Structured Output Revisions #17
Conversation
… and rescue scenario
…to jonathan-dev-rebase Merging Connor H's updates on adding structured output, and removing duplicate requests in getting a task fro each our of the day.
…research/agentic_collab into henry-dev-rebase
… and firehouse location. Merge remote-tracking branch 'origin/chowington-search-and-rescue' into jonathan-dev-rebase
… run_gpt_prompt.py using List[str] throught the file, installed the new requirements.txt, added new gpt-key
…research/agentic_collab into henry-dev-rebase
| if debug or verbose: | ||
| print_run_prompts(prompt_template, persona, gpt_param, | ||
| prompt_input, prompt, output) | ||
|
|
||
| return output, [output, prompt, gpt_param, prompt_input, fail_safe] | ||
|
|
||
| #TODO |
There was a problem hiding this comment.
Can you remove these #TODOs, please? Thanks!
| output = ChatGPT_safe_generate_response(prompt, example_output, special_instruction, 3, fail_safe, | ||
| __chat_func_validate, __chat_func_clean_up, True) | ||
| output = generate_structured_response( | ||
| prompt, | ||
| gpt_param, | ||
| Poignancy, | ||
| 3, | ||
| fail_safe, | ||
| __func_validate, | ||
| __func_clean_up, | ||
| True | ||
| ) |
There was a problem hiding this comment.
Let's wait until @LemmyTron can get his structured outputs conversion of ChatGPT_safe_generate_response so we can use that instead of generate_structured_response
|
|
||
| #TODO | ||
| class Conversation(BaseModel): | ||
| log:str |
There was a problem hiding this comment.
Maybe "summary" would make more sense, rather than "log"?
| output = ChatGPT_safe_generate_response(prompt, example_output, special_instruction, 3, fail_safe, | ||
| __chat_func_validate, __chat_func_clean_up, True) | ||
| output = ChatGPT_safe_generate_response( | ||
| prompt, | ||
| gpt_param, | ||
| Conversation, | ||
| 3, | ||
| fail_safe, | ||
| __func_validate, | ||
| __func_clean_up, | ||
| True | ||
| ) | ||
|
|
||
|
|
There was a problem hiding this comment.
ChatGPT_safe_generate_response doesn't take a response format--this probably throws an error? Let's wait until Henry has finished his function
| def __func_clean_up(gpt_response, prompt=""): | ||
| return ''.join(gpt_response.split("---")[0]).strip() | ||
| def __func_clean_up(gpt_response:GameObject, prompt=""): | ||
| return ''.join(gpt_response.object.split("---")[0]).strip() |
There was a problem hiding this comment.
Looks like you forgot to replace safe_generate_response with generate_structured_response below these lines
| output = ChatGPT_safe_generate_response(prompt, example_output, special_instruction, 3, fail_safe, | ||
| __chat_func_validate, __chat_func_clean_up, True) | ||
| output = ChatGPT_safe_generate_response( | ||
| prompt, | ||
| gpt_param, | ||
| Pronunciatio, | ||
| 3, | ||
| fail_safe, | ||
| __func_validate, | ||
| __func_clean_up, | ||
| True | ||
| ) |
There was a problem hiding this comment.
Same here--can't use this function for structured outputs
| output = ChatGPT_safe_generate_response(prompt, example_output, special_instruction, 3, fail_safe, | ||
| __chat_func_validate, __chat_func_clean_up, True) | ||
| output = ChatGPT_safe_generate_response( | ||
| prompt, | ||
| gpt_param, | ||
| Idea_Summary, | ||
| 3, | ||
| fail_safe, | ||
| __func_validate, | ||
| __func_clean_up, | ||
| True | ||
| ) |
There was a problem hiding this comment.
Same here--can't use this function for structured outputs
|
Thanks for the PR, Daniel! Looks like you had merged Henry's and Jonathan's branches into your own--that makes it difficult for me to tell which are actually your changes, unfortunately. I think we should wait until they have merged their PRs before merging this one |
Structured Output Changes for
run_gpt_prompt_generate_whisper_inner_thought
run_gpt_prompt_event_poignancy
run_gpt_prompt_summarize_conversation
run_gpt_prompt_action_game_object
run_gpt_prompt_pronunciation
run_gpt_prompt_summarize_ideas