Merge branch 'main' of https://github.com/jina-ai/gptdeploy into refactor-langchain

# Conflicts: # src/apis/gpt.py # src/cli.py # src/options/generate/generator.py # src/options/generate/prompt_system.py # src/options/generate/prompt_tasks.py
2025-12-20 23:24:20 +01:00 · 2023-04-18 12:17:52 +02:00
parent 86f50e1c86 c11d68fdf7
commit 14b2c1e75a
15 changed files with 112 additions and 85 deletions
--- a/README.md
+++ b/README.md
@@ -100,54 +100,67 @@ jc delete <microservice id>

 <img src="res/teaser.png" alt="QR Code Generator" width="600" />

+### Chemical Formula Visualization
+```bash
+generate --description "Convert a chemical formula into a 2D chemical structure diagram" --test "C=C, CN=C=O, CCC(=O)O" --path microservice
+```
+
+<img src="res/chemical_formula_example.png" alt="Chemical Formula Visualization" width="600" />
+
 ### Animal Detector
 ```bash

-gptdeploy generate --description "Given an image, return the image with bounding boxes of all animals (https://pjreddie.com/media/files/yolov3.weights, https://raw.githubusercontent.com/pjreddie/darknet/master/cfg/yolov3.cfg)" --test "https://images.unsplash.com/photo-1444212477490-ca407925329e contains animals" --model gpt-4
+gptdeploy generate --description "Given an image, return the image with bounding boxes of all animals (https://pjreddie.com/media/files/yolov3.weights, https://raw.githubusercontent.com/pjreddie/darknet/master/cfg/yolov3.cfg)" --test "https://images.unsplash.com/photo-1444212477490-ca407925329e contains animals" --model gpt-4 --path microservice
 ```

 <img src="res/animal_detector_example.png" alt="Animal Detector" width="600" />

 ### Meme Generator
 ```bash
-gptdeploy generate --description "Generate a meme from an image and a caption" --test "Surprised Pikachu: https://media.wired.com/photos/5f87340d114b38fa1f8339f9/master/w_1600%2Cc_limit/Ideas_Surprised_Pikachu_HD.jpg, TOP:When you discovered GPTDeploy" --model gpt-4
+gptdeploy generate --description "Generate a meme from an image and a caption" --test "Surprised Pikachu: https://media.wired.com/photos/5f87340d114b38fa1f8339f9/master/w_1600%2Cc_limit/Ideas_Surprised_Pikachu_HD.jpg, TOP:When you discovered GPTDeploy" --model gpt-4 --path microservice
 ```

 <img src="res/meme_example.png" alt="Meme Generator" width="600" />

 ### Rhyme Generator
 ```bash
-gptdeploy generate --description "Given a word, return a list of rhyming words using the datamuse api" --test "hello" --model gpt-4
+gptdeploy generate --description "Given a word, return a list of rhyming words using the datamuse api" --test "hello" --model gpt-4 --path microservice
 ```
 <img src="res/rhyme_generator_example.png" alt="Rhyme Generator" width="600" />

 ### Word Cloud Generator
 ```bash
-gptdeploy generate --description "Generate a word cloud from a given text" --test "Lorem ipsum dolor sit amet, consectetur adipiscing elit." --model gpt-4
+gptdeploy generate --description "Generate a word cloud from a given text" --test "Lorem ipsum dolor sit amet, consectetur adipiscing elit." --model gpt-4 --path microservice
 ```
 <img src="res/word_cloud_example.png" alt="Word Cloud Generator" width="600" />

 ### 3d model info
 ```bash
-gptdeploy generate --description "Given a 3d object, return vertex count and face count" --test "https://raw.githubusercontent.com/polygonjs/polygonjs-assets/master/models/wolf.obj" --model gpt-4
+gptdeploy generate --description "Given a 3d object, return vertex count and face count" --test "https://raw.githubusercontent.com/polygonjs/polygonjs-assets/master/models/wolf.obj" --model gpt-4 --path microservice
 ```
 <img src="res/obj_info_example.png" alt="3D Model Info" width="600" />

