* upgrade version of llm-guard

output.py

@@ -5,25 +5,8 @@ from typing import Dict, List
 
 import streamlit as st
 from llm_guard.input_scanners.anonymize import default_entity_types
-from llm_guard.output_scanners import (
-    JSON,
-    BanSubstrings,
-    BanTopics,
-    Bias,
-    Code,
-    Deanonymize,
-    FactualConsistency,
-    Language,
-    LanguageSame,
-    MaliciousURLs,
-    NoRefusal,
-    Regex,
-    Relevance,
-    Sensitive,
-)
+from llm_guard.output_scanners import get_scanner_by_name
 from llm_guard.output_scanners.relevance import all_models as relevance_models
-from llm_guard.output_scanners.sentiment import Sentiment
-from llm_guard.output_scanners.toxicity import Toxicity
 from llm_guard.vault import Vault
 from streamlit_tags import st_tags
 
@@ -145,7 +128,14 @@ def init_settings() -> (List, Dict):
 
             st_cd_mode = st.selectbox("Mode", ["allowed", "denied"], index=0)
 
-        settings["Code"] = {"languages": st_cd_languages, "mode": st_cd_mode}
+        allowed_languages = None
+        denied_languages = None
+        if st_cd_mode == "allowed":
+            allowed_languages = st_cd_languages
+        elif st_cd_mode == "denied":
+            denied_languages = st_cd_languages
+
+        settings["Code"] = {"allowed": allowed_languages, "denied": denied_languages}
 
     if "JSON" in st_enabled_scanners:
         st_json_expander = st.sidebar.expander(
@@ -181,61 +171,26 @@ def init_settings() -> (List, Dict):
             st_lan_valid_language = st.multiselect(
                 "Languages",
                 [
-                    "af",
                     "ar",
                     "bg",
-                    "bn",
-                    "ca",
-                    "cs",
-                    "cy",
-                    "da",
                     "de",
                     "el",
                     "en",
                     "es",
-                    "et",
-                    "fa",
-                    "fi",
                     "fr",
-                    "gu",
-                    "he",
                     "hi",
-                    "hr",
-                    "hu",
-                    "id",
                     "it",
                     "ja",
-                    "kn",
-                    "ko",
-                    "lt",
-                    "lv",
-                    "mk",
-                    "ml",
-                    "mr",
-                    "ne",
                     "nl",
-                    "no",
-                    "pa",
                     "pl",
                     "pt",
-                    "ro",
                     "ru",
-                    "sk",
-                    "sl",
-                    "so",
-                    "sq",
-                    "sv",
                     "sw",
-                    "ta",
-                    "te",
                     "th",
-                    "tl",
                     "tr",
-                    "uk",
                     "ur",
                     "vi",
-                    "zh-cn",
-                    "zh-tw",
+                    "zh",
                 ],
                 default=["en"],
             )
@@ -322,9 +277,16 @@ def init_settings() -> (List, Dict):
                 "Redact", value=False, help="Replace the matched bad patterns with [REDACTED]"
             )
 
+        good_patterns = None
+        bad_patterns = None
+        if st_regex_type == "good":
+            good_patterns = st_regex_patterns
+        elif st_regex_type == "bad":
+            bad_patterns = st_regex_patterns
+
         settings["Regex"] = {
-            "patterns": st_regex_patterns,
-            "type": st_regex_type,
+            "good_patterns": good_patterns,
+            "bad_patterns": bad_patterns,
             "redact": st_redact,
         }
 
@@ -427,86 +389,25 @@ def init_settings() -> (List, Dict):
 def get_scanner(scanner_name: str, vault: Vault, settings: Dict):
     logger.debug(f"Initializing {scanner_name} scanner")
 
