Merge pull request #49 from SAP/develop

Release v0.2.1

Author: marcorosa

.github/workflows/lint-frontend.yml

@@ -35,8 +35,10 @@ jobs:
           npm ci
 
       - name: Run linters
-        uses: wearerequired/lint-action@v2
+        uses: reviewdog/action-eslint@v1
         with:
-          eslint: true
-          eslint_dir: frontend
-          eslint_extensions: js,ts
+          github_token: ${{ secrets.GITHUB_TOKEN }}
+          reporter: github-pr-review
+          eslint_flags: 'frontend/'
+          fail_on_error: true
+          level: warning

.reuse/dep5

@@ -1,3 +1,14 @@
+# Version: v0.2.1
+
+* [#34](https://github.com/SAP/STARS/pull/34): Support aicore-mistralai models
+* [#36](https://github.com/SAP/STARS/pull/36): Fix/frontend
+* [#37](https://github.com/SAP/STARS/pull/37): Bump prismjs from 1.29.0 to 1.30.0 in /frontend
+* [#39](https://github.com/SAP/STARS/pull/39): Support remote ollama models
+* [#40](https://github.com/SAP/STARS/pull/40): [chore]:Reuse API update - Migration from dep5 file to TOML file.
+* [#47](https://github.com/SAP/STARS/pull/47): Bump flask-cors from 5.0.0 to 6.0.0 in /backend-agent
+* [#48](https://github.com/SAP/STARS/pull/48): Fix js linter action
+
+
 # Version: v0.2.0
 
 * [#25](https://github.com/SAP/STARS/pull/25): Bump katex from 0.16.10 to 0.16.21 in /frontend

CHANGELOG.md

@@ -0,0 +1,32 @@
+version = 1
+SPDX-PackageName = "Smart Threat AI Reporting Scanner (STARS)"
+SPDX-PackageSupplier = "ospo@sap.com"
+SPDX-PackageDownloadLocation = "https://github.com/SAP/stars"
+SPDX-PackageComment = "The code in this project may include calls to APIs (\"API Calls\") of\n SAP or third-party products or services developed outside of this project\n (\"External Products\").\n \"APIs\" means application programming interfaces, as well as their respective\n specifications and implementing code that allows software to communicate with\n other software.\n API Calls to External Products are not licensed under the open source license\n that governs this project. The use of such API Calls and related External\n Products are subject to applicable additional agreements with the relevant\n provider of the External Products. In no event shall the open source license\n that governs this project grant any rights in or to any External Products,or\n alter, expand or supersede any terms of the applicable additional agreements.\n If you have a valid license agreement with SAP for the use of a particular SAP\n External Product, then you may make use of any API Calls included in this\n project's code for that SAP External Product, subject to the terms of such\n license agreement. If you do not have a valid license agreement for the use of\n a particular SAP External Product, then you may only make use of any API Calls\n in this project for that SAP External Product for your internal, non-productive\n and non-commercial test and evaluation of such API Calls. Nothing herein grants\n you any rights to use or access any SAP External Product, or provide any third\n parties the right to use of access any SAP External Product, through API Calls."
+
+[[annotations]]
+path = "**"
+precedence = "aggregate"
+SPDX-FileCopyrightText = "2024 SAP SE or an SAP affiliate company and stars contributors"
+SPDX-License-Identifier = "Apache-2.0"
+
+[[annotations]]
+path = "backend-agent/cli.py"
+precedence = "aggregate"
+SPDX-FileCopyrightText = ["", "2024 SAP SE or an SAP affiliate company and STARS contributors", "2021 Sandflow Consulting LLC"]
+SPDX-License-Identifier = "Apache-2.0 and BSD-3-Clause"
+SPDX-FileComment = "these files contain content from SAP and ttconv: cli.py is overall written by SAP, but contains a code snippet from src/main/python/ttconv/tt.py file taken from ttconv"
+
+[[annotations]]
+path = "backend-agent/libs/artprompt.py"
+precedence = "aggregate"
+SPDX-FileCopyrightText = ["", "2024 SAP SE or an SAP affiliate company and STARS contributors", "2024 UW-NSL"]
+SPDX-License-Identifier = "Apache-2.0 and MIT"
+SPDX-FileComment = "these files contain content from SAP and UW-NSL: the original content was written by UW-NSL, but it was refactored, simplified, and modified by SAP to be more suitable to this project"
+
+[[annotations]]
+path = "frontend/src/app/app.component.spec.ts"
+precedence = "aggregate"
+SPDX-FileCopyrightText = ["", "2010-2020 Google LLC. https://angular.io/license"]
+SPDX-License-Identifier = "MIT"
+SPDX-FileComment = "these files contain content from Angular 11.1.0-next.1 (it contains a code snippet from integration/trusted-types/src/app/app.component.spec.ts file taken from angular)"