+### 2d rendering of 3d model
+```bash
+gptdeploy generate --description "create a 2d rendering of a whole 3d object and x,y,z object rotation using trimesh and pyrender.OffscreenRenderer with os.environ['PYOPENGL_PLATFORM'] = 'egl' and freeglut3-dev library" --test "input: https://graphics.stanford.edu/courses/cs148-10-summer/as3/code/as3/teapot.obj output: assert the image is not completely white or black" --model gpt-4 --path microservice
+```
+<img src="res/obj_render_example.gif" alt="2D Rendering of 3D Model" width="600" />
+
 ### Table extraction
 ```bash
-gptdeploy generate --description "Given a URL, extract all tables as csv" --test "http://www.ins.tn/statistiques/90" --model gpt-4
+gptdeploy generate --description "Given a URL, extract all tables as csv" --test "http://www.ins.tn/statistiques/90" --model gpt-4 --path microservice
 ```
 <img src="res/table_extraction_example.png" alt="Table Extraction" width="600" />

 ### Audio to mel spectrogram
 ```bash
-gptdeploy generate --description "Create mel spectrograms from audio file" --test "https://cdn.pixabay.com/download/audio/2023/02/28/audio_550d815fa5.mp3" --model gpt-4
+gptdeploy generate --description "Create mel spectrograms from audio file" --test "https://cdn.pixabay.com/download/audio/2023/02/28/audio_550d815fa5.mp3" --model gpt-4 --path microservice
 ```
 <img src="res/audio_to_mel_example.png" alt="Audio to Mel Spectrogram" width="600" />

 ### Text to speech
 ```bash
-gptdeploy generate --description "Convert text to speech" --test "Hello, welcome to GPT Deploy!" --model gpt-4
+gptdeploy generate --description "Convert text to speech" --test "Hello, welcome to GPT Deploy!" --model gpt-4 --path microservice
 ```
 <a href=res/text_to_speech_example.wav><img src="res/text_to_speech_example.png" alt="Text to Speech" width="600" /></a>

@@ -158,20 +171,20 @@ gptdeploy generate --description "Convert text to speech" --test "Hello, welcome

 ### Heatmap Generator
 ```bash
-gptdeploy generate --description "Create a heatmap from an image and a list of relative coordinates" --test "https://images.unsplash.com/photo-1574786198875-49f5d09fe2d2, [[0.1, 0.2], [0.3, 0.4], [0.5, 0.6], [0.2, 0.1], [0.7, 0.2], [0.4, 0.2]]" --model gpt-4
+gptdeploy generate --description "Create a heatmap from an image and a list of relative coordinates" --test "https://images.unsplash.com/photo-1574786198875-49f5d09fe2d2, [[0.1, 0.2], [0.3, 0.4], [0.5, 0.6], [0.2, 0.1], [0.7, 0.2], [0.4, 0.2]]" --model gpt-4 --path microservice
 ```
 <img src="res/heatmap_example.png" alt="Heatmap Generator" width="600" />

 ### QR Code Generator
 ```bash
-gptdeploy generate --description "Generate QR code from URL" --test "https://www.example.com" --model gpt-4
+gptdeploy generate --description "Generate QR code from URL" --test "https://www.example.com" --model gpt-4 --path microservice
 ```
 <img src="res/qr_example.png" alt="QR Code Generator" width="600" />

 ### Mandelbrot Set Visualizer

 ```bash
-gptdeploy generate --description "Visualize the Mandelbrot set with custom parameters" --test "center=-0+1i, zoom=1.0, size=800x800, iterations=1000" --model gpt-4
+gptdeploy generate --description "Visualize the Mandelbrot set with custom parameters" --test "center=-0+1i, zoom=1.0, size=800x800, iterations=1000" --model gpt-4 --path microservice
 ```
 <img src="res/mandelbrot_example.png" alt="Mandelbrot Set Visualizer" width="600" />

@@ -341,14 +354,6 @@ gptdeploy generate --description "Visualize the Mandelbrot set with custom param

 [//]: # (## Upcoming Challenges)

