diff --git a/rtfs/_types/results.py b/rtfs/_types/results.py
index c01ce7c..d29ccff 100644
--- a/rtfs/_types/results.py
+++ b/rtfs/_types/results.py
@@ -9,6 +9,6 @@ class NodeResponse(TypedDict):
 
 
 class Response(TypedDict):
-    results: dict[str, NodeResponse]
+    results: dict[str, NodeResponse] | None
     query_time: float
     commit_sha: str | None
diff --git a/rtfs/app.py b/rtfs/app.py
index 9b4bbfe..03e3fc6 100644
--- a/rtfs/app.py
+++ b/rtfs/app.py
@@ -53,24 +53,16 @@ def current_rtfs(state: State) -> Indexes:
 
 
 @get(path="/", dependencies={"rtfs": Provide(current_rtfs, sync_to_thread=False)})
-async def get_rtfs(query: dict[str, str], rtfs: Indexes) -> Response[Mapping[str, Any]]:
-    query_search = query.get("search")
-    library = query.get("library", "").lower()
-    format_ = query.get("format", "url")
-
-    if not query_search or not library:
+async def get_rtfs(search: str, library: str, direct: bool | None, rtfs: Indexes) -> Response[Mapping[str, Any]]:
+    if not search or not library:
         return Response(content={"available_libraries": rtfs.libraries}, media_type=MediaType.JSON, status_code=200)
-    elif not query_search:
+    elif not search:
         return Response({"error": "Missing `search` query parameter."}, media_type=MediaType.JSON, status_code=400)
     elif not library:
         return Response({"error": "Missing `library` query parameter."}, media_type=MediaType.JSON, status_code=400)
 
-    if format_ not in ("url", "source"):
-        return Response(
-            {"error": "The `format` parameter must be `url` or `source`."}, media_type=MediaType.JSON, status_code=400
-        )
+    result = rtfs.get_direct(library, search) if direct else rtfs.get_query(library, search)
 
