aljaz/gpt-engineer
1
0
Fork 0
mirror of https://github.com/aljazceru/gpt-engineer.git synced 2025-12-18 13:15:00 +01:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into add-makefile

Browse Source
This commit is contained in:
PatillaCode
2023-06-18 03:41:31 +02:00
parent 95cac98e2b fa7585f900
commit f6fc442426
10 changed files with 121 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.gitignore vendored
View File

@@ -19,6 +19,7 @@ ENV/
# IDE-specific files
.vscode/
.idea/
# Compiled Python modules
*.pyc
@@ -36,3 +37,4 @@ archive
# any log file
*log.txt
todo

10
README.md
View File

@@ -3,6 +3,8 @@
GPT Engineer is made to be easy to adapt, extend, and make your agent learn how you want your code to look. It generates an entire codebase based on a prompt.
[Demo](https://twitter.com/antonosika/status/1667641038104674306) 👶🤖
## Project philosophy
- Simple to get value
- Flexible and easy to add new own "AI steps". See `steps.py`.
@@ -13,10 +15,6 @@ GPT Engineer is made to be easy to adapt, extend, and make your agent learn how
- Simplicity, all computation is "resumable" and persisted to the filesystem
![output](https://github.com/AntonOsika/gpt-engineer/assets/4467025/a6938d43-2ac1-4cf1-98d1-93eea1bdfce4)
## Usage
**Setup**:
@@ -40,9 +38,9 @@ Contributors welcome! If you are unsure what to add, check out the ideas listed
## Features
You can specify the "identity" of the AI agent by editing the files in the `identity` folder.
Editing the identity, and evolving the main_prompt, is currently how you make the agent remember things between projects.
Editing the identity, and evolving the `main_prompt`, is currently how you make the agent remember things between projects.
Each step in steps.py will have its communication history with GPT4 stored in the logs folder, and can be rerun with scripts/rerun_edited_message_logs.py.
Each step in `steps.py` will have its communication history with GPT4 stored in the logs folder, and can be rerun with `scripts/rerun_edited_message_logs.py`.
## Contributing
If you want to contribute, please check out the [projects](https://github.com/AntonOsika/gpt-engineer/projects?query=is%3Aopen) or [issues tab](https://github.com/AntonOsika/gpt-engineer/issues) in the GitHub repo and please read the [contributing document](.github/CONTRIBUTING.md) on how to contribute.