-[//]: # (### Chemical Structure Drawing)
-
-[//]: # (```bash)
-
-[//]: # (gptdeploy generate --description "Convert a chemical formula into a 2D chemical structure diagram" --test "C6H6")
-
-[//]: # (```)
-
 [//]: # (### Color Palette Generator)

 [//]: # (```bash)
--- a/res/chemical_formula_example.png
+++ b/res/chemical_formula_example.png
--- a/res/obj_render_example.gif
+++ b/res/obj_render_example.gif
--- a/setup.py
+++ b/setup.py
@@ -7,7 +7,7 @@ def read_requirements():

 setup(
    name='gptdeploy',
-    version='0.18.16',
+    version='0.18.18',
    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',
--- a/src/init.py
+++ b/src/init.py
@@ -1,2 +1,2 @@
-__version__ = '0.18.16'
+__version__ = '0.18.18'
 from src.cli import main
--- a/src/apis/gpt.py
+++ b/src/apis/gpt.py
@@ -15,7 +15,9 @@ from src.utils.string_tools import print_colored


 class GPTSession:
-    def __init__(self, model: str = 'gpt-4'):
+    def __init__(self, task_description, test_description, model: str = 'gpt-4', ):
+        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'

@@ -51,7 +53,7 @@ If you have updated it already, please restart your terminal.
            return False

    def get_conversation(self, system_definition_examples: List[str] = ['executor', 'docarray', 'client']):
-        return _GPTConversation(self.model_name, system_definition_examples)
+        return _GPTConversation(self.model_name, self.task_description, self.test_description, system_definition_examples)


 class AssistantStreamingStdOutCallbackHandler(StreamingStdOutCallbackHandler):
@@ -62,7 +64,7 @@ class AssistantStreamingStdOutCallbackHandler(StreamingStdOutCallbackHandler):


 class _GPTConversation:
-    def __init__(self, model: str, system_definition_examples: List[str] = ['executor', 'docarray', 'client']):
+    def __init__(self, model: str, task_description, test_description, system_definition_examples: List[str] = ['executor', 'docarray', 'client']):
        self.chat = ChatOpenAI(
            model_name=model,
            streaming=True,
@@ -85,7 +87,7 @@ class _GPTConversation:
        return response

    @staticmethod
-    def _create_system_message(system_definition_examples: List[str] = []) -> SystemMessage:
+    def _create_system_message(task_description, test_description, system_definition_examples: List[str] = []) -> SystemMessage:
        system_message = system_message_base
        if 'executor' in system_definition_examples:
            system_message += f'\n{executor_example}'
@@ -93,4 +95,5 @@ class _GPTConversation:
            system_message += f'\n{docarray_example}'
        if 'client' in system_definition_examples:
            system_message += f'\n{client_example}'
+        # create from template
        return SystemMessage(content=system_message)
--- a/src/apis/jina_cloud.py
+++ b/src/apis/jina_cloud.py
@@ -248,10 +248,18 @@ def update_client_line_in_file(file_path, host):
        file.write(replaced_content)


-def remove_after_stderr(relevant_lines):
+def shorten_logs(relevant_lines):
+    # handle duplicate error messages
    for index, line in enumerate(relevant_lines):
        if '--- Captured stderr call ----' in line:
-            return relevant_lines[:index]
+            relevant_lines = relevant_lines[:index]
+    # filter pip install logs
+    relevant_lines = [line for line in relevant_lines if ' Requirement already satisfied: ' not in line]
+    # filter version not found logs
+    for index, line in enumerate(relevant_lines):
+        if 'ERROR: Could not find a version that satisfies the requirement ' in line:
+            start_and_end = line[:150] + '...' + line[-150:]
+            relevant_lines[index] = start_and_end
    return relevant_lines


@@ -270,9 +278,9 @@ def process_error_message(error_message):
    if last_matching_line_index is not None:
        relevant_lines = lines[last_matching_line_index:]

-    relevant_lines = remove_after_stderr(relevant_lines)
+    relevant_lines = shorten_logs(relevant_lines)