-    result = rtfs.get_query(library, query_search)
     if result is None:
         return Response(
             content={
diff --git a/rtfs/fuzzy.py b/rtfs/fuzzy.py
new file mode 100644
index 0000000..4c2a774
--- /dev/null
+++ b/rtfs/fuzzy.py
@@ -0,0 +1,345 @@
+"""
+This Source Code Form is subject to the terms of the Mozilla Public
+License, v. 2.0. If a copy of the MPL was not distributed with this
+file, You can obtain one at http://mozilla.org/MPL/2.0/.
+"""
+
+# help with: http://chairnerd.seatgeek.com/fuzzywuzzy-fuzzy-string-matching-in-python/
+
+from __future__ import annotations
+
+import heapq
+import re
+from difflib import SequenceMatcher
+from typing import TYPE_CHECKING, Literal, TypeVar, overload
+
+if TYPE_CHECKING:
+    from collections.abc import Callable, Generator, Iterable, Sequence
+
+T = TypeVar("T")
+
+
+def ratio(a: str, b: str) -> int:
+    m = SequenceMatcher(None, a, b)
+    return int(round(100 * m.ratio()))
+
+
+def quick_ratio(a: str, b: str) -> int:
+    m = SequenceMatcher(None, a, b)
+    return int(round(100 * m.quick_ratio()))
+
+
+def partial_ratio(a: str, b: str) -> int:
+    short, long = (a, b) if len(a) <= len(b) else (b, a)
+    m = SequenceMatcher(None, short, long)
+
+    blocks = m.get_matching_blocks()
+
+    scores: list[float] = []
+    for i, j, _ in blocks:
+        start = max(j - i, 0)
+        end = start + len(short)
+        o = SequenceMatcher(None, short, long[start:end])
+        r = o.ratio()
+
+        if 100 * r > 99:
+            return 100
+        scores.append(r)
+
+    return int(round(100 * max(scores)))
+
+
+_word_regex = re.compile(r"\W", re.IGNORECASE)
+
+
+def _sort_tokens(a: str) -> str:
+    a = _word_regex.sub(" ", a).lower().strip()
+    return " ".join(sorted(a.split()))
+
+
+def token_sort_ratio(a: str, b: str) -> int:
+    a = _sort_tokens(a)
+    b = _sort_tokens(b)
+    return ratio(a, b)
+
+
+def quick_token_sort_ratio(a: str, b: str) -> int:
+    a = _sort_tokens(a)
+    b = _sort_tokens(b)
+    return quick_ratio(a, b)
+
+
+def partial_token_sort_ratio(a: str, b: str) -> int:
+    a = _sort_tokens(a)
+    b = _sort_tokens(b)
+    return partial_ratio(a, b)
+
+
+@overload
+def _extraction_generator(
+    query: str,
+    choices: Sequence[str],
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+) -> Generator[tuple[str, int], None, None]: ...
+
+
+@overload
+def _extraction_generator(
+    query: str,
+    choices: dict[str, T],
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+) -> Generator[tuple[str, int, T], None, None]: ...
+
+
+def _extraction_generator(
+    query: str,
+    choices: Sequence[str] | dict[str, T],
+    scorer: Callable[[str, str], int] = quick_ratio,
+    score_cutoff: int = 0,
+) -> Generator[tuple[str, int, T] | tuple[str, int], None, None]:
+    if isinstance(choices, dict):
+        for key, value in choices.items():
+            score = scorer(query, key)
+            if score >= score_cutoff:
+                yield (key, score, value)
+    else:
+        for choice in choices:
+            score = scorer(query, choice)
+            if score >= score_cutoff:
+                yield (choice, score)
+
+
+@overload
+def extract(
+    query: str,
+    choices: Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+    limit: int | None = ...,
+) -> list[tuple[str, int]]: ...
+
+
+@overload
+def extract(
+    query: str,
+    choices: dict[str, T],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+    limit: int | None = ...,
+) -> list[tuple[str, int, T]]: ...
+
+
+def extract(
+    query: str,
+    choices: dict[str, T] | Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = quick_ratio,
+    score_cutoff: int = 0,
+    limit: int | None = 10,
+) -> list[tuple[str, int]] | list[tuple[str, int, T]]:
+    it = _extraction_generator(query, choices, scorer, score_cutoff)
+    key = lambda t: t[1]
+    if limit is not None:
+        return heapq.nlargest(limit, it, key=key)  # type: ignore
+    return sorted(it, key=key, reverse=True)  # type: ignore
+
+
+@overload
+def extract_one(
+    query: str,
+    choices: Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+) -> tuple[str, int] | None: ...
+
+
+@overload
+def extract_one(
+    query: str,
+    choices: dict[str, T],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+) -> tuple[str, int, T] | None: ...
+
+
+def extract_one(
+    query: str,
+    choices: dict[str, T] | Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = quick_ratio,
+    score_cutoff: int = 0,
+) -> tuple[str, int] | None | tuple[str, int, T] | None:
+    it = _extraction_generator(query, choices, scorer, score_cutoff)
+    key = lambda t: t[1]
+    try:
+        return max(it, key=key)
+    except:
+        # iterator could return nothing
+        return None
+
+
+@overload
+def extract_or_exact(
+    query: str,
+    choices: Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+    limit: int | None = ...,
+) -> list[tuple[str, int]]: ...
+
+
+@overload
+def extract_or_exact(
+    query: str,
+    choices: dict[str, T],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+    limit: int | None = ...,
+) -> list[tuple[str, int, T]]: ...
+
+
+def extract_or_exact(
+    query: str,
+    choices: dict[str, T] | Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = quick_ratio,
+    score_cutoff: int = 0,
+    limit: int | None = None,
+) -> list[tuple[str, int]] | list[tuple[str, int, T]]:
+    matches = extract(query, choices, scorer=scorer, score_cutoff=score_cutoff, limit=limit)
+    if len(matches) == 0:
+        return []
+
+    if len(matches) == 1:
+        return matches
+
+    top = matches[0][1]
+    second = matches[1][1]
+
+    # check if the top one is exact or more than 30% more correct than the top
+    if top == 100 or top > (second + 30):
+        return [matches[0]]  # type: ignore
+
+    return matches
+
+
+@overload
+def extract_matches(
+    query: str,
+    choices: Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+) -> list[tuple[str, int]]: ...
+
+
+@overload
+def extract_matches(
+    query: str,
+    choices: dict[str, T],
+    *,
+    scorer: Callable[[str, str], int] = ...,
+    score_cutoff: int = ...,
+) -> list[tuple[str, int, T]]: ...
+
+
+def extract_matches(
+    query: str,
+    choices: dict[str, T] | Sequence[str],
+    *,
+    scorer: Callable[[str, str], int] = quick_ratio,
+    score_cutoff: int = 0,
+) -> list[tuple[str, int]] | list[tuple[str, int, T]]:
+    matches = extract(query, choices, scorer=scorer, score_cutoff=score_cutoff, limit=None)
+    if len(matches) == 0:
+        return []
+
+    top_score = matches[0][1]
+    to_return = []
+    index = 0
+    while True:
+        try:
+            match = matches[index]
+        except IndexError:
+            break
+        else:
+            index += 1
+
+        if match[1] != top_score:
+            break
+
+        to_return.append(match)
+    return to_return
+
+
+@overload
+def finder(
+    text: str,
+    collection: Iterable[T],
+    *,
+    key: Callable[[T], str] | None = ...,
+    raw: Literal[True],
+) -> list[tuple[int, int, T]]: ...
+
+
+@overload
+def finder(
+    text: str,
+    collection: Iterable[T],
+    *,
+    key: Callable[[T], str] | None = ...,
+    raw: Literal[False],
+) -> list[T]: ...
+
+
+@overload
+def finder(
+    text: str,
+    collection: Iterable[T],
+    *,
+    key: Callable[[T], str] | None = ...,
+    raw: bool = ...,
+) -> list[T]: ...
+
+
+def finder(
+    text: str,
+    collection: Iterable[T],
+    *,
+    key: Callable[[T], str] | None = None,
+    raw: bool = False,
+) -> list[tuple[int, int, T]] | list[T]:
+    suggestions: list[tuple[int, int, T]] = []
+    text = str(text)
+    pat = ".*?".join(map(re.escape, text))
+    regex = re.compile(pat, flags=re.IGNORECASE)
+    for item in collection:
+        to_search = key(item) if key else str(item)
+        r = regex.search(to_search)
+        if r:
+            suggestions.append((len(r.group()), r.start(), item))
+
+    def sort_key(tup: tuple[int, int, T]) -> tuple[int, int, str | T]:
+        if key:
+            return tup[0], tup[1], key(tup[2])
+        return tup
+
+    if raw:
+        return sorted(suggestions, key=sort_key)
+    else:
+        return [z for _, _, z in sorted(suggestions, key=sort_key)]
+
+
+def find(text: str, collection: Iterable[str], *, key: Callable[[str], str] | None = None) -> str | None:
+    try:
+        return finder(text, collection, key=key)[0]
+    except IndexError:
+        return None
diff --git a/rtfs/index.py b/rtfs/index.py
index b171960..a717799 100644
--- a/rtfs/index.py
+++ b/rtfs/index.py
@@ -2,7 +2,6 @@
 