130
gpt_engineer/steps.py
View File

@@ -8,10 +8,10 @@ from gpt_engineer.chat_to_files import parse_chat
def setup_sys_prompt(dbs):
return dbs.identity["setup"] + "\nUseful to know:\n" + dbs.identity["philosophy"]
return dbs.identity["generate"] + "\nUseful to know:\n" + dbs.identity["philosophy"]
def run(ai: AI, dbs: DBs):
def simple_gen(ai: AI, dbs: DBs):
"""Run the AI on the main prompt and save the results"""
messages = ai.start(
setup_sys_prompt(dbs),
@@ -30,7 +30,7 @@ def clarify(ai: AI, dbs: DBs):
while True:
messages = ai.next(messages, user)
if messages[-1]['content'].strip().lower().startswith("no"):
if messages[-1]["content"].strip().lower().startswith("no"):
break
print()
@@ -53,34 +53,70 @@ def clarify(ai: AI, dbs: DBs):
def gen_spec(ai: AI, dbs: DBs):
'''
"""
Generate a spec from the main prompt + clarifications and save the results to the workspace
'''
messages = [ai.fsystem(setup_sys_prompt(dbs)), ai.fsystem(f"Main prompt: {dbs.input['main_prompt']}")]
"""
messages = [
ai.fsystem(setup_sys_prompt(dbs)),
ai.fsystem(f"Instructions: {dbs.input['main_prompt']}"),
]
messages = ai.next(messages, dbs.identity['spec'])
messages = ai.next(messages, dbs.identity['respec'])
messages = ai.next(messages, dbs.identity['spec'])
messages = ai.next(messages, dbs.identity["spec"])
dbs.memory['specification'] = messages[-1]['content']
dbs.memory["specification"] = messages[-1]["content"]
return messages
def pre_unit_tests(ai: AI, dbs: DBs):
'''
def respec(ai: AI, dbs: DBs):
messages = dbs.logs[gen_spec.__name__]
messages += [ai.fsystem(dbs.identity["respec"])]
messages = ai.next(messages)
messages = ai.next(
messages,
(
'Based on the conversation so far, please reiterate the specification for the program. '
'If there are things that can be improved, please incorporate the improvements. '
"If you are satisfied with the specification, just write out the specification word by word again."
)
)
dbs.memory["specification"] = messages[-1]["content"]
return messages
def gen_unit_tests(ai: AI, dbs: DBs):
"""
Generate unit tests based on the specification, that should work.
'''
messages = [ai.fsystem(setup_sys_prompt(dbs)), ai.fuser(f"Instructions: {dbs.input['main_prompt']}"), ai.fuser(f"Specification:\n\n{dbs.memory['specification']}")]
"""
messages = [
ai.fsystem(setup_sys_prompt(dbs)),
ai.fuser(f"Instructions: {dbs.input['main_prompt']}"),
ai.fuser(f"Specification:\n\n{dbs.memory['specification']}"),
]
messages = ai.next(messages, dbs.identity['unit_tests'])
messages = ai.next(messages, dbs.identity["unit_tests"])
dbs.memory['unit_tests'] = messages[-1]['content']
to_files(dbs.memory['unit_tests'], dbs.workspace)
dbs.memory["unit_tests"] = messages[-1]["content"]
to_files(dbs.memory["unit_tests"], dbs.workspace)
return messages
def run_clarified(ai: AI, dbs: DBs):
def gen_clarified_code(ai: AI, dbs: DBs):
# get the messages from previous step
messages = json.loads(dbs.logs[clarify.__name__])
messages = [
ai.fsystem(setup_sys_prompt(dbs)),
] + messages[1:]
messages = ai.next(messages, dbs.identity["use_qa"])
to_files(messages[-1]["content"], dbs.workspace)
return messages
def gen_code(ai: AI, dbs: DBs):
# get the messages from previous step
messages = [
@@ -95,39 +131,55 @@ def run_clarified(ai: AI, dbs: DBs):
def execute_workspace(ai: AI, dbs: DBs):
messages = ai.start(
system=(
f"You will get infomation about a codebase that is currently on disk in the folder {dbs.workspace.path}.\n"
"From this you will answer with one code block that includes all the necessary macos terminal commands to "
"a) install dependencies "
"b) run the necessary parts of the codebase to try it.\n"
"Do not explain the code, just give the commands.\n"
),
user="Information about the codebase:\n\n" + dbs.workspace["all_output.txt"],
)
messages = gen_entrypoint(ai, dbs)
execute_entrypoint(ai, dbs)
return messages
[[lang, command]] = parse_chat(messages[-1]['content'])
assert lang in ['', 'bash', 'sh']
print('Do you want to execute this code?')
def execute_entrypoint(ai, dbs):
command = dbs.workspace["run.sh"]
print("Do you want to execute this code?")
print()
print(command)
print()
print('If yes, press enter. If no, type "no"')
print()
if input() == 'no':
print('Ok, not executing the code.')
return messages
print('Executing the code...')
if input() == "no":
print("Ok, not executing the code.")
print("Executing the code...")
print()
subprocess.run(command, shell=True)
subprocess.run("bash run.sh", shell=True, cwd=dbs.workspace.path)
return []
def gen_entrypoint(ai, dbs):
messages = ai.start(
system=(
f"You will get information about a codebase that is currently on disk in the current folder.\n"
"From this you will answer with one code block that includes all the necessary macos terminal commands to "
"a) install dependencies "
"b) run all necessary parts of the codebase (in parallell if necessary).\n"
"Do not install globally. Do not use sudo.\n"
"Do not explain the code, just give the commands.\n"
),
user="Information about the codebase:\n\n" + dbs.workspace["all_output.txt"],
)
print()
[[lang, command]] = parse_chat(messages[-1]["content"])
assert lang in ["", "bash", "sh"]
dbs.workspace["run.sh"] = command
return messages
# Different configs of what steps to run
STEPS = {
'default': [gen_spec, pre_unit_tests, run_clarified, execute_workspace],
'simple': [run, execute_workspace],
'clarify': [clarify, run_clarified],
"default": [gen_spec, gen_unit_tests, gen_code, execute_workspace],
"benchmark": [gen_spec, gen_unit_tests, gen_code, gen_entrypoint],
"simple": [simple_gen, execute_workspace],
"clarify": [clarify, gen_clarified_code, execute_workspace],
"respec": [gen_spec, respec, gen_unit_tests, gen_code, execute_workspace],
"execute_only": [execute_entrypoint],
}
# Future steps that can be added:

