diff --git a/guardrails/utils/tokenization_utils.py b/guardrails/utils/tokenization_utils.py
new file mode 100644
index 000000000..6dab87cdd
--- /dev/null
+++ b/guardrails/utils/tokenization_utils.py
@@ -0,0 +1,206 @@
+# This file contains code adapted from the WordTokenizers.jl
+# https://github.com/JuliaText/WordTokenizers.jl project.
+# It is subject to the license terms in the Apache License file
+# found in the top-level directory of this distribution.
+# This file has been modified by Guardrails AI on September 27 2024.
+
+import re
+
+
+def replace_til_no_change(input_text, pattern, replacement):
+    while True:
+        new_text = re.sub(pattern, replacement, input_text)
+        if new_text == input_text:
+            break
+        input_text = new_text
+    return input_text
+
+
+def postproc_splits(sentences, separator):
+    """
+    Applies heuristic rules to repair sentence splitting errors.
+    Developed for use as postprocessing for the GENIA sentence
+    splitter on PubMed abstracts, with minor tweaks for
+    full-text documents.
+
+    `sentences` should be a string, with line breaks on sentence boundaries.
+    Returns a similar string, but more correct.
+
+    Based on
+    https://github.com/ninjin/geniass/blob/master/geniass-postproc.pl
+    Which is
+    (c) 2010 Sampo Pyysalo. No rights reserved, i.e. do whatever you like with this.
+    Which draws in part on heuristics included in Yoshimasa Tsuruoka's
+    medss.pl script.
+    """
+
+    # Remove Windows line endings
+    sentences = sentences.replace("\r", "")
+
+    # Breaks sometimes missing after "?", "safe" cases
+    sentences = re.sub(
+        r"\b([a-z]+\?)\s+([A-Z][a-z]+)\b", rf"\1{separator}\2", sentences
+    )
+    # Breaks sometimes missing after ".", "safe" cases
+    sentences = re.sub(
+        r"\b([a-z]+ \.)\s+([A-Z][a-z]+)\b", rf"\1{separator}\2", sentences
+    )
+
+    # No breaks producing lines only containing sentence-ending punctuation
+    sentences = re.sub(rf"{separator}([.!?]+){separator}", r"\1" + separator, sentences)
+
+    # No breaks inside parentheses/brackets
+    sentences = replace_til_no_change(
+        sentences,
+        r"\[([^\[\]\(\)]*)" + re.escape(separator) + r"([^\[\]\(\)]*)\]",
+        r"[\1 \2]",
+    )
+    sentences = replace_til_no_change(
+        sentences,
+        r"\(([^\[\]\(\)]*)" + re.escape(separator) + r"([^\[\]\(\)]*)\)",
+        r"(\1 \2)",
+    )
+    # Standard mismatched with possible intervening
+    sentences = replace_til_no_change(
+        sentences,
+        r"\[([^\[\]]{0,250})" + re.escape(separator) + r"([^\[\]]{0,250})\]",
+        r"[\1 \2]",
+    )
+    sentences = replace_til_no_change(
+        sentences,
+        r"\(([^\(\)]{0,250})" + re.escape(separator) + r"([^\(\)]{0,250})\)",
+        r"(\1 \2)",
+    )
+
+    # Line breaks within quotes
+    sentences = replace_til_no_change(
+        sentences,
+        r'"([^"\n]{0,250})' + re.escape(separator) + r'([^"\n]{0,250})"',
+        r'"\1 \2"',
+    )
+    sentences = replace_til_no_change(
+        sentences,
+        r"'([^'\n]{0,250})" + re.escape(separator) + r"([^'\n]{0,250})'",
+        r"'\1 \2'",
+    )
+
+    # Nesting to depth one
+    sentences = replace_til_no_change(
+        sentences,
+        r"\[((?:[^\[\]]|\[[^\[\]]*\]){0,250})"
+        + re.escape(separator)
+        + r"((?:[^\[\]]|\[[^\[\]]*\]){0,250})\]",
+        r"[\1 \2]",
+    )
+    sentences = replace_til_no_change(
+        sentences,
+        r"\(((?:[^\(\)]|\([^\(\)]*\)){0,250})"
+        + re.escape(separator)
+        + r"((?:[^\(\)]|\([^\(\)]*\)){0,250})\)",
+        r"(\1 \2)",
+    )
+
+    # No break after periods followed by a non-uppercase "normal word"
+    sentences = re.sub(rf"\.{separator}([a-z]{{3,}}[a-z-]*[ .:,])", r". \1", sentences)
+
+    # No break after a single letter other than I
+    sentences = re.sub(rf"(\b[A-HJ-Z]\.){separator}", r"\1 ", sentences)
+
+    # No break before coordinating conjunctions (CC)
+    coordinating_conjunctions = ["and", "or", "but", "nor", "yet"]
+    for cc in coordinating_conjunctions:
+        sentences = re.sub(rf"{separator}({cc}\s)", r" \1", sentences)
+
+    # No break before prepositions (IN)
+    prepositions = [
+        "of",
+        "in",
+        "by",
+        "as",
+        "on",
+        "at",
+        "to",
+        "via",
+        "for",
+        "with",
+        "that",
+        "than",
+        "from",
+        "into",
+        "upon",
+        "after",
+        "while",
+        "during",
+        "within",
+        "through",
+        "between",
+        "whereas",
+        "whether",
+    ]
+    for prep in prepositions:
+        sentences = re.sub(rf"{separator}({prep}\s)", r" \1", sentences)
+
+    # No sentence breaks in the middle of specific abbreviations
+    sentences = re.sub(rf"(\be\.){separator}(g\.)", r"\1 \2", sentences)
+    sentences = re.sub(rf"(\bi\.){separator}(e\.)", r"\1 \2", sentences)
+    sentences = re.sub(rf"(\bi\.){separator}(v\.)", r"\1 \2", sentences)
+
+    # No sentence break after specific abbreviations
+    abbreviations = [
+        r"e\. ?g\.",
+        r"i\. ?e\.",
+        r"i\. ?v\.",
+        r"vs\.",
+        r"cf\.",
+        r"Dr\.",
+        r"Mr\.",
+        r"Ms\.",
+        r"Mrs\.",
+        r"Prof\.",
+        r"Ph\.?D\.",
+        r"Jr\.",
+        r"St\.",
+        r"Mt\.",
+        r"etc\.",
+        r"Fig\.",
+        r"vol\.",
+        r"Vols\.",
+        r"no\.",
+        r"Nos\.",
+        r"et\.",
+        r"al\.",
+        r"i\. ?v\.",
+        r"inc\.",
+        r"Ltd\.",
+        r"Co\.",
+        r"Corp\.",
+        r"Dept\.",
+        r"est\.",
+        r"Asst\.",
+        r"approx\.",
+        r"dr\.",
+        r"fig\.",
+        r"mr\.",
+        r"mrs\.",
+        r"ms\.",
+        r"prof\.",
+        r"rep\.",
+        r"jr\.",
+        r"sen\.",
+        r"st\.",
+        r"vs\.",
+        r"i\. ?e\.",
+    ]
+    for abbr in abbreviations:
+        sentences = re.sub(
+            rf"(\b{abbr}){separator}", r"\1", sentences, flags=re.IGNORECASE
+        )
+
+    return sentences
+
+
+def split_sentences(text, separator="abcdsentenceseperatordcba"):
+    # Use the separator in the regex
+    text = re.sub(r"([?!.])(?=\s|$)", rf"\1{separator}", text)
+    text = postproc_splits(text, separator)
+    return re.split(rf"\n?{separator} ?\n?", text)
diff --git a/guardrails/validator_base.py b/guardrails/validator_base.py
index d2c8e53db..ba86d2989 100644
--- a/guardrails/validator_base.py
+++ b/guardrails/validator_base.py
@@ -4,18 +4,19 @@
 #   - [ ] Remove validator_base.py in 0.6.x
 
