Added feature to generate and explore ideas

- A new feature that allows generation and exploration of ideas. When the analysis is complete the user can request the variation of the original task to explore and execute the varied versions. User has an ability to set the degree of variation on scale 1-5.
- Updated workflow map to indicate forking, as related to the previous point
- Theme toggle moved under settings
- A few bug fixes

Author: pgalko

PROMPT_TEMPLATES_sample.yaml

@@ -939,6 +939,49 @@ theorist_system: |
 
   Remember to tailor your response appropriately based on whether there is relevant previous context or not.
 
+###########################################
+### EXPLORE IDEAS #########################
+###########################################
+
+ideas_explorer: |
+  Generate 5 questions that explore variations of our current inquiry. Each question should represent a small mutation in the problem space - subtle shifts that may reveal profound insights.
+
+  Variation Level: {}
+
+  Level 1: Refine focus - adjust scope, granularity, or specific parameters within the same inquiry
+  Level 2: Shift angle - examine the same subject from a different perspective or framework  
+  Level 3: Test boundaries - explore what happens at the edges of our current assumptions
+  Level 4: Blend elements - combine two aspects from our current exploration in a new way
+  Level 5: Follow emergence - pursue interesting patterns that appeared unexpectedly in our investigation
+
+  Guidelines for generating variations:
+  - Each variation should be ONE STEP away from the current inquiry
+  - Maintain clear connection to the existing conversation context and any data present
+  - Small changes often reveal large insights - prefer subtle over dramatic
+  - Build variations from elements already present in our discussion
+  - Each question should feel like a natural "what if we considered this slightly differently?"
+  - Preserve the investigative thread while gently pulling in new directions
+
+  Consider how discovery actually works:
+  - Profound insights emerge from minor adjustments in perspective
+  - The most valuable questions are often hiding just outside our current view
+  - Progress happens through accumulation of small exploratory steps
+  - The adjacent possible is always only one step away
+
+  Ground it in the context of the conversation so far, and the given dataset. Format the response as a numbered markdown list.
+  The Task is routed to the Research Specialist, and the response will be formatted as follows:
+
+  1. **question_title:**
+  question
+  2. **question_title:**
+  question
+  3. **question_title:**
+  question
+  4. **question_title:**
+  question
+  5. **question_title:**
+  question
+
 ###########################################
 ### DATAFRAME INSPECTOR AGENT PROMPTS #####
 ###########################################

bambooai/messages/default_prompts.yaml

@@ -939,6 +939,52 @@ theorist_system: |
 
   Remember to tailor your response appropriately based on whether there is relevant previous context or not.
 
+###########################################
+### EXPLORE IDEAS #########################
+###########################################
+
+ideas_explorer: |
+  Generate 5 questions that explore variations of our current inquiry. Each question should represent a small mutation in the problem space - subtle shifts that may reveal profound insights.
+
+  Variation Level: {}
+  
+  Level descriptions (in ascending order from lovest variation to the highest):
+  Level 1: Refine focus - adjust scope, granularity, or specific parameters within the same inquiry
+  Level 2: Shift angle - examine the same subject from a different perspective or framework  
+  Level 3: Test boundaries - explore what happens at the edges of our current assumptions
+  Level 4: Blend elements - combine two aspects from our current exploration in a new way
+  Level 5: Follow emergence - pursue interesting patterns that appeared unexpectedly in our investigation
+
+  Your task: Generate 5 different variations, ALL at Level {}. Each question should explore a different aspect while maintaining the same variation distance from the original inquiry.
+
+  Guidelines for generating variations:
+  - Each variation should be ONE STEP away from the current inquiry
+  - Maintain clear connection to the existing conversation context and any data present
+  - Small changes often reveal large insights - prefer subtle over dramatic
+  - Build variations from elements already present in our discussion
+  - Each question should feel like a natural "what if we considered this slightly differently?"
+  - Preserve the investigative thread while gently pulling in new directions
+
+  Consider how discovery actually works:
+  - Profound insights emerge from minor adjustments in perspective
+  - The most valuable questions are often hiding just outside our current view
+  - Progress happens through accumulation of small exploratory steps
+  - The adjacent possible is always only one step away
+
+  Ground it in the context of the conversation so far, and the given dataset. Format the response as a numbered markdown list.
+  The Task is routed to the Research Specialist, and the response will be formatted as follows:
+
+  1. **question_title:**
+  question
+  2. **question_title:**
+  question
+  3. **question_title:**
+  question
+  4. **question_title:**
+  question
+  5. **question_title:**
+  question
+
 ###########################################
 ### DATAFRAME INSPECTOR AGENT PROMPTS #####
 ###########################################