-    if scanner_name == "BanSubstrings":
-        return BanSubstrings(
-            substrings=settings["substrings"],
-            match_type=settings["match_type"],
-            case_sensitive=settings["case_sensitive"],
-            redact=settings["redact"],
-            contains_all=settings["contains_all"],
-        )
-
-    if scanner_name == "BanTopics":
-        return BanTopics(topics=settings["topics"], threshold=settings["threshold"])
-
-    if scanner_name == "Bias":
-        return Bias(threshold=settings["threshold"], use_onnx=True)
-
     if scanner_name == "Deanonymize":
-        return Deanonymize(vault=vault)
-
-    if scanner_name == "JSON":
-        return JSON(required_elements=settings["required_elements"], repair=settings["repair"])
-
-    if scanner_name == "Language":
-        return Language(valid_languages=settings["valid_languages"])
-
-    if scanner_name == "LanguageSame":
-        return LanguageSame()
-
-    if scanner_name == "Code":
-        mode = settings["mode"]
-
-        allowed_languages = None
-        denied_languages = None
-        if mode == "allowed":
-            allowed_languages = settings["languages"]
-        elif mode == "denied":
-            denied_languages = settings["languages"]
-
-        return Code(allowed=allowed_languages, denied=denied_languages, use_onnx=True)
-
-    if scanner_name == "MaliciousURLs":
-        return MaliciousURLs(threshold=settings["threshold"], use_onnx=True)
-
-    if scanner_name == "NoRefusal":
-        return NoRefusal(threshold=settings["threshold"])
-
-    if scanner_name == "FactualConsistency":
-        return FactualConsistency(minimum_score=settings["minimum_score"])
+        settings["vault"] = vault
 
-    if scanner_name == "Regex":
-        match_type = settings["type"]
-
-        good_patterns = None
-        bad_patterns = None
-        if match_type == "good":
-            good_patterns = settings["patterns"]
-        elif match_type == "bad":
-            bad_patterns = settings["patterns"]
-
-        return Regex(
-            good_patterns=good_patterns, bad_patterns=bad_patterns, redact=settings["redact"]
-        )
-
-    if scanner_name == "Relevance":
-        return Relevance(threshold=settings["threshold"], model=settings["model"])
-
-    if scanner_name == "Sensitive":
-        return Sensitive(
-            entity_types=settings["entity_types"],
-            redact=settings["redact"],
-            threshold=settings["threshold"],
-            use_onnx=True,
-        )
-
-    if scanner_name == "Sentiment":
-        return Sentiment(threshold=settings["threshold"])
-
-    if scanner_name == "Toxicity":
-        return Toxicity(threshold=settings["threshold"], use_onnx=True)
+    if scanner_name in [
+        "BanTopics",
+        "Bias",
+        "Code",
+        "Language",
+        "LanguageSame",
+        "MaliciousURLs",
+        "NoRefusal",
+        "FactualConsistency",
+        "Relevance",
+        "Sensitive",
+        "Toxicity",
+    ]:
+        settings["use_onnx"] = True
 
-    raise ValueError("Unknown scanner name")
+    return get_scanner_by_name(scanner_name, settings)
 
 
 def scan(

prompt.py

@@ -4,19 +4,7 @@ from datetime import timedelta
 from typing import Dict, List
 
 import streamlit as st
-from llm_guard.input_scanners import (
-    Anonymize,
-    BanSubstrings,
-    BanTopics,
-    Code,
-    Language,
-    PromptInjection,
-    Regex,
-    Secrets,
-    Sentiment,
-    TokenLimit,
-    Toxicity,
-)
+from llm_guard.input_scanners import get_scanner_by_name
 from llm_guard.input_scanners.anonymize import default_entity_types
 from llm_guard.input_scanners.prompt_injection import ALL_MODELS as PI_ALL_MODELS
 from llm_guard.vault import Vault
@@ -181,9 +169,16 @@ def init_settings() -> (List, Dict):
 
             st_cd_mode = st.selectbox("Mode", ["allowed", "denied"], index=0)
 
+        allowed_languages = None
+        denied_languages = None
+        if st_cd_mode == "allowed":
+            allowed_languages = st_cd_languages
+        elif st_cd_mode == "denied":
+            denied_languages = st_cd_languages
+
         settings["Code"] = {
-            "languages": st_cd_languages,
-            "mode": st_cd_mode,
+            "allowed": allowed_languages,
+            "denied": denied_languages,
         }
 
     if "Language" in st_enabled_scanners:
@@ -196,61 +191,26 @@ def init_settings() -> (List, Dict):
             st_lan_valid_language = st.multiselect(
                 "Languages",
                 [
-                    "af",
                     "ar",
                     "bg",
-                    "bn",
-                    "ca",
-                    "cs",
-                    "cy",
-                    "da",
                     "de",
                     "el",
                     "en",
                     "es",
-                    "et",
-                    "fa",
-                    "fi",
                     "fr",
-                    "gu",
-                    "he",
                     "hi",
-                    "hr",
-                    "hu",
-                    "id",
                     "it",
                     "ja",
-                    "kn",
-                    "ko",
-                    "lt",
-                    "lv",
-                    "mk",
-                    "ml",
-                    "mr",
-                    "ne",
                     "nl",
-                    "no",
-                    "pa",
                     "pl",
                     "pt",
-                    "ro",
                     "ru",
-                    "sk",
-                    "sl",
-                    "so",
-                    "sq",
-                    "sv",
                     "sw",
-                    "ta",
-                    "te",
                     "th",
-                    "tl",
                     "tr",
-                    "uk",
                     "ur",
                     "vi",
-                    "zh-cn",
-                    "zh-tw",
+                    "zh",
                 ],
                 default=["en"],
             )
@@ -303,9 +263,16 @@ def init_settings() -> (List, Dict):
                 "Redact", value=False, help="Replace the matched bad patterns with [REDACTED]"
             )
 