 import asyncio
+import contextlib
 from functools import partial
 import inspect
 import logging
 from collections import defaultdict
 from dataclasses import dataclass
+import re
 from string import Template
 from typing import Any, Callable, Dict, List, Optional, Type, TypeVar, Union
 from typing_extensions import deprecated
 from warnings import warn
 import warnings
 
-import nltk
 import requests
 from langchain_core.runnables import Runnable
 
@@ -31,12 +32,9 @@
 from guardrails.types.on_fail import OnFailAction
 from guardrails.utils.safe_get import safe_get
 from guardrails.utils.hub_telemetry_utils import HubTelemetry
-
-#   See: https://github.com/guardrails-ai/guardrails/issues/829
-try:
-    nltk.data.find("tokenizers/punkt")
-except LookupError:
-    nltk.download("punkt")
+from guardrails.utils.tokenization_utils import (
+    postproc_splits,
+)
 
 
 ### functions to get chunks ###
@@ -48,21 +46,46 @@ def split_sentence_str(chunk: str):
     return [fragments[0] + ".", ".".join(fragments[1:])]
 
 
-def split_sentence_nltk(chunk: str):
+def split_sentence_word_tokenizers_jl_separator(
+    chunk: str, separator: str = "abcdsentenceseperatordcba"
+):
     """
-    NOTE: this approach currently does not work
-    Use a sentence tokenizer to split the chunk into sentences.
+    Use a sentence tokenizer to detect if at least one sentence is present in the chunk.
+    We return the first sentence and the remaining chunks without the first sentence.
+
+    We perform the first step of WordTokenizers.jl's split_sentences function to
+    detect possible sentence boundaries before calling the sentence tokenizer.
+
+    Args:
+        chunk (str): The text to split into sentences.
 
