feat: Customizable operation output type

Added function to dynamically change operation output based on prompt directory files. Improves program flexibility and control.

Seanium_Brain.py

@@ -1,5 +1,6 @@
 import streamlit as st
 from modules import utilities as util
+from modules import model_data
 import brain
 import check_update
 import time
@@ -19,6 +20,7 @@ model_options = ['text-davinci-003', 'text-curie-001', 'text-babbage-001', 'text
 header = st.container()
 body = st.container()
 LOG_PATH = '.user/log'
+PROMPT_PATH = '.user/prompt'
 SESSION_TIME = st.session_state['SESSION_TIME']
 CURRENT_LOG_FILE = f'{LOG_PATH}/log_{SESSION_TIME}.log'
 
@@ -58,13 +60,43 @@ def save_as():
         )
 
 
+def process_response(query, target_model, prompt_file: str, data: model_data.param):
+    # check if exclude model is not target model
+    file_name = util.get_file_name(prompt_file)
+    print(f'Processing {file_name}...')
+    with st.spinner(f'Thinking on {file_name}...'):
+        results = brain.run(query, target_model, prompt_file,
+                            data.temp,
+                            data.max_tokens,
+                            data.top_p,
+                            data.frequency_penalty,
+                            data.present_penalty)
+        # displaying results
+        st.header(f'📃{file_name}')
+        st.success(results)
+        time.sleep(1)
+        log(results, delimiter=f'{file_name.upper()}')
+
+
 # sidebar
 with st.sidebar:
     st.title('Settings')
-    output_types = st.multiselect('Output Types', ['Answer', 'Summary'], default=['Answer'])
-    answer_model = st.selectbox('Answer Model', model_options)
-    if util.contains(output_types, 'Summary'):
-        summary_model = st.selectbox('Summary Model', model_options)
+
+    prompt_files = util.scan_directory(PROMPT_PATH)
+    prompt_file_names = [util.get_file_name(file) for file in prompt_files]
+    prompt_dictionary = dict(zip(prompt_file_names, prompt_files))
+    # remove 'my-info' from prompt dictionary
+    prompt_dictionary.pop('my-info')
+
+    operations = st.multiselect('Operations', list(prompt_dictionary.keys()), default=list(prompt_dictionary.keys())[0])
+    other_models = []
+    question_model = st.selectbox('Question Model', model_options)
+
+    operations_no_question = [op for op in operations if op != 'question']
+
+    for operation in operations_no_question:
+        model = st.selectbox(f'{operation} Model', model_options)
+        other_models.append(model)
 
     temp = st.slider('Temperature', 0.0, 1.0, value=0.1)
     max_tokens = st.slider('Max Tokens', 850, 4500, value=1000)
@@ -75,6 +107,14 @@ with st.sidebar:
     chunk_size = st.slider('Chunk Size', 1500, 4500, value=4000)
     chunk_count = st.slider('Answer Count', 1, 5, value=1)
 
+    param = model_data.param(temp=temp,
+                             max_tokens=max_tokens,
+                             top_p=top_p,
+                             frequency_penalty=freq_panl,
+                             present_penalty=pres_panl,
+                             chunk_size=chunk_size,
+                             chunk_count=chunk_count)
+
     if st.button('Clear Log', on_click=clear_log):
         st.success('Log Cleared')
 with header:
@@ -95,23 +135,33 @@ def execute_brain(q):
 
     # thinking on answer
     with st.spinner('Thinking on Answer'):
-        answer = brain.run_answer(q, answer_model, temp, max_tokens, top_p, freq_panl, pres_panl,
+        answer = brain.run_answer(q, question_model, temp, max_tokens, top_p, freq_panl, pres_panl,
                                   chunk_count=chunk_count)
-        if util.contains(output_types, 'Answer'):
+        if util.contains(operations, 'question'):
             # displaying results
             st.header('💬Answer')
             st.success(answer)
+            time.sleep(1)
             log(answer, delimiter='ANSWER')
 
