Merge pull request #4 from tomkat-cr/GS-66_fix_lcel_issues

GS-66 - Fix LCEL issues

.env.example

@@ -21,7 +21,7 @@ DEFAULT_LANG=en
 # Application debug (0,1)
 APP_DEBUG=1
 # Application environment: dev, qa, staging, prod
-APP_STAGE=dev
+# APP_STAGE=dev
 # Application super admin email
 APP_SUPERADMIN_EMAIL=xxxx
 # Application secret ket (to set password encryption)
@@ -96,16 +96,55 @@ FRONTEND_PATH=../exampleapp_frontend
 # Local python version
 PYTHON_VERSION=3.11.5
 #
+# IAAS Cloud provider
+# Available options: `aws`, `gcp`, `azure`
+CLOUD_PROVIDER=aws
+#
+# Enable/disable Cloud Provider secrets (instead of environment variables).
+# Available options: `1` to enable, `0` to disable. Defaults to: 1
+# GET_SECRETS_ENABLED=0
+#
+# Fine grained Cloud Provider secrets management:
+#
+# Enable/disable Cloud Provider envvars.
+# Available options: `1` to enable, `0` to disable. Defaults to: 1
+# Set to "0" in local development environment so envvars like APP_CORS_ORIGIN can be
+# set by the scripts and .env file and access QA resources from DEV.
+# GET_SECRETS_ENVVARS=0
+#
+# Enable/disable Cloud Provider critical secrets.
+# Available options: `1` to enable, `0` to disable. Defaults to: 1
+# Set to "0" in local development environment so envvars like APP_DB_URI can be
+# set by the scripts and .env file and access QA resources from DEV.
+# GET_SECRETS_CRITICAL=0
+#
 # AWS Configuration
 # https://console.aws.amazon.com
-AWS_S3_BUCKET_NAME_FE=aws-s3-bucket-name
+#
+# Region for this App all AWS services
 AWS_REGION=aws-region
-AWS_API_GATEWAY_STAGE=aws-api-gateway-stage
+#
+# AWS Deployment type
+# Available options: `lambda`, `fargate`, `ec2`. Defaults to: lambda
+AWS_DEPLOYMENT_TYPE=lambda
+#
+# AWS base name for Lambda Functions, API Gateway, EC2, ELB, etc.
 AWS_LAMBDA_FUNCTION_NAME=aws-lambda-function-name
-AWS_LAMBDA_FUNCTION_ROLE=aws-lambda-function-role
-AWS_SSL_CERTIFICATE_ARN=arn:aws:acm:AWS-REGION:AWS-ACCOUNT:certificate/AWS-CERTIFICATE-UUD
 #
-# Development URL masking external hostname (storage URL encryption)
+# AWS Lambda function role:
+# These variables are used only if deploy without AWS SAM (deploy_without_sam) in big_lambdas_manager.sh. SAM generates this role automatically
+AWS_LAMBDA_FUNCTION_ROLE_QA=exampleapp-api_handler-role-qa
+AWS_LAMBDA_FUNCTION_ROLE_STAGING=exampleapp-api_handler-role-staging
+AWS_LAMBDA_FUNCTION_ROLE_DEMO=exampleapp-api_handler-role-demo
+AWS_LAMBDA_FUNCTION_ROLE_PROD=exampleapp-api_handler-role-prod
+#