-    Because using the tokenizer is expensive, we only use it if there
-    is a period present in the chunk.
+    Returns:
+        List[str]: A list of two strings. The first string is the first sentence
+            in the chunk. The second string is the remaining text in the chunk.
     """
     # using the sentence tokenizer is expensive
     # we check for a . to avoid wastefully calling the tokenizer
-    if "." not in chunk:
+
+    # check at least 3 characters have been accumulated before splitting
+    is_minimum_length = False
+    with contextlib.suppress(IndexError):
+        chunk[2]
+        is_minimum_length = True
+
+    # check for potential line endings, which is what split_sentences does
+    chunk_with_potential_line_endings, count = re.subn(
+        r"([?!.])(?=\s|$)", rf"\1{separator}", chunk
+    )
+    any_potential_line_endings = count > 0
+    if not is_minimum_length or not any_potential_line_endings:
         return []
-    sentences = nltk.sent_tokenize(chunk)
-    if len(sentences) == 0:
+
+    sentences = postproc_splits(chunk_with_potential_line_endings, separator)
+    sentences = re.split(rf"\n?{separator} ?\n?", sentences)
+    # if not more than one sentence, we haven't accumulated enough for a validation
+    if len(sentences) <= 1:
         return []
+
     # return the sentence
     # then the remaining chunks that aren't finished accumulating
     return [sentences[0], "".join(sentences[1:])]
@@ -266,7 +289,7 @@ def _chunking_function(self, chunk: str) -> List[str]:
         Returns:
             list[str]: The text chunked into some subset.
         """
-        return split_sentence_str(chunk)
+        return split_sentence_word_tokenizers_jl_separator(chunk)
 
     def validate_stream(
         self, chunk: Any, metadata: Dict[str, Any], **kwargs
diff --git a/poetry.lock b/poetry.lock
index f2dabcbfb..eafa419f2 100644
--- a/poetry.lock
+++ b/poetry.lock
@@ -1,5 +1,21 @@
 # This file is automatically @generated by Poetry 1.8.4 and should not be changed by hand.
 
+[[package]]
+name = "aiocache"
+version = "0.12.3"
+description = "multi backend asyncio cache"
+optional = false
+python-versions = "*"
+files = [
+    {file = "aiocache-0.12.3-py2.py3-none-any.whl", hash = "sha256:889086fc24710f431937b87ad3720a289f7fc31c4fd8b68e9f918b9bacd8270d"},
+    {file = "aiocache-0.12.3.tar.gz", hash = "sha256:f528b27bf4d436b497a1d0d1a8f59a542c153ab1e37c3621713cb376d44c4713"},
+]
+
+[package.extras]
+memcached = ["aiomcache (>=0.5.2)"]
+msgpack = ["msgpack (>=0.5.5)"]
+redis = ["redis (>=4.2.0)"]
+
 [[package]]
 name = "aiohappyeyeballs"
 version = "2.4.3"
@@ -307,6 +323,23 @@ types-python-dateutil = ">=2.8.10"
 doc = ["doc8", "sphinx (>=7.0.0)", "sphinx-autobuild", "sphinx-autodoc-typehints", "sphinx_rtd_theme (>=1.3.0)"]
 test = ["dateparser (==1.*)", "pre-commit", "pytest", "pytest-cov", "pytest-mock", "pytz (==2021.1)", "simplejson (==3.*)"]
 
+[[package]]
+name = "asgiref"
+version = "3.8.1"
+description = "ASGI specs, helper code, and adapters"
+optional = false
+python-versions = ">=3.8"
+files = [
+    {file = "asgiref-3.8.1-py3-none-any.whl", hash = "sha256:3e1e3ecc849832fe52ccf2cb6686b7a55f82bb1d6aee72a58826471390335e47"},
+    {file = "asgiref-3.8.1.tar.gz", hash = "sha256:c343bd80a0bec947a9860adb4c432ffa7db769836c64238fc34bdc3fec84d590"},
+]
+
+[package.dependencies]
+typing-extensions = {version = ">=4", markers = "python_version < \"3.11\""}
+
+[package.extras]
+tests = ["mypy (>=0.800)", "pytest", "pytest-asyncio"]
+
 [[package]]
 name = "asttokens"
 version = "2.4.1"
@@ -489,7 +522,7 @@ css = ["tinycss2 (>=1.1.0,<1.3)"]
 name = "blinker"
 version = "1.8.2"
 description = "Fast, simple object-to-object and broadcast signaling"
-optional = false
+optional = true
 python-versions = ">=3.8"
 files = [
     {file = "blinker-1.8.2-py3-none-any.whl", hash = "sha256:1779309f71bf239144b9399d06ae925637cf6634cf6bd131104184531bf67c01"},
@@ -537,17 +570,6 @@ urllib3 = [
 [package.extras]
 crt = ["awscrt (==0.22.0)"]
 
