Feat: Add uv python enviorment, dspy chatbot, ollama, an easy frontend by flask #359

Skywind5487 · 2026-01-07T07:38:55Z

基於 #308 的一個分支。
總之就是加上一些chatBot的code、寫了文檔(use_chatbot_guide.md)、寫了3個自動啟動腳本。
用gemini簡單的搓了前端。

新增了調查者與反對者兩個ChatBot。反對派的prompt可以參考 #170 。

dspy跟uv好用。

預設是已經把整個README.md當作 Ai 的輸入先輸入進去。

使用方法: 先一鍵安裝再一鍵啟動

Co-authored-by: Kelvin <cychandt@connect.ust.hk>

… cache

…ulnerability detection and challenge generation

…port an investigator chatbot

…las-World chatbot

… modules, and add history management. Use dspy to replace whole old ollama logic

…el configuration

Skywind5487 · 2026-01-07T07:45:29Z

前端示意圖

免責聲明同 #170 。

本ChatBot只用於交流分享。並不保證能解決任何模型幻覺、不保證產出的內容正確、不保證指出的問題到位。
不保證ChatBot能解決任何問題，專業的事推薦找專業的人。
對ChatBot不做任何推薦、只做分享。若要使用這個ChatBot請謹慎考慮，是否有任何不好的問題。若有人使用這份Prompt而有任何後果，本人已告知於前。
若有人拿 prompt去為非作歹，並非本人的本意。
如果有任何人覺得自己心裡有任何狀況，建議尋求心理諮商等等的專業人士
本ChatBot同樣不保證任何心理或神經的療效。若有需求，請洽專業的單位，例如諮商師、生命線 1995；張老師服務專線 1980；台灣自殺防治中心 0800 788 995。(來源: BBC AI機器人「聽見我、陪伴我、回應我」，但它會誘導人自殺嗎？)

Copilot

Pull request overview

This pull request adds a chatbot system to the Atlas-World project, featuring dual AI personas (Investigator and Opposition) built with DSPy framework. The implementation includes a Flask-based web backend, an interactive HTML/JavaScript frontend, comprehensive Chinese documentation, and automated setup scripts for deployment with uv and Ollama.

Key Changes

Implements DSPy-based chatbot modules with configurable LLM backends (Ollama, OpenAI, Gemini, Claude)
Adds Flask REST API with /investigate, /oppose, and /reset endpoints
Creates a responsive web UI with Markdown rendering and MathJax support

Reviewed changes

Copilot reviewed 15 out of 18 changed files in this pull request and generated 23 comments.

Show a summary per file

File	Description
`use_chatbot_guide.md`	Chinese user guide with installation and usage instructions
`templates/index.html`	Frontend interface with chat functionality and markdown rendering
`script/chatbot.py`	Flask backend with DSPy integration and endpoint handlers
`script/llm_config.py`	Interactive LLM configuration module with multi-provider support
`script/llm_module/investigator/*`	Investigator chatbot module and signature definition
`script/llm_module/opposition/*`	Opposition chatbot module with extensive prompt file
`pyproject.toml`	Project dependencies configuration for uv
`.python-version`	Python version specification (3.13)
`.gitignore`	Updated ignore patterns for Python and environment files
`.vscode/settings.json`	VS Code workspace configuration
`.idea/.gitignore`	IntelliJ IDEA ignore patterns

Files not reviewed (1)

.idea/.gitignore: Language not supported

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-07T07:52:01Z

script/llm_module/opposition/module.py

+    def forward(self, system_prompt, history):
+        # history 應為 dspy.History 物件
+        result = self.predictor(system_prompt=system_prompt, history=history)
+        return result.response


The opposition module's signature defines response as an OutputField (line 15 of opposition/signature.py) but the module's forward method returns result.response as a string (line 15 of opposition/module.py). However, in chatbot.py line 70, the result is used directly without accessing the .response attribute, which is inconsistent with how the investigator module is used. This suggests the forward method should return the full result object, not just the response string.