REUSE.toml

@@ -23,35 +23,21 @@
 logger.addHandler(status.trace_logging)
 
 AICORE_MODELS = {
-    'openai':
+    'aicore-ibm':
     [
-        'gpt-35-turbo',
-        'gpt-35-turbo-0125',
-        'gpt-35-turbo-16k',
-        'gpt-4',
-        'gpt-4-32k',
-        'gpt-4o',
-        'gpt-4o-mini'
+        'ibm--granite-13b-chat'
+    ],
+    'aicore-mistralai':
+    [
+        'mistralai--mistral-large-instruct',
     ],
-    'opensource':
+    'aicore-opensource':
     [
         'mistralai--mixtral-8x7b-instruct-v01',
         'meta--llama3.1-70b-instruct',
         'meta--llama3-70b-instruct'
     ],
-    'vertexai':
-    [
-        'text-bison',
-        'chat-bison',
-        'gemini-1.0-pro',
-        'gemini-1.5-pro',
-        'gemini-1.5-flash'
-    ],
-    'ibm':
-    [
-        'ibm--granite-13b-chat'
-    ],
-    'bedrock':
+    'aws-bedrock':
     [
         'amazon--titan-text-lite',
         'amazon--titan-text-express',
@@ -62,7 +48,25 @@
         'amazon--nova-pro',
         'amazon--nova-lite',
         'amazon--nova-micro'
-    ]
+    ],
+    'azure-openai':
+    [
+        'gpt-35-turbo',
+        'gpt-35-turbo-0125',
+        'gpt-35-turbo-16k',
+        'gpt-4',
+        'gpt-4-32k',
+        'gpt-4o',
+        'gpt-4o-mini'
+    ],
+    'gcp-vertexai':
+    [
+        'text-bison',
+        'chat-bison',
+        'gemini-1.0-pro',
+        'gemini-1.5-pro',
+        'gemini-1.5-flash'
+    ],
 }
 
 
@@ -79,25 +83,30 @@ def from_model_name(cls, model_name: str) -> 'LLM':
         Create a specific LLM object from the name of the model.
         Useful because the user can specify only the name in the agent.
         """
-        if 'gpt' in model_name:
+        # Foundation-models scenarios in AI Core
+        if model_name in AICORE_MODELS['azure-openai']:
             # The agent sometimes autocorrects gpt-35-turbo to gpt-3.5-turbo,
             # so we handle this behavior here.
             if model_name == 'gpt-3.5-turbo':
                 model_name = 'gpt-35-turbo'
             return AICoreOpenAILLM(model_name)
-        if model_name in AICORE_MODELS['opensource']:
-            return AICoreOpenAILLM(model_name, False)
-        if model_name in AICORE_MODELS['ibm']:
+        if model_name in AICORE_MODELS['aicore-ibm']:
             # IBM models are compatible with OpenAI completion API
             return AICoreOpenAILLM(model_name)
-        if model_name in AICORE_MODELS['vertexai']:
-            return AICoreGoogleVertexLLM(model_name)
-        if model_name in AICORE_MODELS['bedrock']:
+        if model_name in AICORE_MODELS['aicore-opensource']:
+            return AICoreOpenAILLM(model_name, False)
+        if model_name in AICORE_MODELS['aicore-mistralai']:
+            return AICoreOpenAILLM(model_name, False)
+        if model_name in AICORE_MODELS['aws-bedrock']:
             if 'titan' in model_name:
                 # Titan models don't support system prompts
                 return AICoreAmazonBedrockLLM(model_name, False)
             else:
                 return AICoreAmazonBedrockLLM(model_name)
+        if model_name in AICORE_MODELS['gcp-vertexai']:
+            return AICoreGoogleVertexLLM(model_name)
+
+        # Custom models
         if model_name == 'mistral':
             return LocalOpenAILLM(
                 os.getenv('MISTRAL_MODEL_NAME', ''),
@@ -109,9 +118,17 @@ def from_model_name(cls, model_name: str) -> 'LLM':
         # possible local ollama instance. If it not even served there, then an
         # exception is raised because such model has either an incorrect name
         # or it has not been deployed.
+        ollama_host = os.getenv('OLLAMA_HOST')
+        ollama_port = os.getenv('OLLAMA_PORT', 11434)
         try:
-            ollama.show(model_name)
-            return OllamaLLM(model_name)
+            if ollama_host:
+                # The model is served in a remote ollama instance
+                remote_ollama_host = f'{ollama_host}:{ollama_port}'
+                return OllamaLLM(model_name, remote_ollama_host)
+            else:
+                # The model is served in a local ollama instance
+                ollama.show(model_name)
+                return OllamaLLM(model_name)
         except (ollama.ResponseError, httpx.ConnectError):
             raise ValueError(f'Model {model_name} not found')
 