-[[package]]
-name = "cachelib"
-version = "0.9.0"
-description = "A collection of cache libraries in the same API interface."
-optional = false
-python-versions = ">=3.7"
-files = [
-    {file = "cachelib-0.9.0-py3-none-any.whl", hash = "sha256:811ceeb1209d2fe51cd2b62810bd1eccf70feba5c52641532498be5c675493b3"},
-    {file = "cachelib-0.9.0.tar.gz", hash = "sha256:38222cc7c1b79a23606de5c2607f4925779e37cdcea1c2ad21b8bae94b5425a5"},
-]
-
 [[package]]
 name = "cachetools"
 version = "5.5.0"
@@ -1489,6 +1511,26 @@ files = [
 [package.dependencies]
 python-dateutil = ">=2.4"
 
+[[package]]
+name = "fastapi"
+version = "0.115.2"
+description = "FastAPI framework, high performance, easy to learn, fast to code, ready for production"
+optional = false
+python-versions = ">=3.8"
+files = [
+    {file = "fastapi-0.115.2-py3-none-any.whl", hash = "sha256:61704c71286579cc5a598763905928f24ee98bfcc07aabe84cfefb98812bbc86"},
+    {file = "fastapi-0.115.2.tar.gz", hash = "sha256:3995739e0b09fa12f984bce8fa9ae197b35d433750d3d312422d846e283697ee"},
+]
+
+[package.dependencies]
+pydantic = ">=1.7.4,<1.8 || >1.8,<1.8.1 || >1.8.1,<2.0.0 || >2.0.0,<2.0.1 || >2.0.1,<2.1.0 || >2.1.0,<3.0.0"
+starlette = ">=0.37.2,<0.41.0"
+typing-extensions = ">=4.8.0"
+
+[package.extras]
+all = ["email-validator (>=2.0.0)", "fastapi-cli[standard] (>=0.0.5)", "httpx (>=0.23.0)", "itsdangerous (>=1.1.0)", "jinja2 (>=2.11.2)", "orjson (>=3.2.1)", "pydantic-extra-types (>=2.0.0)", "pydantic-settings (>=2.0.0)", "python-multipart (>=0.0.7)", "pyyaml (>=5.3.1)", "ujson (>=4.0.1,!=4.0.2,!=4.1.0,!=4.2.0,!=4.3.0,!=5.0.0,!=5.1.0)", "uvicorn[standard] (>=0.12.0)"]
+standard = ["email-validator (>=2.0.0)", "fastapi-cli[standard] (>=0.0.5)", "httpx (>=0.23.0)", "jinja2 (>=2.11.2)", "python-multipart (>=0.0.7)", "uvicorn[standard] (>=0.12.0)"]
+
 [[package]]
 name = "fastcore"
 version = "1.4.2"
@@ -1559,7 +1601,7 @@ typing = ["typing-extensions (>=4.12.2)"]
 name = "flask"
 version = "3.0.3"
 description = "A simple framework for building complex web applications."
-optional = false
+optional = true
 python-versions = ">=3.8"
 files = [
     {file = "flask-3.0.3-py3-none-any.whl", hash = "sha256:34e815dfaa43340d1d15a5c3a02b8476004037eb4840b34910c6e21679d288f3"},
@@ -1578,50 +1620,6 @@ Werkzeug = ">=3.0.0"
 async = ["asgiref (>=3.2)"]
 dotenv = ["python-dotenv"]
 
-[[package]]
-name = "flask-caching"
-version = "2.3.0"
-description = "Adds caching support to Flask applications."
-optional = false
-python-versions = ">=3.8"
-files = [
-    {file = "Flask_Caching-2.3.0-py3-none-any.whl", hash = "sha256:51771c75682e5abc1483b78b96d9131d7941dc669b073852edfa319dd4e29b6e"},
-    {file = "flask_caching-2.3.0.tar.gz", hash = "sha256:d7e4ca64a33b49feb339fcdd17e6ba25f5e01168cf885e53790e885f83a4d2cf"},
-]
-
-[package.dependencies]
-cachelib = ">=0.9.0,<0.10.0"
-Flask = "*"
-
-[[package]]
-name = "flask-cors"
-version = "5.0.0"
-description = "A Flask extension adding a decorator for CORS support"
-optional = false
-python-versions = "*"
-files = [
-    {file = "Flask_Cors-5.0.0-py2.py3-none-any.whl", hash = "sha256:b9e307d082a9261c100d8fb0ba909eec6a228ed1b60a8315fd85f783d61910bc"},
-    {file = "flask_cors-5.0.0.tar.gz", hash = "sha256:5aadb4b950c4e93745034594d9f3ea6591f734bb3662e16e255ffbf5e89c88ef"},
-]
-
-[package.dependencies]
-Flask = ">=0.9"
-
-[[package]]
-name = "flask-sqlalchemy"
-version = "3.1.1"
-description = "Add SQLAlchemy support to your Flask application."
-optional = false
-python-versions = ">=3.8"
-files = [
-    {file = "flask_sqlalchemy-3.1.1-py3-none-any.whl", hash = "sha256:4ba4be7f419dc72f4efd8802d69974803c37259dd42f3913b0dcf75c9447e0a0"},
-    {file = "flask_sqlalchemy-3.1.1.tar.gz", hash = "sha256:e4b68bb881802dda1a7d878b2fc84c06d1ee57fb40b874d3dc97dabfa36b8312"},
-]
-
-[package.dependencies]
-flask = ">=2.2.5"
-sqlalchemy = ">=2.0.16"
-
 [[package]]
 name = "fonttools"
 version = "4.54.1"
@@ -2154,34 +2152,33 @@ protobuf = ["grpcio-tools (>=1.67.0)"]
 