bambooai/messages/prompts.py

@@ -16,9 +16,9 @@
     "analyst_selector_system", "analyst_selector_user",
     "planner_system", "planner_user_gen", "planner_user_df",
     "planner_user_gen_reasoning", "planner_user_df_reasoning",
-    "theorist_system", "dataframe_inspector_system", "dataframe_inspector_user",
-    "google_search_query_generator_system", "google_search_react_system",
-    "google_search_summarizer_system",
+    "theorist_system", "ideas_explorer", "dataframe_inspector_system", 
+    "dataframe_inspector_user","google_search_query_generator_system", 
+    "google_search_react_system","google_search_summarizer_system",
     "code_generator_system_df", "code_generator_system_gen",
     "code_generator_user_df_plan", "code_generator_user_df_no_plan",
     "code_generator_user_gen_plan", "code_generator_user_gen_no_plan",

requirements.txt

@@ -20,7 +20,7 @@ flask
 ipython
 nbformat
 python-dotenv
-lxml[html_clean]
+lxml_html_clean
 gunicorn
 google-cloud-storage
 statsmodels

setup.py

@@ -6,7 +6,7 @@
 
 setup(
     name='bambooai',
-    version='0.4.19',
+    version='0.4.20',
     description='A lightweight library for working with pandas dataframes using natural language queries',
     long_description=long_description,
     long_description_content_type="text/markdown",
@@ -39,7 +39,7 @@
         'ipython',
         'nbformat',
         'python-dotenv',
-        'lxml[html_clean]',
+        'lxml_html_clean',
         'gunicorn',
         'google-cloud-storage',
         'statsmodels',
@@ -49,6 +49,7 @@
         'yfinance',
         'sweatstack'
     ],
+    python_requires=">=3.10",
     classifiers=[
         'Development Status :: 3 - Alpha',
         'License :: OSI Approved :: MIT License',

web_app/app.py

@@ -889,10 +889,13 @@ def query():
     thread_id = request.json['thread_id']
     chain_id = request.json['chain_id']
     image = request.json.get('image')
-    user_code= request.json.get('user_code')
+    user_code = request.json.get('user_code')
+    branching_cv = request.json.get('branching_cv')
 
     if user_code:
         user_input = "User manually edited your code, and requested to run it, and return the result."
+    elif branching_cv:
+        user_input = bamboo_ai_instance.prompts.ideas_explorer.format(branching_cv, branching_cv)
 
     bamboo_ai_instance.output_manager.add_user_input(user_input)
 

web_app/static/css/styles.css

@@ -1836,6 +1836,116 @@ body, html {
     opacity: 0.8;
 }
 
+/* Branching CV Icons Popup */
+.branching-slider-popup {
+    position: absolute;
+    bottom: calc(100% + 10px);
+    left: 70px; /* Positioned to align with suggest questions button */
+    background: var(--dropdown-bg);
+    border: 1px solid var(--border-color);
+    border-radius: 8px;
+    padding: 4px;
+    display: none;
+    flex-direction: column;
+    gap: 2px;
+    min-width: 40px;
+    box-shadow: 0 2px 10px var(--shadow-color);
+    z-index: 1000;
+}
+
+.branching-slider-popup::before {
+    content: '';
+    position: absolute;
+    bottom: -10px;
+    left: 0;
+    width: 100%;
+    height: 10px;
+}
+
+.branching-slider-popup::after {
+    content: '';
+    position: absolute;
+    bottom: -5px;
+    left: 20px;
+    transform: rotate(45deg);
+    width: 10px;
+    height: 10px;
+    background: var(--dropdown-bg);
+    border-right: 1px solid var(--border-color);
+    border-bottom: 1px solid var(--border-color);
+}
+
+.branching-icon {
+    background: none;
+    border: none;
+    cursor: pointer;
+    color: var(--text-secondary);
+    display: flex;
+    align-items: center;
+    justify-content: center;
+    padding: 6px;
+    border-radius: 4px;
+    transition: all 0.2s;
+    width: 100%;
+}
+
+/* Size scaling from smallest to largest */
+.branching-icon[data-value="1"] svg {
+    width: 14px;
+    height: 14px;
+}
+
+.branching-icon[data-value="2"] svg {
+    width: 16px;
+    height: 16px;
+}
+
+.branching-icon[data-value="3"] svg {
+    width: 18px;
+    height: 18px;
+}
+
+.branching-icon[data-value="4"] svg {
+    width: 20px;
+    height: 20px;
+}
+
+.branching-icon[data-value="5"] svg {
+    width: 22px;
+    height: 22px;
+}
+
+.branching-icon svg {
+    stroke: currentColor;
+    fill: none;
+    transition: all 0.2s;
+}
+
+.branching-icon:hover {
+    background: var(--bg-tertiary);
+    color: var(--accent-color);
+}
+
+/* Dark theme adjustments */
+[data-theme="dark"] .branching-slider-popup {
+    background: var(--dropdown-bg);
+    border-color: var(--border-color);
+}
+
+[data-theme="dark"] .branching-slider-popup::after {
+    background: var(--dropdown-bg);
+    border-color: var(--border-color);
+}
+
+[data-theme="dark"] .branching-icon {
+    color: var(--text-secondary);
+}
+
+[data-theme="dark"] .branching-icon:hover {
+    color: var(--accent-color);
+    background: var(--bg-tertiary);
+}
+
 .switch-container {
     display: flex;
     align-items: center;
@@ -2958,32 +3068,39 @@ code.language-json {
     color: var(--text-tertiary);
 }
 