3
identity/setup → identity/generate
View File

@@ -1,7 +1,8 @@
You will get instructions for code to write.
You will write a very long answer. Make sure that every detail of the architecture is, in the end, implemented as code.
You will first lay out the names of the core classes, functions, methods that will be necessary, As well as a quick comment on their purpose.
Think step by step and reason yourself to the right decisions to make sure we get it right.
You will first lay out the names of the core classes, functions, methods that will be necessary, as well as a quick comment on their purpose.
Then you will output the content of each file, with syntax below.
(You will start with the "entrypoint" file, then go to the ones that are imported by that file, and so on.)
Make sure that files contain all imports, types etc. Make sure that code in different files are compatible with each other.

2
identity/qa
View File

@@ -1,3 +1,3 @@
You will read instructions and not carry them out, only seek to clarify them.
Specifically you will first summarise a list of super short bullets of areas that need clarification.
Then you will pick one clarifying question, and wait for an answer from the user.
Then you will pick one clarifying question, and wait for an answer from the user.

14
identity/respec
View File

@@ -1,8 +1,12 @@
You are a pragmatic principal engineer at Google. You have been asked to review a specification for a new feature.
You are a pragmatic principal engineer at Google.
You have been asked to review a specification for a new feature by a previous version of yourself
You have been asked to give feedback on the following:
- Is there anything that might not work the way the user expects?
- Is there anything missing for the program to fully work?
- Is there anything that can be simplified without decreasing quality?
- Is there anything that might not work the way intended by the instructions?
- Is there anything in the specification missing for the program to work as expected?
- Is there anything that can be simplified without significant drawback?
You are asked to make educated assumptions for each unclear item. For each of these, communicate which assumptions you'll make when implementing the feature.
You are asked to make educated assumptions for each unclear item.
For each of these, communicate which assumptions you'll make when implementing the feature.
Think step by step to make sure we don't miss anything.

13
identity/spec
View File

@@ -1,8 +1,11 @@
You are a super smart developer and an AI developed to write programs. You have been asked to make a specification for a program.
You are a super smart developer. You have been asked to make a specification for a program.
Please generate a specification based on the given input. First, be super explicit about what the program should do, which features it should have and give details about anything that might be unclear. **Don't leave anything unclear or undefined.**
Think step by step to make sure we get a high quality specification and we don't miss anything.
First, be super explicit about what the program should do, which features it should have
and give details about anything that might be unclear. **Don't leave anything unclear or undefined.**
Second, lay out the names of the core classes, functions, methods that will be necessary, As well as a quick comment on their purpose.
Then write out which non-standard dependencies you'll have to use.
Second, lay out the names of the core classes, functions, methods that will be necessary,
as well as a quick comment on their purpose.
Finally list all non-standard dependencies you'll have to use.
This specification will be used later as the basis for your implementation.
This specification will be used later as the basis for the implementation.

2
identity/unit_tests
View File

@@ -1,3 +1,3 @@
You are a super smart developer and an AI developed to use Test Driven Development to write tests according to a specification.
You are a super smart developer using Test Driven Development to write tests according to a specification.
Please generate tests based on the above specification. The tests should be as simple as possible, but still cover all the functionality.

3
identity/use_qa
View File

@@ -1,6 +1,7 @@
Please now remember the steps:
First lay out the names of the core classes, functions, methods that will be necessary, As well as a quick comment on their purpose.
Think step by step and reason yourself to the right decisions to make sure we get it right.
First lay out the names of the core classes, functions, methods that will be necessary, as well as a quick comment on their purpose.
Then output the content of each file, with syntax below.
(You will start with the "entrypoint" file, then go to the ones that are imported by that file, and so on.)
Make sure that files contain all imports, types, variables etc. The code should be fully functional. If anything is unclear, just make assumptions. Make sure that code in different files are compatible with each other.

2
tests/test_db.py
View File

@@ -1,4 +1,4 @@
from ..db import DB
from gpt_engineer.db import DB
def test_db():