-    response = '\n'.join(relevant_lines[-25:]).strip()
+    response = '\n'.join(relevant_lines[-100:]).strip()

    # the following code tests the case that the docker file is corrupted and can not be parsed
    # the method above will not return a relevant error message in this case
@@ -282,21 +290,3 @@ def process_error_message(error_message):
    if not response and last_line.startswith('error: '):
        return last_line
    return response
-
-
-def build_docker(path):
-    # The command to build the Docker image
-    cmd = f"docker build -t micromagic {path}"
-
-    # Run the command and capture the output
-    process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True)
-    stdout, stderr = process.communicate()
-
-    # Check if there was an error
-    if process.returncode != 0:
-        error_message = stderr.decode("utf-8")
-        relevant_error_message = process_error_message(error_message)
-        return relevant_error_message
-    else:
-        print("Docker build completed successfully.")
-        return ''
--- a/src/cli.py
+++ b/src/cli.py
@@ -64,13 +64,12 @@ def generate(
            return

    from src.options.generate.generator import Generator
-    generator = Generator(model=model)
+    generator = Generator(description, test, model=model)
    with get_openai_callback() as cb:
-        generator.generate(description, test, path)
+        generator.generate(path)
        print(f"Prompt/Completion/Total Tokens: {cb.prompt_tokens}/{cb.completion_tokens}/{cb.total_tokens}")
        print(f"Total Cost on OpenAI (USD): ${cb.total_cost}")

-
@main.command()
@path_param
 def run(path):
--- a/src/constants.py
+++ b/src/constants.py
@@ -30,5 +30,6 @@ MAX_DEBUGGING_ITERATIONS = 10
 DEMO_TOKEN = '45372338e04f5a41af949024db929d46'

 PROBLEMATIC_PACKAGES = [
-    'Pyrender', 'Trimesh', 'ModernGL', 'PyOpenGL', 'Pyglet', 'pythreejs', 'panda3d' # because they need a screen
+    # 'Pyrender', 'Trimesh',
+    'ModernGL', 'PyOpenGL', 'Pyglet', 'pythreejs', 'panda3d' # because they need a screen
 ]
--- a/src/options/init.py
+++ b/src/options/init.py
@@ -1,13 +1,19 @@
 import os


-def get_latest_folder(path):
-    return max([os.path.join(path, f) for f in os.listdir(path) if os.path.isdir(os.path.join(path, f))])
+def get_latest_folder(path, max_fn=max):
+    return max_fn([os.path.join(path, f) for f in os.listdir(path) if os.path.isdir(os.path.join(path, f))])
+
+def version_max_fn(path_list):
+    version_list = [int(path.split('/')[-1].replace('v', '')) for path in path_list]
+    max_version = max(version_list)
+    max_index = version_list.index(max_version)
+    return path_list[max_index]

 def get_latest_version_path(microservice_path):
    executor_name_path = get_latest_folder(microservice_path)
    latest_approach_path = get_latest_folder(executor_name_path)
-    latest_version_path = get_latest_folder(latest_approach_path)
+    latest_version_path = get_latest_folder(latest_approach_path, max_fn=version_max_fn)
    return latest_version_path

 def get_executor_name(microservice_path):
--- a/src/options/generate/generator.py
+++ b/src/options/generate/generator.py
@@ -17,22 +17,23 @@ from src.utils.string_tools import print_colored


 class Generator:
-    def __init__(self, model='gpt-4'):
-        self.gpt_session = gpt.GPTSession(model=model)
+    def __init__(self, task_description, test_description, model='gpt-4'):
+        self.gpt_session = gpt.GPTSession(task_description, test_description, model=model)
+        self.task_description = task_description
+        self.test_description = test_description

    def extract_content_from_result(self, plain_text, file_name, match_single_block=False):
        pattern = fr"^\*\*{file_name}\*\*\n```(?:\w+\n)?([\s\S]*?)```"
        match = re.search(pattern, plain_text, re.MULTILINE)
        if match:
            return match.group(1).strip()
-        else:
+        elif match_single_block:
            # Check for a single code block
            single_code_block_pattern = r"^```(?:\w+\n)?([\s\S]*?)```"
            single_code_block_match = re.findall(single_code_block_pattern, plain_text, re.MULTILINE)
-            if match_single_block and len(single_code_block_match) == 1:
+            if len(single_code_block_match) == 1:
                return single_code_block_match[0].strip()
-            else:
-                return ''
+        return ''