 [[package]]
 name = "guardrails-api"
-version = "0.0.3"
+version = "0.1.0a1"
 description = "Guardrails API"
 optional = false
 python-versions = "<4,>=3.8"
 files = [
-    {file = "guardrails_api-0.0.3-py3-none-any.whl", hash = "sha256:22f2ae1ea34fb36f9407078cd81924f756968477b4f079f0d82b329b1441552f"},
-    {file = "guardrails_api-0.0.3.tar.gz", hash = "sha256:f4e9b192ee3f2bfcba1453f4cca9d68aca5ccfe22adc82f104c449b283136b5d"},
+    {file = "guardrails_api-0.1.0a1-py3-none-any.whl", hash = "sha256:4685447b2ffbddce77de14d165e28a12437154fabc203678bc33edb2670d647c"},
+    {file = "guardrails_api-0.1.0a1.tar.gz", hash = "sha256:0ea1e0b8ac7c240bb58f4649dd547a7772c94a245ff9701f9b688872d0bc3563"},
 ]
 
 [package.dependencies]
+aiocache = ">=0.11.1"
 boto3 = ">=1.34.115,<2"
-flask = ">=3.0.3,<4"
-Flask-Caching = ">=2.3.0,<3"
-Flask-Cors = ">=4.0.1,<6"
-Flask-SQLAlchemy = ">=3.1.1,<4"
-guardrails-ai = ">=0.5.6"
+fastapi = ">=0.114.1"
+guardrails-ai = ">=0.5.12"
 jsonschema = ">=4.22.0,<5"
 litellm = ">=1.39.3,<2"
 opentelemetry-api = ">=1.0.0,<2"
 opentelemetry-exporter-otlp-proto-grpc = ">=1.0.0,<2"
 opentelemetry-exporter-otlp-proto-http = ">=1.0.0,<2"
-opentelemetry-instrumentation-flask = ">=0.12b0,<1"
+opentelemetry-instrumentation-fastapi = ">=0.48b0"
 opentelemetry-sdk = ">=1.0.0,<2"
 psycopg2-binary = ">=2.9.9,<3"
 referencing = ">=0.35.1,<1"
 requests = ">=2.32.3"
+SQLAlchemy = ">=2.0.34"
 typer = ">=0.9.4,<1"
-Werkzeug = ">=3.0.3,<4"
+uvicorn = ">=0.30.6"
 
 [package.extras]
 dev = ["coverage", "gunicorn (>=22.0.0,<23)", "pytest", "pytest-mock", "ruff"]
@@ -2547,7 +2544,7 @@ arrow = ">=0.15.0"
 name = "itsdangerous"
 version = "2.2.0"
 description = "Safely pass data to untrusted environments and back."
-optional = false
+optional = true
 python-versions = ">=3.8"
 files = [
     {file = "itsdangerous-2.2.0-py3-none-any.whl", hash = "sha256:c6242fc49e35958c8b15141343aa660db5fc54d4f13a1db01a3f5891b98700ef"},
@@ -2760,7 +2757,7 @@ files = [
 name = "joblib"
 version = "1.4.2"
 description = "Lightweight pipelining with Python functions"
-optional = false
+optional = true
 python-versions = ">=3.8"
 files = [
     {file = "joblib-1.4.2-py3-none-any.whl", hash = "sha256:06d478d5674cbc267e7496a410ee875abd68e4340feff4490bcb7afb88060ae6"},
@@ -4403,31 +4400,6 @@ files = [
     {file = "nh3-0.2.18.tar.gz", hash = "sha256:94a166927e53972a9698af9542ace4e38b9de50c34352b962f4d9a7d4c927af4"},
 ]
 
-[[package]]
-name = "nltk"
-version = "3.8.1"
-description = "Natural Language Toolkit"
-optional = false
-python-versions = ">=3.7"
-files = [
-    {file = "nltk-3.8.1-py3-none-any.whl", hash = "sha256:fd5c9109f976fa86bcadba8f91e47f5e9293bd034474752e92a520f81c93dda5"},
-    {file = "nltk-3.8.1.zip", hash = "sha256:1834da3d0682cba4f2cede2f9aad6b0fafb6461ba451db0efb6f9c39798d64d3"},
-]
-
-[package.dependencies]
-click = "*"
-joblib = "*"
-regex = ">=2021.8.3"
-tqdm = "*"
-
-[package.extras]
-all = ["matplotlib", "numpy", "pyparsing", "python-crfsuite", "requests", "scikit-learn", "scipy", "twython"]
-corenlp = ["requests"]
-machine-learning = ["numpy", "python-crfsuite", "scikit-learn", "scipy"]
-plot = ["matplotlib"]
-tgrep = ["pyparsing"]
-twitter = ["twython"]
-
 [[package]]
 name = "nodeenv"
 version = "1.9.1"
@@ -4844,45 +4816,47 @@ setuptools = ">=16.0"
 wrapt = ">=1.0.0,<2.0.0"
 