Suggested change

return result.response

return result

Copilot · 2026-01-07T07:52:01Z

script/llm_module/investigator/module.py

+    def forward(self, system_prompt, user_query, history):
+        return self.investigate(
+            system_prompt=system_prompt, 
+            user_query=user_query, 
+            history=history
+        )


The investigator module's forward method returns the full result object with .response attribute accessible (lines 13-16), but in chatbot.py line 50, it accesses result.response. This is correct, but inconsistent with the opposition module pattern. For consistency, either both modules should return just the response string, or both should return the full result object.

Copilot · 2026-01-07T07:52:01Z

script/llm_config.py

+    else:
+        print(" 選擇無效，預設用 openai/gpt-5.2")
+        model_name = "openai/gpt-5.2"
+        api_key = os.getenv("OPENAI_API_KEY")
+        lm = dspy.LM(model_name, api_key=api_key, cache=False)


The default model fallback uses "openai/gpt-5.2" which does not exist. If a user provides an invalid choice, the application will fail when trying to use this non-existent model. Use a real, existing model as the fallback, such as "openai/gpt-4o" or "openai/gpt-4o-mini".

Copilot · 2026-01-07T07:52:02Z

use_chatbot_guide.md

+
+## 1. 用到的工具簡介
+
+- uv


Spelling error: "enviorment" should be "environment" in the PR title. This typo should also be corrected.

Copilot · 2026-01-07T07:52:02Z

script/llm_config.py

+        idx = input("選擇模型 (1-5): ").strip()
+        mapping = {
+            "1": "openai/gpt-5.2",
+            "2": "openai/gpt-4o",


The mapping dictionary for OpenAI models is incomplete. It only defines mappings for options "1" and "2", but the prompt offers 5 options (1-5). Options 3, 4, and 5 will fallback to the default "openai/gpt-5.2" without warning. Either complete the mapping or update the prompt to match available options.

Suggested change

"2": "openai/gpt-4o",

"2": "openai/gpt-4o",

"3": "openai/gpt-4o-mini",

"4": "openai/o4-mini",

"5": "openai/o3-mini",

Copilot · 2026-01-07T07:52:06Z

script/llm_module/opposition/signature.py

+    with open(__file__.replace("signature.py", "prompt.md"), "r", encoding="utf-8") as f:
+        return f.read()


Error handling issue: If the prompt.md file cannot be found or read (e.g., incorrect path, missing file, encoding errors), the load_prompt() function will raise an unhandled exception during class definition, which will crash the application at import time. Add proper error handling with a fallback mechanism.

Suggested change

with open(__file__.replace("signature.py", "prompt.md"), "r", encoding="utf-8") as f:

return f.read()

prompt_path = __file__.replace("signature.py", "prompt.md")

try:

with open(prompt_path, "r", encoding="utf-8") as f:

return f.read()

except (OSError, UnicodeError):

# Fallback: use a built-in prompt if the external file is unavailable

return (

"作為 Atlas-World 的反對派或監督者，根據文明憲法與原則，"

"對目前的對話歷史提出挑戰、建設性批評或不同的觀點。"

)

Copilot · 2026-01-07T07:52:06Z

templates/index.html

+  <script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>
+
+  <script>
+    window.MathJax = {
+      tex: { inlineMath: [['$', '$'], ['\\(', '\\)']] },
+      svg: { fontCache: 'global' }
+    };
+  </script>
+  <script id="MathJax-script" async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>


The HTML hardcodes external CDN URLs without integrity checks (SRI - Subresource Integrity). This creates a security vulnerability where if the CDN is compromised, malicious code could be injected. Add integrity attributes with hash values to the script tags for marked.js and MathJax to ensure the loaded scripts haven't been tampered with.

Copilot · 2026-01-07T07:52:06Z

templates/index.html

+          // 1. 強力清除 "粗體內部" 的空格
+          // 將 `** 文字 **` 或 `** 文字**` 或 `**文字 **` 統一轉為 `**文字**`
+          // 原理：找到 ** 開頭，中間包含非星號的內容，結尾是 **，並去掉頭尾的空白
+          content = content.replace(/\*\*\s*([^*]+?)\s*\*\*/g, '**$1**');


The regex pattern on line 249 attempts to remove spaces inside bold markdown, but it's overly aggressive and will incorrectly transform legitimate bold text. For example, "some bold text" would have internal spaces removed, potentially breaking words. This pattern assumes all spaces inside bold markers are unwanted, which is a flawed assumption for natural language text.