+# AWS SSL certificate ARN (used by big_lambdas_manager.sh)
+AWS_SSL_CERTIFICATE_ARN=arn:aws:acm:AWS-REGION:AWS-ACCOUNT:certificate/AWS-CERTIFICATE-UUID
+#
+# AWS S3 bucket name (used by set_fe_cloudfront_domain.sh to set the CloudFront domain name in the frontend for the CORS config)
+AWS_S3_BUCKET_NAME_FE=aws-s3-bucket-name
+#
+# Development URL masking external hostname (for features like AI Vision)
 URL_MASK_EXTERNAL_HOSTNAME=app-dev.exampleapp.com
 #
 # SMTP Mail configuration
@@ -119,28 +158,58 @@ SMTP_DEFAULT_SENDER=sender_email
 DOCKER_ACCOUNT=docker_account_username
 #
 # Local development environment run configuration
+#
 # Options are: uvicorn, gunicorn, chalice, chalice_docker
+#
 # Chalice case: "chalice" to use http (running without docker) or "chalice_docker" to use https (with docker)
 # http:
 # RUN_METHOD="chalice"
 # https:
 RUN_METHOD="chalice_docker"
 #
 # Tests configuration
+#
 # Testing enndpoint
 TEST_APP_URL=http://app.exampleapp.local:5002
 #
+# Default App main code directory
+# for Chalice:
+# https://aws.github.io/chalice/topics/packaging.html
+# APP_DIR='.'
+# for FastAPI:
+# https://fastapi.tiangolo.com/tutorial/bigger-applications/?h=directory+structure#an-example-file-structure
+# APP_DIR='app'
+# for Flask:
+# https://flask.palletsprojects.com/en/2.3.x/tutorial/layout/
+# APP_DIR='flaskr'
+#
+# Default App entry point code file
+# for Chalice:
+# https://aws.github.io/chalice/topics/packaging.html
+# APP_MAIN_FILE='app'
+# for FastAPI:
+# https://fastapi.tiangolo.com/tutorial/bigger-applications/?h=directory+structure#an-example-file-structure
+# APP_MAIN_FILE='main'
+# for Flask:
+# https://flask.palletsprojects.com/en/2.3.x/tutorial/factory/
+# APP_MAIN_FILE='__init__'
+#
 # Flask configuration
 FLASK_APP=index.py
 #
-# AI technology (openai/langchain APIs)
-# AI_TECHNOLOGY="openai"
-AI_TECHNOLOGY="langchain"
+# Local frontend port (defautls to 3000)
+FRONTEND_LOCAL_PORT=3000
+# Local backend API port (defaults to 5001)
+BACKEND_LOCAL_PORT=5001
 #
 ####################
 # AI Configuration #
 ####################
 #
+# AI technology (openai/langchain APIs)
+# AI_TECHNOLOGY="openai"
+AI_TECHNOLOGY="langchain"
+#
 # Google configuration
 # https://console.cloud.google.com/apis/credentials
 GOOGLE_API_KEY=google_console_api_key
@@ -150,6 +219,7 @@ GOOGLE_CSE_ID=google_console_cse_key
 # OpenAI configuration
 # https://platform.openai.com/api-keys
 OPENAI_API_KEY=openai_api_key
+# https://openai.com/api/pricing/
 OPENAI_MODEL=gpt-3.5-turbo
 OPENAI_TEMPERATURE=0.7
 #