    def write_config_yml(self, microservice_name, dest_folder):
        config_content = f'''
@@ -58,10 +59,10 @@ metas:
            test,
            path,
            microservice_name,
-            package,
+            packages,
            num_approach,
    ):
-        MICROSERVICE_FOLDER_v1 = get_microservice_path(path, microservice_name, package, num_approach, 1)
+        MICROSERVICE_FOLDER_v1 = get_microservice_path(path, microservice_name, packages, num_approach, 1)
        os.makedirs(MICROSERVICE_FOLDER_v1)

        print_colored('', '############# Microservice #############', 'blue')
@@ -168,7 +169,6 @@ metas:
        persist_file(playground_content, os.path.join(microservice_path, 'app.py'))

    def debug_microservice(self, path, microservice_name, num_approach, packages, description, test):
-        error_before = ''
        for i in range(1, MAX_DEBUGGING_ITERATIONS):
            print('Debugging iteration', i)
            print('Trying to build the microservice. Might take a while...')
@@ -178,20 +178,22 @@ metas:
            error = process_error_message(log_hubble)
            if error:
                print('An error occurred during the build process. Feeding the error back to the assistent...')
-                self.do_debug_iteration(description, error, error_before, next_microservice_path,
+                self.do_debug_iteration(description, error, next_microservice_path,
                                        previous_microservice_path, test)
-                error_before = error
+                if i == MAX_DEBUGGING_ITERATIONS - 1:
+                    raise self.MaxDebugTimeReachedException('Could not debug the microservice.')
            else:
                print('Successfully build microservice.')
                break
-            if i == MAX_DEBUGGING_ITERATIONS - 1:
-                raise self.MaxDebugTimeReachedException('Could not debug the microservice.')
+
        return get_microservice_path(path, microservice_name, packages, num_approach, i)

-    def do_debug_iteration(self, description, error, error_before, next_microservice_path, previous_microservice_path,
+    def do_debug_iteration(self, description, error, next_microservice_path, previous_microservice_path,
                           test):
        os.makedirs(next_microservice_path)
        file_name_to_content = get_all_microservice_files_with_content(previous_microservice_path)
+
+        summarized_error = self.summarize_error(error)
        is_dependency_issue = self.is_dependency_issue(error, file_name_to_content['Dockerfile'])
        if is_dependency_issue:
            all_files_string = self.files_to_string({
@@ -199,11 +201,11 @@ metas:
                key in ['requirements.txt', 'Dockerfile']
            })
            user_query = template_solve_dependency_issue.format(
-                description=description, error=error, all_files_string=all_files_string,
+                description=description, summarized_error=summarized_error, all_files_string=all_files_string,
            )
        else:
            user_query = template_solve_code_issue.format(
-                description=description, error=error, all_files_string=self.files_to_string(file_name_to_content),
+                description=description, summarized_error=summarized_error, all_files_string=self.files_to_string(file_name_to_content),
            )
        conversation = self.gpt_session.get_conversation()
        returned_files_raw = conversation.chat(user_query)
@@ -211,6 +213,7 @@ metas:
            updated_file = self.extract_content_from_result(returned_files_raw, file_name)
            if updated_file and (not is_dependency_issue or file_name in ['requirements.txt', 'Dockerfile']):
                file_name_to_content[file_name] = updated_file
+                print(f'Updated {file_name}')
        for file_name, content in file_name_to_content.items():
            persist_file(content, os.path.join(next_microservice_path, file_name))

@@ -240,12 +243,12 @@ metas:
            template_generate_possible_packages.format(description=description)
        )
        packages_csv_string = self.extract_content_from_result(packages_raw, 'packages.csv')
-        packages = [package.split(',') for package in packages_csv_string.split('\n')]
-        packages = packages[:NUM_IMPLEMENTATION_STRATEGIES]
-        return packages
+        packages_list = [[pkg.strip() for pkg in packages_string.split(',')] for packages_string in packages_csv_string.split('\n')]
+        packages_list = packages_list[:NUM_IMPLEMENTATION_STRATEGIES]
+        return packages_list