+        good_patterns = None
+        bad_patterns = None
+        if st_regex_type == "good":
+            good_patterns = st_regex_patterns
+        elif st_regex_type == "bad":
+            bad_patterns = st_regex_patterns
+
         settings["Regex"] = {
-            "patterns": st_regex_patterns,
-            "type": st_regex_type,
+            "good_patterns": good_patterns,
+            "bad_patterns": bad_patterns,
             "redact": st_redact,
         }
 
@@ -392,74 +359,15 @@ def get_scanner(scanner_name: str, vault: Vault, settings: Dict):
     logger.debug(f"Initializing {scanner_name} scanner")
 
     if scanner_name == "Anonymize":
-        return Anonymize(
-            vault=vault,
-            allowed_names=settings["allowed_names"],
-            hidden_names=settings["hidden_names"],
-            entity_types=settings["entity_types"],
-            preamble=settings["preamble"],
-            use_faker=settings["use_faker"],
-            threshold=settings["threshold"],
-            use_onnx=True,
-        )
-
-    if scanner_name == "BanSubstrings":
-        return BanSubstrings(
-            substrings=settings["substrings"],
-            match_type=settings["match_type"],
-            case_sensitive=settings["case_sensitive"],
-            redact=settings["redact"],
-            contains_all=settings["contains_all"],
-        )
-
-    if scanner_name == "BanTopics":
-        return BanTopics(topics=settings["topics"], threshold=settings["threshold"])
-
-    if scanner_name == "Code":
-        mode = settings["mode"]
-
-        allowed_languages = None
-        denied_languages = None
-        if mode == "allowed":
-            allowed_languages = settings["languages"]
-        elif mode == "denied":
-            denied_languages = settings["languages"]
-
-        return Code(allowed=allowed_languages, denied=denied_languages, use_onnx=True)
-
-    if scanner_name == "Language":
-        return Language(valid_languages=settings["valid_languages"])
+        settings["vault"] = vault
 
     if scanner_name == "PromptInjection":
-        return PromptInjection(threshold=settings["threshold"], models=PI_ALL_MODELS, use_onnx=True)
-
-    if scanner_name == "Regex":
-        match_type = settings["type"]
-
-        good_patterns = None
-        bad_patterns = None
-        if match_type == "good":
-            good_patterns = settings["patterns"]
-        elif match_type == "bad":
-            bad_patterns = settings["patterns"]
-
-        return Regex(
-            good_patterns=good_patterns, bad_patterns=bad_patterns, redact=settings["redact"]
-        )
-
-    if scanner_name == "Secrets":
-        return Secrets(redact_mode=settings["redact_mode"])
-
-    if scanner_name == "Sentiment":
-        return Sentiment(threshold=settings["threshold"])
-
-    if scanner_name == "TokenLimit":
-        return TokenLimit(limit=settings["limit"], encoding_name=settings["encoding_name"])
+        settings["models"] = PI_ALL_MODELS
 
-    if scanner_name == "Toxicity":
-        return Toxicity(threshold=settings["threshold"], use_onnx=True)
+    if scanner_name in ["Anonymize", "BanTopics", "Code", "PromptInjection", "Toxicity"]:
+        settings["use_onnx"] = True
 
-    raise ValueError("Unknown scanner name")
+    return get_scanner_by_name(scanner_name, settings)
 
 
 def scan(

requirements.txt

@@ -1,5 +1,5 @@
-llm-guard==0.3.1
-llm-guard[onnxruntime]==0.3.1
-pandas==2.1.2
-streamlit==1.28.1
+llm-guard==0.3.2
+llm-guard[onnxruntime]==0.3.2
+pandas==2.1.3
+streamlit==1.28.2
 streamlit-tags==1.2.8