CHANGELOG.md

@@ -17,6 +17,23 @@ This project adheres to [Semantic Versioning](http://semver.org/) and [Keep a Ch
 ### Breaks
 
 
+## 0.1.8 (2024-07-18)
+
+### New
+Add "langchain-google-community" due to a deprecation notice about GoogleSearchAPIWrapper [GS-66].
+Add Langchain Tools description length validation, to avoid descriptions > 1024 chars.
+
+### Changes
+Update dependecies "langchain" (from ^0.1.20 to ^0.2.3), "langchain-core", "langchain-openai" (from ^0.1.6 to ^0.1.8), "tiktoken" (from 0.6 to ^0.7.0) to be able to add "langchain-google-community" [GS-66].
+Update env.example. to have the GS BE Core latest updates.
+AWS_API_GATEWAY_STAGE env. var. removed from env.example.
+All DEBUGs turned off to save logs/AWS Cloudwatch space.
+
+### Fixes
+Fix audio processing issues in FastAPI due to AWS API Gateway limitations, sending the base64 encoded files back [GS-95].
+Change: minor linting changes.
+
+
 ## 0.1.7 (2024-06-06)
 
 ### New

genericsuite_ai/fastapilib/endpoints/ai_chatbot_endpoint.py

@@ -14,8 +14,14 @@
 from fastapi.security import HTTPBasic
 
 # from genericsuite.util.framework_abs_layer import Response, BlueprintOne
-from genericsuite.fastapilib.framework_abstraction import Response, BlueprintOne
+from genericsuite.fastapilib.framework_abstraction import (
+    Response,
+    BlueprintOne,
+)
 from genericsuite.util.app_logger import log_debug
+from genericsuite.util.utilities import (
+    send_file_text_text,
+)
 
 from genericsuite.fastapilib.util.parse_multipart import download_file_fa
 from genericsuite.fastapilib.util.dependencies import (
@@ -33,6 +39,7 @@
 #     assign_app_gpt_functions
 # )
 
+
 class VisionImageAnalyzerRequest(BaseModel):
     """
     This class is used to parse the query parameters for the
@@ -53,7 +60,10 @@ class TranscribeAudioRequest(BaseModel):
     other: str
 
 
-DEBUG = True
+DEBUG = False
+
+# send_file_fa() mode
+SEND_FILE_AS_BINARY = False
 
 # Set FastAPI router
 # router = APIRouter()
@@ -75,6 +85,18 @@ def remove_temp_file(file_path: str) -> None:
 def send_file_fa(
     file_to_send: str,
     background_tasks: BackgroundTasks,
+) -> Any:
+    """
+    Send the file back, the FastAPI way
+    """
+    if SEND_FILE_AS_BINARY:
+        return send_binary_file_fa(file_to_send, background_tasks)
+    return send_base64_file_fa(file_to_send, background_tasks)
+
+
+def send_binary_file_fa(
+    file_to_send: str,
+    background_tasks: BackgroundTasks,
 ) -> FileResponse:
     """
     Send the file back, the FastAPI way
@@ -87,6 +109,19 @@ def send_file_fa(
     return FileResponse(file_to_send)
 
 
+def send_base64_file_fa(
+    file_to_send: str,
+    background_tasks: BackgroundTasks,
+) -> Any:
+    """
+    Return the file content as GenericSuite way, Base64 encoded.
+    This approach worked for audio file and the ai_chatbot in Chalice
+    because AWS API Gateway won't allow binary responses.
+    """
+    _ = DEBUG and log_debug("Returning file content the Genericsuite way")
+    return send_file_text_text(file_to_send)
+
+
 @router.post('/chatbot', tags='chatbot')
 async def ai_chatbot_endpoint(
     request: FaRequest,
@@ -129,7 +164,7 @@ async def ai_chatbot_endpoint(
 async def vision_image_analyzer_endpoint(
     request: FaRequest,
     file: UploadFile = File(...),
-        # file: Annotated[UploadFile, File(...)] = None,
+    # file: Annotated[UploadFile, File(...)] = None,
     current_user: str = Depends(get_current_user),
     query_params: VisionImageAnalyzerRequest = Depends(),
 ) -> Response:
@@ -142,15 +177,17 @@ async def vision_image_analyzer_endpoint(
     :return: The text with the image analysis.
     """
     _ = DEBUG and \
-        log_debug('1) vision_image_analyzer_endpoint | query_params:' +
+        log_debug(
+            '1) vision_image_analyzer_endpoint | query_params:' +
             f' {request.query_params}')
     gs_request, other_params = get_default_fa_request(
         current_user=current_user, query_params=query_params.dict(),
         preserve_nones=True)
     router.set_current_request(request, gs_request)
     uploaded_file_path = await download_file_fa(file)
     _ = DEBUG and \
-        log_debug('2) vision_image_analyzer_endpoint | uploaded_file_path:' +
+        log_debug(
+            '2) vision_image_analyzer_endpoint | uploaded_file_path:' +
             f' {uploaded_file_path} | request: {request}')
     return vision_image_analyzer_endpoint_model(
         request=gs_request,
@@ -165,7 +202,7 @@ async def vision_image_analyzer_endpoint(
 async def transcribe_audio_endpoint(
     request: FaRequest,
     file: UploadFile = File(...),
-       # file: Annotated[UploadFile, File()] = None,
+    # file: Annotated[UploadFile, File()] = None,
     current_user: str = Depends(get_current_user),
     query_params: TranscribeAudioRequest = Depends(),
 ) -> Response:
@@ -178,7 +215,8 @@ async def transcribe_audio_endpoint(
     :return: The transcription result.
     """
     _ = DEBUG and \
-        log_debug('transcribe_audio_endpoint | query_params:' +
+        log_debug(
+            'transcribe_audio_endpoint | query_params:' +
             f' {request.query_params}')
     gs_request, other_params = get_default_fa_request(
         current_user=current_user, query_params=query_params.dict(),

genericsuite_ai/lib/ai_chatbot_commons.py

@@ -23,7 +23,7 @@
     get_current_date_time
 )
 