-    # thinking on summary
-    if util.contains(output_types, 'Summary'):
-        with st.spinner('Thinking on Summary'):
-            time.sleep(2)
-            summary = brain.run_summary(answer, summary_model, temp, max_tokens, top_p, freq_panl, pres_panl)
-            # displaying results
-            st.header('📃Summary')
-            st.success(summary)
-            log(summary, delimiter='SUMMARY')
+    # thinking on other outputs
+
+    if len(operations_no_question) > 0:
+        for i in range(len(operations_no_question)):
+            prompt_path = prompt_dictionary[operations_no_question[i]]
+            other_model = other_models[i]
+            process_response(answer, other_model, prompt_path, param)
+
+
+
+    # # thinking on summary
+    # if util.contains(output_types, 'Summary'):
+    #     with st.spinner('Thinking on Summary'):
+    #         summary = brain.run_summary(answer, summary_model, temp, max_tokens, top_p, freq_panl, pres_panl)
+    #         # displaying results
+    #         st.header('📃Summary')
+    #         st.success(summary)
+    #         log(summary, delimiter='SUMMARY')
 
 
 # main

brain.py

@@ -44,17 +44,15 @@ def run_answer(query, model, temp, max_tokens, top_p, freq_penl, pres_penl, chun
         answers.append(answer)
 
     all_answers = '\n\n'.join(answers)
-    # print('\n\n============ANSWER============\n\n', all_answers)
     return all_answers
 
 
-def run_summary(query, model, temp, max_tokens, top_p, freq_penl, pres_penl):
+def run(query, model, prompt_file, temp, max_tokens, top_p, freq_penl, pres_penl):
     chunks = textwrap.wrap(query, 10000)
-    summaries = []
+    responses = []
     for chunk in chunks:
-        prompt = util.read_file(f'{prompt_dir}/summarize.txt').replace('<<SUM>>', chunk)
-        summary = gpt.gpt3(prompt, model, temp, max_tokens, top_p, freq_penl, pres_penl)
-        summaries.append(summary)
-    all_summary = '\n\n'.join(summaries)
-    # print('\n\n============SUMMRY============\n\n', all_summary)
-    return all_summary
+        prompt = util.read_file(prompt_file).replace('<<DATA>>', chunk)
+        response = gpt.gpt3(prompt, model, temp, max_tokens, top_p, freq_penl, pres_penl)
+        responses.append(response)
+    all_response = '\n\n'.join(responses)
+    return all_response

example_prompt/summarize.txt

@@ -1,6 +1,6 @@
 Write notes about the following document. Summarize the text into a single clear， detailed dot points that are easy to read， and in a way that it makes sense.
 Include most important data.
 
-<<SUM>>
+<<DATA>>
 
 NOTES:

modules/model_data.py

@@ -0,0 +1,9 @@
+class param:
+    def __init__(self, temp, max_tokens, top_p, frequency_penalty, present_penalty, chunk_count, chunk_size):
+        self.temp = temp
+        self.max_tokens = max_tokens
+        self.top_p = top_p
+        self.frequency_penalty = frequency_penalty
+        self.present_penalty = present_penalty
+        self.chunk_count = chunk_count
+        self.chunk_size = chunk_size

modules/prompt_data.py

@@ -0,0 +1,6 @@
+
+class prompt:
+    def __init__(self, name, path, model):
+        self.name = name
+        self.path = path
+        self.model = model

modules/utilities.py

@@ -29,12 +29,22 @@ def extract_string(text, delimiter, force=False, join=True, split_mode=False):
 
 
 def remove_oldest_file(directory, max_files):
-    files = glob.glob(f'{directory}/*')
+    files = scan_directory(directory)
     if len(files) >= max_files:
         oldest_file = min(files, key=os.path.getctime)
         os.remove(oldest_file)
 
 
+def scan_directory(directory):
+    files = glob.glob(f'{directory}/*')
+    return files
+
+
+# get file name without extension
+def get_file_name(filepath):
+    return os.path.splitext(os.path.basename(filepath))[0]
+
+
 def create_path_not_exist(path):
     directory = os.path.dirname(path)
     if not os.path.exists(directory):