 [[package]]
-name = "opentelemetry-instrumentation-flask"
+name = "opentelemetry-instrumentation-asgi"
 version = "0.48b0"
-description = "Flask instrumentation for OpenTelemetry"
+description = "ASGI instrumentation for OpenTelemetry"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "opentelemetry_instrumentation_flask-0.48b0-py3-none-any.whl", hash = "sha256:26b045420b9d76e85493b1c23fcf27517972423480dc6cf78fd6924248ba5808"},
-    {file = "opentelemetry_instrumentation_flask-0.48b0.tar.gz", hash = "sha256:e03a34428071aebf4864ea6c6a564acef64f88c13eb3818e64ea90da61266c3d"},
+    {file = "opentelemetry_instrumentation_asgi-0.48b0-py3-none-any.whl", hash = "sha256:ddb1b5fc800ae66e85a4e2eca4d9ecd66367a8c7b556169d9e7b57e10676e44d"},
+    {file = "opentelemetry_instrumentation_asgi-0.48b0.tar.gz", hash = "sha256:04c32174b23c7fa72ddfe192dad874954968a6a924608079af9952964ecdf785"},
 ]
 
 [package.dependencies]
-importlib-metadata = ">=4.0"
+asgiref = ">=3.0,<4.0"
 opentelemetry-api = ">=1.12,<2.0"
 opentelemetry-instrumentation = "0.48b0"
-opentelemetry-instrumentation-wsgi = "0.48b0"
 opentelemetry-semantic-conventions = "0.48b0"
 opentelemetry-util-http = "0.48b0"
-packaging = ">=21.0"
 
 [package.extras]
-instruments = ["flask (>=1.0)"]
+instruments = ["asgiref (>=3.0,<4.0)"]
 
 [[package]]
-name = "opentelemetry-instrumentation-wsgi"
+name = "opentelemetry-instrumentation-fastapi"
 version = "0.48b0"
-description = "WSGI Middleware for OpenTelemetry"
+description = "OpenTelemetry FastAPI Instrumentation"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "opentelemetry_instrumentation_wsgi-0.48b0-py3-none-any.whl", hash = "sha256:c6051124d741972090fe94b2fa302555e1e2a22e9cdda32dd39ed49a5b34e0c6"},
-    {file = "opentelemetry_instrumentation_wsgi-0.48b0.tar.gz", hash = "sha256:1a1e752367b0df4397e0b835839225ef5c2c3c053743a261551af13434fc4d51"},
+    {file = "opentelemetry_instrumentation_fastapi-0.48b0-py3-none-any.whl", hash = "sha256:afeb820a59e139d3e5d96619600f11ce0187658b8ae9e3480857dd790bc024f2"},
+    {file = "opentelemetry_instrumentation_fastapi-0.48b0.tar.gz", hash = "sha256:21a72563ea412c0b535815aeed75fc580240f1f02ebc72381cfab672648637a2"},
 ]
 
 [package.dependencies]
 opentelemetry-api = ">=1.12,<2.0"
 opentelemetry-instrumentation = "0.48b0"
+opentelemetry-instrumentation-asgi = "0.48b0"
 opentelemetry-semantic-conventions = "0.48b0"
 opentelemetry-util-http = "0.48b0"
 
+[package.extras]
+instruments = ["fastapi (>=0.58,<1.0)"]
+
 [[package]]
 name = "opentelemetry-proto"
 version = "1.27.0"
@@ -7471,6 +7445,24 @@ pure-eval = "*"
 [package.extras]
 tests = ["cython", "littleutils", "pygments", "pytest", "typeguard"]
 
+[[package]]
+name = "starlette"
+version = "0.40.0"
+description = "The little ASGI library that shines."
+optional = false
+python-versions = ">=3.8"
+files = [
+    {file = "starlette-0.40.0-py3-none-any.whl", hash = "sha256:c494a22fae73805376ea6bf88439783ecfba9aac88a43911b48c653437e784c4"},
+    {file = "starlette-0.40.0.tar.gz", hash = "sha256:1a3139688fb298ce5e2d661d37046a66ad996ce94be4d4983be019a23a04ea35"},
+]
+
+[package.dependencies]
+anyio = ">=3.4.0,<5"
+typing-extensions = {version = ">=3.10.0", markers = "python_version < \"3.10\""}
+
+[package.extras]
+full = ["httpx (>=0.22.0)", "itsdangerous", "jinja2", "python-multipart (>=0.0.7)", "pyyaml"]
+
 [[package]]
 name = "sympy"
 version = "1.13.1"
@@ -8105,6 +8097,25 @@ secure = ["certifi", "cryptography (>=1.9)", "idna (>=2.0.0)", "pyopenssl (>=17.
 socks = ["pysocks (>=1.5.6,!=1.5.7,<2.0)"]
 zstd = ["zstandard (>=0.18.0)"]
 
