Adding a mechanism for self-improvement #353
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Excellent! It seems that the HANDOVER_FILE is necessary. Could you kindly provide it as well? |
geekan
reviewed
Sep 26, 2023
| "QaEngineer": {"name": "Edward", "next": None, "prev": "Engineer"}, | ||
| } | ||
|
|
||
| def print_with_color(text, color="red"): |
| print(f"{color_codes[color]} {text} {color_codes['reset']}") | ||
|
|
||
|
|
||
|
|
| """ | ||
|
|
||
| async def run(self, handover_msg, *args, **kwargs) -> ActionOutput: | ||
| import re |
| import re | ||
| #prev_role = handover_msg[0].to_dict()["role"] | ||
| #prev_msg = handover_msg[0].to_dict()["content"] | ||
| if isinstance(handover_msg, list): |
| @@ -38,6 +38,22 @@ def invest(self, investment: float): | |||
| CONFIG.max_budget = investment | |||
| logger.info(f'Investment: ${investment}.') | |||
|
|
|||
| def improvement(self, initial=False, roles=None): | |||
| handover_file = CONFIG.handover_file | |||
| if initial: | |||
There was a problem hiding this comment.
Would it be better to do this in CONFIG?
| msgs = self.environment.memory.get_by_action(Feedback) | ||
| if isinstance(msgs, list): | ||
| for msg in msgs: | ||
| logger.info(f"{msg.role}'s feedback: {msg.content}") |
There was a problem hiding this comment.
This actually prints the feedback result, which is different from the function name. Should we optimize the function name?
| Now, rewrite your "{role}" constraints in 30 words: | ||
| """ | ||
|
|
||
| # def print_with_color(text, color="red"): |
| # } | ||
| # print(f"{color_codes[color]} {text} {color_codes['reset']}") | ||
|
|
||
| class Reflect(): |
There was a problem hiding this comment.
class Reflect instead of class Reflect()
| class Reflect(): | ||
| def from_feedback(role, constraints): | ||
|
|
||
| chat = OpenAIGPTAPI() |
stellaHSR
reviewed
Sep 26, 2023
|
|
||
| if run_tests: | ||
| # developing features: run tests on the spot and identify bugs | ||
| # (bug fixing capability comes soon!) | ||
| company.hire([QaEngineer()]) | ||
|
|
||
| company.invest(investment) | ||
|
|
||
| if self_improvement: |
There was a problem hiding this comment.
you need add this first, before hire()
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary of Updates
1️⃣ Internal Feedback Action:
Added the Feedback action in
metagpt/actions/internal_feedback.pyto allow internal evaluation among agents (currently supporting the handover process).2️⃣ Action Functions Enhancement:
Introduced two new methods,
_add_action_at_head()and_add_action_at_tail(), for improved clarity and ease of defining actions.3️⃣ Long-Term Memory Utilization:
Commenced utilization of the previously underutilized
longterm_memory, currently supporting storage in a JSON file. Future improvements will involve leveraging existing FAISS support.4️⃣ Reflective Mechanism:
Incorporated the
reflect.pymodule withinmetagpt/learn.5️⃣ Initialization with Feedback:
Implemented feedback-driven constraints initialization during role setup, marking the initial phase of our journey towards self-improvement.
These updates address several urgent issues, including effective long-term memory utilization, internal evaluation mechanisms, and the integration of historical data into the self-improvement mechanism. And there is still ample room for improvement beyond these updates. 😄
Quick usage:
python startup.py "Write a cli snake game" --self_improvement TrueNote: Unit tests are pending implementation due to time constraints and will be added in the near days.