Merge branch 'main' of github.com:jina-ai/microchain into feat_going_meta

.gitignore

@@ -1,4 +1,4 @@
-/executor_level2/
+/microservice/
 
 .env
 config.yml

setup.py

@@ -7,7 +7,7 @@ def read_requirements():
 
 setup(
     name='gptdeploy',
-    version='0.18.18',
+    version='0.18.19',
     description='Use natural language interface to generate, deploy and update your microservice infrastructure.',
     long_description=open('README.md', 'r', encoding='utf-8').read(),
     long_description_content_type='text/markdown',

src/__init__.py

@@ -1,2 +1,2 @@
-__version__ = '0.18.18'
+__version__ = '0.18.19'
 from src.cli import main

src/apis/gpt.py

@@ -12,6 +12,8 @@ from langchain.schema import HumanMessage, SystemMessage, BaseMessage
 from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
 from requests.exceptions import ConnectionError
 
+from src.constants import PRICING_GPT4_PROMPT, PRICING_GPT4_GENERATION, PRICING_GPT3_5_TURBO_PROMPT, \
+    PRICING_GPT3_5_TURBO_GENERATION, CHARS_PER_TOKEN
 from src.options.generate.templates_system import template_system_message_base, executor_example, docarray_example, client_example
 from src.utils.string_tools import print_colored
 
@@ -21,7 +23,23 @@ class GPTSession:
         self.task_description = task_description
         self.test_description = test_description
         self.configure_openai_api_key()
-        self.model_name = 'gpt-4' if model == 'gpt-4' and self.is_gpt4_available() else 'gpt-3.5-turbo'
+        if model == 'gpt-4' and self.is_gpt4_available():
+            self.pricing_prompt = PRICING_GPT4_PROMPT
+            self.pricing_generation = PRICING_GPT4_GENERATION
+        else:
+            if model == 'gpt-4':
+                print_colored('GPT version info', 'GPT-4 is not available. Using GPT-3.5-turbo instead.', 'yellow')
+            model = 'gpt-3.5-turbo'
+            self.pricing_prompt = PRICING_GPT3_5_TURBO_PROMPT
+            self.pricing_generation = PRICING_GPT3_5_TURBO_GENERATION
+        self.model_name = model
+        self.chars_prompt_so_far = 0
+        self.chars_generation_so_far = 0
+
+    def get_conversation(self, system_definition_examples: List[str] = ['executor', 'docarray', 'client']):
+        return _GPTConversation(
+            self.model_name, self.cost_callback, self.task_description, self.test_description, system_definition_examples
+        )
 
     @staticmethod
     def configure_openai_api_key():
@@ -51,11 +69,20 @@ If you have updated it already, please restart your terminal.
                     continue
             return True
         except openai.error.InvalidRequestError:
-            print_colored('GPT version info', 'GPT-4 is not available. Using GPT-3.5-turbo instead.', 'yellow')
             return False
 
-    def get_conversation(self, system_definition_examples: List[str] = ['executor', 'docarray', 'client']):
+    def cost_callback(self, chars_prompt, chars_generation):
-        return _GPTConversation(self.model_name, self.task_description, self.test_description, system_definition_examples)
+        self.chars_prompt_so_far += chars_prompt
+        self.chars_generation_so_far += chars_generation
+        print('\n')
+        money_prompt = self._calculate_money_spent(self.chars_prompt_so_far, self.pricing_prompt)
+        money_generation = self._calculate_money_spent(self.chars_generation_so_far, self.pricing_generation)
+        print('Total money spent so far on openai.com:', f'${money_prompt + money_generation}')
+        print('\n')
+
+    @staticmethod
+    def _calculate_money_spent(num_chars, price):
+        return round(num_chars / CHARS_PER_TOKEN * price / 1000, 3)
 
 
 class AssistantStreamingStdOutCallbackHandler(StreamingStdOutCallbackHandler):
@@ -65,7 +92,7 @@ class AssistantStreamingStdOutCallbackHandler(StreamingStdOutCallbackHandler):
 
 
 class _GPTConversation:
-    def __init__(self, model: str, task_description, test_description, system_definition_examples: List[str] = ['executor', 'docarray', 'client']):
+    def __init__(self, model: str, cost_callback, task_description, test_description, system_definition_examples: List[str] = ['executor', 'docarray', 'client']):
         self._chat = ChatOpenAI(
             model_name=model,
             streaming=True,
@@ -73,6 +100,7 @@ class _GPTConversation:
             verbose=True,
             temperature=0,
         )
+        self.cost_callback = cost_callback
         self.messages: List[BaseMessage] = []
         self.system_message = self._create_system_message(task_description, test_description, system_definition_examples)
         if os.environ['VERBOSE'].lower() == 'true':
@@ -96,6 +124,7 @@ class _GPTConversation:
 