@@ -127,7 +144,17 @@ def _calculate_list_of_supported_models(self) -> list[str]:
         if os.getenv('MISTRAL_URL'):
             models.append('mistral')
         try:
-            ollama_models = [m['name'] for m in ollama.list()['models']]
+            ollama_models = []
+            ollama_host = os.getenv('OLLAMA_HOST')
+            ollama_port = os.getenv('OLLAMA_PORT', 11434)
+            if ollama_host:
+                # The model is served in a remote ollama instance
+                remote_ollama_host = f'{ollama_host}:{ollama_port}'
+                ollama_models = [m['model'] for m in
+                                 ollama.Client(remote_ollama_host).list()
+                                 ['models']]
+            else:
+                ollama_models = [m['name'] for m in ollama.list()['models']]
             return models + ollama_models
         except httpx.ConnectError:
             return models
@@ -325,28 +352,30 @@ def generate_completions_for_messages(self,
 
 
 class OllamaLLM(LLM):
-    def __init__(self, model_name: str):
+    def __init__(self, model_name: str, host=None):
         self.model_name = model_name
+        self.client = ollama.Client(host=host)
 
     def __str__(self) -> str:
         return f'{self.model_name}/Ollama LLM'
 
     def generate(self,
                  system_prompt: str,
                  prompt: str,
-                 temperature: float,
-                 max_tokens: int,
-                 n: int) -> list[str]:
+                 max_tokens: int = 4096,
+                 temperature: float = 0.3,
+                 n: int = 1,) -> list[str]:
         try:
             messages = [
                 {'role': 'system', 'content': system_prompt},
                 {'role': 'user', 'content': prompt},
             ]
             generations = [
-                ollama.generate(self.model_name,
-                                prompt, system_prompt,
-                                options={'temperature': temperature})
-                ['response']
+                self.client.generate(model=self.model_name,
+                                     prompt=prompt,
+                                     system=system_prompt,
+                                     options={'temperature': temperature}
+                                     )['response']
                 for _ in range(n)]
             return self._trace_llm_call(messages, Success(generations))
         except Exception as e:
@@ -363,13 +392,14 @@ def generate_completions_for_messages(
             n: int = 1) -> list[str]:
         try:
             generations = [
-                ollama.chat(self.model_name,
-                            messages,
-                            options={'temperature': temperature,
-                                     'top_p': top_p,
-                                     'frequency_penalty': frequency_penalty,
-                                     'presence_penalty': presence_penalty})
-                ['message']['content']
+                self.client.chat(
+                    self.model_name,
+                    messages,
+                    options={'temperature': temperature,
+                             'top_p': top_p,
+                             'frequency_penalty': frequency_penalty,
+                             'presence_penalty': presence_penalty}
+                )['message']['content']
                 for _ in range(n)
             ]
             return self._trace_llm_call(messages, Success(generations))

backend-agent/llm.py

@@ -3,7 +3,7 @@ Deprecated==1.2.15
 python-dotenv==1.0.1
 faiss-cpu~=1.9.0
 Flask==2.3.3
-Flask-Cors==5.0.0
+Flask-Cors==6.0.0
 flask_sock==0.7.0
 langchain~=0.2.16
 PyYAML==6.0.2
@@ -17,7 +17,7 @@ tensorflow-hub
 sentence-transformers
 torchfile
 tensorflow-text
-ollama==0.4.2
+ollama==0.4.7
 weasyprint
 pyrit==0.2.1
 textattack>=0.3.10