Copilot · 2026-01-07T07:52:07Z

script/chatbot.py

+opposer = OppositionModule()
+
+# 使用官方 dspy.History 管理
+chat_history = dspy.History(messages=[])


Global state management issue: The chat_history is a global variable that is shared across all users/sessions. In a multi-user environment, all users would see each other's conversation history, which is a serious privacy and functionality issue. Consider using session-based storage (e.g., Flask sessions or a session manager) to maintain separate histories for each user.

Copilot · 2026-01-07T07:52:07Z

templates/index.html

+      if(!confirm('確定要清空所有對話記憶嗎？')) return;
+      try {
+        const resp = await fetch('/reset', { method: 'POST' });
+        if (!resp.ok) throw new Error(resp.statusText);
+        const data = await resp.json();
+        resDiv.innerHTML = `<span style="color:#27ae60; font-weight:bold;">${data.status || '歷史已重置'}</span>`;
+        resDiv.style.borderLeftColor = '#bdc3c7';
+        setTimeout(() => {
+            if(resDiv.textContent.includes('已重置')) resDiv.innerHTML = '準備就緒。';
+        }, 1500);
+      } catch (err) {
+        alert('重置失敗: ' + err);
+      }
+    });


The confirmation dialog message on line 278 asks in Chinese but doesn't check the return value properly. The function returns early if the user cancels, but then proceeds to send an async request. However, if the reset fails silently on the backend, the UI will show success. Consider adding better error state management.

KageRyo

太狠了

Skywind5487 · 2026-01-07T08:41:16Z

在
一鍵安裝uv+python+各種package.bat
中，要把rem拿掉才有安裝的效果。

3underscoreN

Please rebase to include new commits in #308

heilcheng · 2026-01-07T12:43:00Z

你們真用心

… cache

…ulnerability detection and challenge generation

…port an investigator chatbot

…las-World chatbot

… modules, and add history management. Use dspy to replace whole old ollama logic

…el configuration

Skywind5487 · 2026-01-07T15:10:43Z

Please rebase to include new commits in #308

Done. Thanks for your suggestion.

3underscoreN · 2026-01-08T01:21:36Z

Please rebase to include new commits in #308

Done. Thanks for your suggestion.

太狠了 🛐

Chung1045 · 2026-01-09T03:49:54Z

LGTM, you really take effort in this meme repository lol 😂

WuSandWitch

LGTM

Chung1045 and others added 19 commits January 7, 2026 11:20

Implement local llm feature with ollama

9fd0ddf

Update the local llm approach to use Flask Web Framework

019a943

feat: Add flask template and update the frontend logic

d659d5c

feat: add uv stuff

8680dc5

Update to fix overlooked issues

cbd88dc

Co-authored-by: Kelvin <cychandt@connect.ust.hk>

Update to fix overlooked issue

23dba99

Co-authored-by: Kelvin <cychandt@connect.ust.hk>

feat: update dependencies and add dotenv, Flask, and Flask-CORS

41ca6ea

feat: add .gitignore file to exclude default ignored files and Python…

b356e8d

… cache

feat: update pyproject.toml and add dotenv and flask-cors dependencies

ca138d2

feat: add OppositionModule and OppositionSignature DSPy classes for v…

95ced16

…ulnerability detection and challenge generation

feat: add InvestigatorModule and InvestigatorSignature classes to sup…

f570e25

…port an investigator chatbot

feat: add usage documentation and batch scripts to help launch the At…

b260eca

…las-World chatbot

feat: add llm_config.py for configuring LLM model source selection

986dd06

feat: update LLM initialization to disable caching by default

82720b9

feat: enhance chatbot functionality with investigation and opposition…

9d52c76

… modules, and add history management. Use dspy to replace whole old ollama logic

feat: Add an easy frontend

b1c205a

feat: add settings.json

5b8eea5

fix: disable reloader in app.run to prevent double prompts during mod…

c3c9cc8

…el configuration

feat: disable caching for LLM initialization across all model choices

abd2eb1

Copilot AI review requested due to automatic review settings January 7, 2026 07:38