-/* Theme Toggle Styles */
+/* Theme Toggle in Settings Menu - Match existing menu items exactly */
+.theme-toggle-option .sun-icon,
+.theme-toggle-option .moon-icon {
+    width: 20px;
+    height: 20px;
+    stroke: currentColor;
+    /* Remove any margin - let the default gap: 8px from .menu-option handle spacing */
+}
+
+.theme-toggle-option .theme-label {
+    font-size: 14px; /* Match other menu item text */
+}
+
+/* Hide old theme toggle container */
 .theme-toggle-container {
-    display: inline-block;
-    margin-left: 10px;
+    display: none;
 }
 
-.theme-toggle {
-    background: none;
-    border: none;
-    padding: 4px;
-    cursor: pointer;
-    color: var(--accent-color);
-    display: flex;
-    align-items: center;
-    justify-content: center;
-    transition: transform 0.2s;
+/* Fix submenu trigger - be very specific */
+.integrations-menu-item .integrations-option:hover + .integrations-submenu,
+.integrations-submenu:hover {
+    display: flex !important;
 }
 
-.theme-toggle:hover {
-    transform: scale(1.1);
+/* Ensure the general hover rule doesn't apply to other elements */
+.integrations-menu-item:hover .integrations-submenu {
+    display: none !important;
 }
 
-.theme-toggle svg {
-    width: 20px;
-    height: 20px;
-    stroke: currentColor;
+/* Only show submenu when hovering specifically over the integrations option or the submenu itself */
+.integrations-menu-item .integrations-option:hover + .integrations-submenu,
+.integrations-menu-item .integrations-submenu:hover {
+    display: flex !important;
 }
 
 /* Search Results Styles */
@@ -3530,7 +3647,7 @@ code.language-json {
     margin-top: 0; /* Remove top margin */
 }
 
-#workflowMapContainer g.node:hover rect {
+#workflowMapContainer g.node:hover * {
     stroke: var(--accent-color) !important;
     stroke-width: 1px !important;
 }
@@ -3894,7 +4011,7 @@ code.language-json {
     border: none;
     padding: 0;
     cursor: pointer;
-    color: var(--text-tertiary) !important;
+    color: var(--text-secondary) !important;
     width: 28px;
     height: 28px;
     display: flex;

web_app/static/js/content-rendering.js

@@ -259,6 +259,11 @@ async function handleQuerySubmission(e) {
 //--------------------
 
 function createOrUpdateTab(type, data, id = null, format = null) {
+    // Skip creating query tab entirely if answerTabInteractive is true
+    if (type === 'query' && answerTabInteractive) {
+        return;
+    }
+
     const tabContainer = document.getElementById('tabContainer');
     const contentOutput = document.getElementById('contentOutput');
     let tab = document.getElementById(`tab-${type}`);
@@ -269,7 +274,11 @@ function createOrUpdateTab(type, data, id = null, format = null) {
         tab = document.createElement('div');
         tab.id = `tab-${type}`;
         tab.className = 'tab';
-        tab.textContent = type.charAt(0).toUpperCase() + type.slice(1);
+
+        // Change "Answer" to "Explore" if answerTabInteractive is true
+        const tabName = (type === 'answer' && answerTabInteractive) ? 'Explore' : type.charAt(0).toUpperCase() + type.slice(1);
+        tab.textContent = tabName;
+
         tab.onclick = () => activateTab(type);
         tabContainer.appendChild(tab);
 
@@ -377,7 +386,10 @@ async function updateTabContent(type, data, id = null, format = null) {
             break;
         case 'answer': {
             console.log('Processing answer case');
-            content = `<h3>${type.charAt(0).toUpperCase() + type.slice(1)}:</h3>`;
+
+            // Change header text based on answerTabInteractive
+            const headerText = answerTabInteractive ? 'Explore' : type.charAt(0).toUpperCase() + type.slice(1);
+            content = `<h3>${headerText}:</h3>`;
 
             // First protect LaTeX content
             const { text: protectedContent, placeholders } = protectLatexDelimiters(data);