-    def generate(self, description, test, microservice_path):
-        generated_name = self.generate_microservice_name(description)
+    def generate(self, microservice_path):
+        generated_name = self.generate_microservice_name(self.task_description)
        microservice_name = f'{generated_name}{random.randint(0, 10_000_000)}'
        packages_list = self.get_possible_packages(description)
        packages_list = [
@@ -254,7 +257,7 @@ metas:
        for num_approach, packages in enumerate(packages_list):
            try:
                self.generate_microservice(
-                    description, test, microservice_path, microservice_name, packages, num_approach
+                    self.task_description, self.test_description, microservice_path, microservice_name, packages, num_approach
                )
                final_version_path = self.debug_microservice(
                    microservice_path, microservice_name, num_approach, packages, description, test
@@ -274,3 +277,14 @@ gptdeploy deploy --path {microservice_path}
 '''
                  )
            break
+
+    def summarize_error(self, error):
+        conversation = self.gpt_session.get_conversation([])
+        user_query = f'''
+Here is an error message I encountered during the docker build process:
+"{error}"
+Your task is to summarize the error message as compact and informative as possible while maintaining all information necessary to debug the core issue.
+Warnings are not worth mentioning.
+'''
+        error_summary = conversation.query(user_query)
+        return error_summary
--- a/src/options/generate/prompt_system.py
+++ b/src/options/generate/prompt_system.py
@@ -16,7 +16,7 @@ class MyInfoExecutor(Executor):
        for d in docs:
            content = json.loads(d.text)
            ...
-            d.text = json.dumps(modified_content)
+            d.text = json.dumps(modified_content) # serialized json
        return docs
 ```

@@ -28,7 +28,7 @@ A Document is a python class that represents a single document.
 Here is the protobuf definition of a Document:
 ```
 message DocumentProto {{
-  // used to store json data the executor gets and returns
+  // used to store serialized json data the executor gets and returns
  string text = 1;
 }}
 ```
@@ -71,7 +71,16 @@ print(response[0].text)
 ```'''


-system_message_base = '''It is the year 2021. 
-You are a principal engineer working at Jina - an open source company. 
+system_base_definition = '''It is the year 2021. 
+You are a principal engineer working at Jina - an open source company.
 You accurately satisfy all of the user's requirements.
-Your goal is to build a microservice that: {description}'''
+To be more specific, you help the user to build a microservice with the following requirements:
+```
+{task_description}
+```
+and the following test scenario:
+```
+{test_description}
+```
+
+You must obey the following rules:''' + f'\n{not_allowed_executor}\n{not_allowed_docker}'
--- a/src/options/generate/prompt_tasks.py
+++ b/src/options/generate/prompt_tasks.py
--- a/src/utils/io.py
+++ b/src/utils/io.py
@@ -9,8 +9,8 @@ import docker
 from docker import APIClient


-def get_microservice_path(path, microservice_name, package, num_approach, version):
-    package_path = '_'.join(package)
+def get_microservice_path(path, microservice_name, packages, num_approach, version):
+    package_path = '_'.join(packages)
    return os.path.join(path, microservice_name, f'{num_approach}_{package_path}', f'v{version}')

 def persist_file(file_content, file_path):
--- a/test/test_generator.py
+++ b/test/test_generator.py
@@ -25,5 +25,5 @@ def test_generator(tmpdir):
    # Use mock.patch as a context manager to replace the original methods with the mocks
    with mock.patch("openai.ChatCompletion.create", side_effect=mock_create), \
            mock.patch.object(GPTSession, "configure_openai_api_key", side_effect=mock_get_openai_api_key):
-        generator = Generator()
-        generator.generate("my description", "my test", str(tmpdir))
+        generator = Generator("my description", "my test")
+        generator.generate(str(tmpdir))