Copilot started reviewing on behalf of Skywind5487 January 7, 2026 07:39 View session

Copilot AI reviewed Jan 7, 2026

View reviewed changes

KageRyo approved these changes Jan 7, 2026

View reviewed changes

heilcheng approved these changes Jan 7, 2026

View reviewed changes

orenccl approved these changes Jan 7, 2026

View reviewed changes

3underscoreN suggested changes Jan 7, 2026

View reviewed changes

Update 一鍵安裝uv+python+各種package.bat

5428c50

Skywind5487 added 17 commits January 7, 2026 23:10

feat: Add flask template and update the frontend logic

dcb305d

feat: add uv stuff

e8f0005

feat: update dependencies and add dotenv, Flask, and Flask-CORS

9af2ca1

feat: add .gitignore file to exclude default ignored files and Python…

361e8ef

… cache

feat: update pyproject.toml and add dotenv and flask-cors dependencies

51da118

feat: add OppositionModule and OppositionSignature DSPy classes for v…

d2bc4d1

…ulnerability detection and challenge generation

feat: add InvestigatorModule and InvestigatorSignature classes to sup…

c9df197

…port an investigator chatbot

feat: add usage documentation and batch scripts to help launch the At…

4757e5a

…las-World chatbot

feat: add llm_config.py for configuring LLM model source selection

9c9e951

feat: update LLM initialization to disable caching by default

6c7e884

feat: enhance chatbot functionality with investigation and opposition…

c513c18

… modules, and add history management. Use dspy to replace whole old ollama logic

feat: Add an easy frontend

54d2e0f

feat: add settings.json

a7ad552

fix: disable reloader in app.run to prevent double prompts during mod…

7116830

…el configuration

feat: disable caching for LLM initialization across all model choices

44c813c

Update 一鍵安裝uv+python+各種package.bat

3843544

Merge branch 'main' of https://github.com/Skywind5487/Atlas-World

2a6188a

shihyousyu approved these changes Jan 7, 2026

View reviewed changes

3underscoreN approved these changes Jan 8, 2026

View reviewed changes

WuSandWitch approved these changes Jan 9, 2026

View reviewed changes

-            "2": "openai/gpt-4o",
+            "2": "openai/gpt-4o",
+            "3": "openai/gpt-4o-mini",
+            "4": "openai/o4-mini",
+            "5": "openai/o3-mini",

		with open(__file__.replace("signature.py", "prompt.md"), "r", encoding="utf-8") as f:
		return f.read()

-    with open(__file__.replace("signature.py", "prompt.md"), "r", encoding="utf-8") as f:
-        return f.read()
+    prompt_path = __file__.replace("signature.py", "prompt.md")
+    try:
+        with open(prompt_path, "r", encoding="utf-8") as f:
+            return f.read()
+    except (OSError, UnicodeError):
+        # Fallback: use a built-in prompt if the external file is unavailable
+        return (
+            "作為 Atlas-World 的反對派或監督者，根據文明憲法與原則，"
+            "對目前的對話歷史提出挑戰、建設性批評或不同的觀點。"
+        )

Feat: Add uv python enviorment, dspy chatbot, ollama, an easy frontend by flask #359

Are you sure you want to change the base?

Feat: Add uv python enviorment, dspy chatbot, ollama, an easy frontend by flask #359

Uh oh!

Conversation

Skywind5487 commented Jan 7, 2026

Uh oh!

Skywind5487 commented Jan 7, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Key Changes

Reviewed changes

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

KageRyo left a comment

Choose a reason for hiding this comment

Uh oh!

Skywind5487 commented Jan 7, 2026

Uh oh!

3underscoreN left a comment

Choose a reason for hiding this comment

Uh oh!

heilcheng commented Jan 7, 2026

Uh oh!

Skywind5487 commented Jan 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

3underscoreN commented Jan 8, 2026

Uh oh!

Chung1045 commented Jan 9, 2026

Uh oh!

WuSandWitch left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

Skywind5487 commented Jan 7, 2026 •

edited

Loading