 import ast
 import configparser
-import difflib
 import logging
 import os
 import pathlib
@@ -11,6 +10,8 @@
 
 from yarl import URL
 
+from .fuzzy import extract
+
 LOGGER = logging.getLogger(__name__)
 LOGGER.setLevel(logging.DEBUG)
 VERSION_REGEX = re.compile(r"__version__\s*=\s*(?:'|\")((?:\w|\.)*)(?:'|\")")
@@ -272,4 +273,4 @@ def index_lib(self) -> None:
                 )
 
     def find_matches(self, word: str) -> list[Node]:
-        return [self.nodes[v] for v in difflib.get_close_matches(word, self.keys, cutoff=0.55)]
+        return [self.nodes[v[0]] for v in extract(word, self.keys, score_cutoff=20, limit=3)]
diff --git a/rtfs/indexer.py b/rtfs/indexer.py
index 205a3d3..d224b7d 100644
--- a/rtfs/indexer.py
+++ b/rtfs/indexer.py
@@ -52,6 +52,23 @@ def get_query(self, lib: str, query: str) -> Response | None:
             "commit_sha": self.index[lib].commit,
         }
 
+    def get_direct(self, lib: str, query: str) -> Response | None:
+        if not self._is_indexed:
+            raise RuntimeError("Indexing is not complete.")
+
+        if lib not in self.index:
+            return
+
+        start = time.monotonic()
+        result = self.index[lib].nodes.get(query)
+        end = time.monotonic() - start
+
+        return {
+            "results": {result.name: {"source": result.source, "url": result.url}} if result else None,
+            "query_time": end,
+            "commit_sha": self.index[lib].commit,
+        }
+
     def _do_pull(self, index: Index) -> bool:
         try:
             subprocess.run(["/bin/bash", "-c", f"cd {index.repo_path} && git pull"])