+[[package]]
+name = "uvicorn"
+version = "0.32.0"
+description = "The lightning-fast ASGI server."
+optional = false
+python-versions = ">=3.8"
+files = [
+    {file = "uvicorn-0.32.0-py3-none-any.whl", hash = "sha256:60b8f3a5ac027dcd31448f411ced12b5ef452c646f76f02f8cc3f25d8d26fd82"},
+    {file = "uvicorn-0.32.0.tar.gz", hash = "sha256:f78b36b143c16f54ccdb8190d0a26b5f1901fe5a3c777e1ab29f26391af8551e"},
+]
+
+[package.dependencies]
+click = ">=7.0"
+h11 = ">=0.8"
+typing-extensions = {version = ">=4.0", markers = "python_version < \"3.11\""}
+
+[package.extras]
+standard = ["colorama (>=0.4)", "httptools (>=0.5.0)", "python-dotenv (>=0.13)", "pyyaml (>=5.1)", "uvloop (>=0.14.0,!=0.15.0,!=0.15.1)", "watchfiles (>=0.13)", "websockets (>=10.4)"]
+
 [[package]]
 name = "uvloop"
 version = "0.20.0"
@@ -8283,7 +8294,7 @@ test = ["websockets"]
 name = "werkzeug"
 version = "3.0.4"
 description = "The comprehensive WSGI web application library."
-optional = false
+optional = true
 python-versions = ">=3.8"
 files = [
     {file = "werkzeug-3.0.4-py3-none-any.whl", hash = "sha256:02c9eb92b7d6c06f31a782811505d2157837cea66aaede3e217c7c27c039476c"},
@@ -8663,4 +8674,4 @@ vectordb = ["faiss-cpu", "numpy"]
 [metadata]
 lock-version = "2.0"
 python-versions = "^3.9"
-content-hash = "a662a95423df55ea429547ebc2d91c2fbd38fd3da3b7ef05ac9c790bbdb7b2e0"
+content-hash = "00dc0d05363d8bc86320d18440956938bb6554c7c861b7abf58c616789874fbc"
diff --git a/pyproject.toml b/pyproject.toml
index 625d4a962..8b6d8ac26 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "guardrails-ai"
-version = "0.5.15"
+version = "0.6.0-alpha1"
 description = "Adding guardrails to large language models."
 authors = ["Guardrails AI <contact@guardrailsai.com>"]
 license = "Apache License 2.0"
@@ -28,7 +28,6 @@ rstr = "^3.2.2"
 typing-extensions = "^4.8.0"
 python-dateutil = "^2.8.2"
 tiktoken = ">=0.5.1"
-nltk = ">3.0, <=3.8.1"
 litellm = "^1.37.14"
 sqlvalidator = {version = "^0.0.20", optional = true}
 sqlalchemy = {version = ">=2.0.9", optional = true}
@@ -58,7 +57,7 @@ opentelemetry-exporter-otlp-proto-http = "^1.24.0"
 guardrails-hub-types = "^0.0.4"
 guardrails-api-client = "^0.3.13"
 diff-match-patch = "^20230430"
-guardrails-api = "^0.0.3"
+guardrails-api = ">=0.1.0a1,<0.2.0"
 mlflow = {version = "^2.0.1", optional = true}
 uvloop = {version = "^0.20.0", optional = true}
 semver = "^3.0.2"
diff --git a/tests/integration_tests/test_assets/validators/detect_pii.py b/tests/integration_tests/test_assets/validators/detect_pii.py
index ddcfb87cd..2da94e7d6 100644
--- a/tests/integration_tests/test_assets/validators/detect_pii.py
+++ b/tests/integration_tests/test_assets/validators/detect_pii.py
@@ -1,6 +1,5 @@
 from typing import Any, Callable, Dict, List, Union
 import difflib
-import nltk
 
