//!   Not work for places around borders.
-//! - [DefaultFinder]: combine both, if [FuzzyFinder] got no data, then use [Finder].
-//!
-//! Preprocessed timezone data is distributed via [tzf-rel].
-//!
-//! It's Rust port of [tzf] and also the foundation of
-//! [tzfpy] since `v0.11.0`....
-//!
-//! [tzf]: https://github.com/ringsaturn/tzf
-//! [tzf-rel]: https://github.com/ringsaturn/tzf-rel
-//! [tzfpy]: https://github.com/ringsaturn/tzfpy
-//
+#![doc = include_str!("../README.md")]
+
 use geometry_rs::{Point, Polygon};
 use std::collections::HashMap;
 use std::f64::consts::PI;
@@ -45,6 +23,8 @@ impl Item {
     }
 }
 
+/// Finder works anywhere.
+///
 /// Finder use a fine tuned Ray casting algorithm implement [geometry-rs]
 /// which is Rust port of [geometry] by [Josh Baker].
 ///
@@ -203,6 +183,9 @@ pub fn deg2num(lng: f64, lat: f64, zoom: i64) -> (i64, i64) {
     return (xtile as i64, ytile as i64);
 }
 
+/// FuzzyFinder blazing fast for most places on earth, use a preindex data.
+/// Not work for places around borders.
+///
 /// FuzzyFinder store all preindex's tiles data in a HashMap,
 /// It iterate all zoom levels for input's longitude and latitude to build
 /// map key to to check if in map.
@@ -281,7 +264,8 @@ impl FuzzyFinder {
     }
 }
 
-/// It's most recommend to use, combine both [Finder] and [FuzzyFinder].
+/// It's most recommend to use, combine both [Finder] and [FuzzyFinder],
+/// if [FuzzyFinder] got no data, then use [Finder].
 pub struct DefaultFinder {
     finder: Finder,
     fuzzy_finder: FuzzyFinder,