         if os.environ['VERBOSE'].lower() == 'true':
             print()
+        self.cost_callback(sum([len(m.content) for m in self.messages]), len(response.content))
         self.messages.append(response)
         return response.content
 

src/apis/jina_cloud.py

@@ -156,7 +156,7 @@ streamlit run {os.path.join(microservice_path, "app.py")} --server.port 8081 --s
 
 def run_streamlit_app(app_path):
     subprocess.run(['streamlit', 'run', app_path, 'server.address', '0.0.0.0', '--server.port', '8081', '--', '--host',
-                    'grpc://localhost:8080'])
+                    'http://localhost:8080'])
 
 
 def run_locally(executor_name, microservice_version_path):

src/constants.py

@@ -24,6 +24,12 @@ FILE_AND_TAG_PAIRS = [
 
 FLOW_URL_PLACEHOLDER = 'jcloud.jina.ai'
 
+PRICING_GPT4_PROMPT = 0.03
+PRICING_GPT4_GENERATION = 0.06
+PRICING_GPT3_5_TURBO_PROMPT = 0.002
+PRICING_GPT3_5_TURBO_GENERATION = 0.002
+
+CHARS_PER_TOKEN = 3.4
 
 NUM_IMPLEMENTATION_STRATEGIES = 5
 MAX_DEBUGGING_ITERATIONS = 10

src/options/generate/generator.py

@@ -54,7 +54,7 @@ metas:
 
 
     def generate_and_persist_file(self, section_title, template, destination_folder, file_name, **template_kwargs):
-        print_colored('', f'\n############# {section_title} #############', 'blue')
+        print_colored('', f'\n\n############# {section_title} #############', 'blue')
         conversation = self.gpt_session.get_conversation()
         template_kwargs = {k: v for k, v in template_kwargs.items() if k in template.input_variables}
         content_raw = conversation.chat(
@@ -140,7 +140,7 @@ metas:
         print('\nFirst version of the microservice generated. Start iterating on it to make the tests pass...')
 
     def generate_playground(self, microservice_name, microservice_path):
-        print_colored('', '\n############# Playground #############', 'blue')
+        print_colored('', '\n\n############# Playground #############', 'blue')
 
         file_name_to_content = get_all_microservice_files_with_content(microservice_path)
         conversation = self.gpt_session.get_conversation([])
@@ -223,13 +223,14 @@ metas:
         return 'yes' in answer.lower()
 
     def generate_microservice_name(self, description):
+        print_colored('', '\n\n############# What should be the name of the Microservice? #############', 'blue')
         conversation = self.gpt_session.get_conversation()
         name_raw = conversation.chat(template_generate_microservice_name.format(description=description))
         name = self.extract_content_from_result(name_raw, 'name.txt')
         return name
 
     def get_possible_packages(self):
-        print_colored('', '############# What packages to use? #############', 'blue')
+        print_colored('', '\n\n############# What packages to use? #############', 'blue')
         conversation = self.gpt_session.get_conversation()
         packages_raw = conversation.chat(
             template_generate_possible_packages.format(description=self.task_description)

src/options/generate/templates_system.py

@@ -84,7 +84,7 @@ Here is an example of a client file:
 **client.py**
 ```python
 from jina import Client, Document, DocumentArray
-client = Client(host='{FLOW_URL_PLACEHOLDER}')
+client = Client(host='{FLOW_URL_PLACEHOLDER}', protocol='http')
 d = Document(uri='...')
 d.load_uri_to_blob()
 response = client.post('/', inputs=DocumentArray([d])) # the client must be called on '/'

src/options/generate/templates_user.py

@@ -305,7 +305,7 @@ print(response[0].text) # can also be blob in case of image/audio..., this shoul
 ```
 Note that the response will always be in response[0].text
 You must provide the complete app.py file with the exact same syntax to wrap the code.
-The playground (app.py) must read the host from sys.argv because it will be started with a custom host: streamlit run app.py -- --host grpc://...
+The playground (app.py) must read the host from sys.argv because it will be started with a custom host: streamlit run app.py -- --host http(s)://...
 The playground (app.py) must not let the user configure the host on the ui.
 '''
 )