 from guardrails.validator_base import (
     FailResult,
@@ -65,6 +64,14 @@ def chunking_function(self, chunk: str):
         Because using the tokenizer is expensive, we only use it if
         there is a period present in the chunk.
         """
+        try:
+            import nltk
+        except ImportError:
+            raise ImportError(
+                "nltk is required for sentence splitting. Please install it using "
+                "`poetry add nltk`"
+            )
+
         # using the sentence tokenizer is expensive
         # we check for a . to avoid wastefully calling the tokenizer
         if "." not in chunk:
diff --git a/tests/integration_tests/test_async_streaming.py b/tests/integration_tests/test_async_streaming.py
index cb6fe18c0..e09ab47ff 100644
--- a/tests/integration_tests/test_async_streaming.py
+++ b/tests/integration_tests/test_async_streaming.py
@@ -176,8 +176,7 @@ async def test_async_streaming_fix_behavior_two_validators(mocker):
     assert (
         text
         == """<PERSON>, under golden bridges, roams,
-<LOCATION> hills, his home.
-dreams of fog, and salty air,
+<LOCATION> hills, his home.dreams of fog, and salty air,
 in his heart, he's always there."""
     )
     assert (
diff --git a/tests/integration_tests/test_streaming.py b/tests/integration_tests/test_streaming.py
index 9ae025146..cc3d92181 100644
--- a/tests/integration_tests/test_streaming.py
+++ b/tests/integration_tests/test_streaming.py
@@ -487,8 +487,7 @@ def test_fix_behavior_one_validator(mocker):
     assert (
         text
         == """"john, under golden bridges, roams,
-san francisco's hills, his home.
-dreams of fog, and salty air,
+san francisco's hills, his home.dreams of fog, and salty air,
 in his heart, he's always there."""
     )
     assert (
@@ -530,8 +529,7 @@ def test_fix_behavior_two_validators(mocker):
     assert (
         text
         == """"<PERSON>, under golden bridges, roams,
-<LOCATION> hills, his home.
-dreams of fog, and salty air,
+<LOCATION> hills, his home.dreams of fog, and salty air,
 in his heart, he's always there."""
     )
     assert (
@@ -584,8 +582,7 @@ def test_fix_behavior_three_validators(mocker):
     assert (
         text
         == """"REDACTED!!, under purple!! bridges, roams,
-<LOCATION> hills, his home.
-dreams of fog, and salty air,
+<LOCATION> hills, his home.dreams of fog, and salty air,
 in his heart, he's always there."""
     )
     assert (
diff --git a/tests/unit_tests/classes/test_rc.py b/tests/unit_tests/classes/test_rc.py
index fb71471f0..3ade2e8a7 100644
--- a/tests/unit_tests/classes/test_rc.py
+++ b/tests/unit_tests/classes/test_rc.py
@@ -4,11 +4,6 @@
 
 class TestRC:
     def test_load(self, mocker):
-        # TODO: Re-enable this once we move nltk.download calls to individual validator repos.  # noqa
-        # Right now, it fires during our import chain, causing this to blow up
-        mocker.patch("nltk.data.find")
-        mocker.patch("nltk.download")
-
         expanduser_mock = mocker.patch("guardrails.classes.rc.expanduser")
         expanduser_mock.return_value = "/Home"
 
diff --git a/tests/unit_tests/cli/test_configure.py b/tests/unit_tests/cli/test_configure.py
index e241fa0f7..037efbbcb 100644
--- a/tests/unit_tests/cli/test_configure.py
+++ b/tests/unit_tests/cli/test_configure.py
@@ -61,11 +61,6 @@ def test_configure(mocker, runner, expected_token, enable_metrics, clear_token):
 
 
 def test_save_configuration_file(mocker):
-    # TODO: Re-enable this once we move nltk.download calls to individual validator repos.  # noqa
-    # Right now, it fires during our import chain, causing this to blow up
-    mocker.patch("nltk.data.find")
-    mocker.patch("nltk.download")
-
     expanduser_mock = mocker.patch("guardrails.cli.configure.expanduser")
     expanduser_mock.return_value = "/Home"
 
diff --git a/tests/unit_tests/cli/test_validate.py b/tests/unit_tests/cli/test_validate.py
index 1368c2ef7..2235009e7 100644
--- a/tests/unit_tests/cli/test_validate.py
+++ b/tests/unit_tests/cli/test_validate.py
@@ -2,9 +2,6 @@
 
 
 def test_validate(mocker):
-    mocker.patch("nltk.data.find")
-    mocker.patch("nltk.download")
-
     mock_validate_llm_output = mocker.patch(
         "guardrails.cli.validate.validate_llm_output"
     )
diff --git a/tests/unit_tests/utils/test_docs_utils.py b/tests/unit_tests/utils/test_docs_utils.py
index 6f03857be..9cbf10e00 100644
--- a/tests/unit_tests/utils/test_docs_utils.py
+++ b/tests/unit_tests/utils/test_docs_utils.py
@@ -48,19 +48,6 @@ def test_text_splitter_split(mock_tokenizer):
     assert chunks[3] == "10 11 12 13 14 15 16 17"
 
 
-# @patch('nltk.data.find', side_effect=LookupError)
-# def test_sentence_split_nltk_download_error(mock_nltk_find):
-#     with pytest.raises(ImportError):
-#         sentence_split("This is a test sentence.")
-
-# @patch('nltk.data.find')
-# def test_sentence_split(mock_nltk_find):
-#     mock_nltk_find.return_value = True
-#     result = sentence_split("This is a test sentence.")
-#     assert len(result) == 1
-#     assert result[0] == "This is a test sentence."
-
-
 def test_prompt_template_token_length(mock_tokenizer, mock_prompt_template):
     text_splitter = TextSplitter()
     length = text_splitter.prompt_template_token_length(mock_prompt_template)