-DEBUG = True
+DEBUG = False
 
 
 def get_role(v):

genericsuite_ai/lib/ai_chatbot_endpoint.py

@@ -42,7 +42,7 @@
 )
 from genericsuite_ai.models.billing.billing_utilities import BillingUtilities
 
-DEBUG = True
+DEBUG = False
 
 
 def ai_chatbot_endpoint(
@@ -92,6 +92,7 @@ def ai_chatbot_endpoint(
     # after the app_context_and_set_env(request)
     settings = Config(app_context)
     billing = BillingUtilities(app_context)
+
     def get_model_engine():
         model_engine = \
             f'\n | AI_TECHNOLOGY: {settings.AI_TECHNOLOGY}'
@@ -137,7 +138,8 @@ def get_sub_model_name():
 
     if '[SEND_FILE_BACK]' in ai_chatbot_response.get('response'):
         file_to_send = ai_chatbot_response['response'].split('=')[1]
-        _ = DEBUG and log_debug('AICBEP-3) AI_CHATBOT_ENDPOINT' +
+        _ = DEBUG and log_debug(
+            'AICBEP-3) AI_CHATBOT_ENDPOINT' +
             f' - Sending file: {file_to_send}')
         if sendfile_callable:
             return sendfile_callable(

genericsuite_ai/lib/ai_chatbot_main_langchain.py

@@ -70,7 +70,7 @@
 from genericsuite_ai.models.billing.billing_utilities import BillingUtilities
 
 
-DEBUG = True
+DEBUG = False
 
 NON_AI_TRANSLATOR = 'google_translate'
 NON_AGENT_PROMPT = 'mediabros/gs_non_agent_lcel'
@@ -532,6 +532,30 @@ def run_lcel_chain(
     return exec_result
 
 
+def verify_tools(tools, conv_response):
+    """
+    Verify if the tools are valid, checking the Tool description lenght.
+    """
+    if conv_response["error"]:
+        return conv_response
+    bad_tools = [
+        (tuple(
+            [index, f"Lenght: {len(tools[index].description)}", tools[index]]
+        ))
+        for index in range(len(tools))
+        if len(tools[index].description) > 1024
+    ]
+    _ = DEBUG and \
+        log_debug('>>> VERIFY_TOOLS | LONG DESC tools[list]:' +
+            f' {bad_tools}')
+    if len(bad_tools) > 0:
+        conv_response["error"] = True
+        conv_response["error_message"] = \
+            "ERROR: Too long Tool description(s): " + \
+            ', '.join([f'{v[2].name} ({v[1]})' for v in bad_tools])
+    return conv_response
+
+
 def run_assistant(
     conv_response: dict,
     llm: Any,
@@ -547,6 +571,7 @@ def run_assistant(
         log_debug('>>> 1) RUN_ASSISTANT | LANGCHAIN_AGENT_TYPE:' +
             f' {settings.LANGCHAIN_AGENT_TYPE}')
 
+    conv_response = verify_tools(tools, conv_response)
     if conv_response["error"]:
         return conv_response
 

genericsuite_ai/lib/ai_vision.py

@@ -45,7 +45,7 @@
 )
 from genericsuite_ai.models.billing.billing_utilities import BillingUtilities
 
-DEBUG = True
+DEBUG = False
 cac = CommonAppContext()
 
 

genericsuite_ai/lib/web_search.py

@@ -28,7 +28,7 @@
 )
 from genericsuite_ai.config.config import Config
 
-DEBUG = True
+DEBUG = False
 
 cac = CommonAppContext()