🚀 CLI Chatbot Using the OpenAI Responses API

📖 Overview

A command-line chatbot powered by the OpenAI Responses API.
It supports conversation history, file analysis, instant saving, and web search.
Uses event streaming and previous_response_id to maintain context efficiently.

---

✨ Features

• 🗨️ Conversational interface with context-aware replies
• ⌨️ Advanced input options:
  • Multi-line prompt input (Enter = new line, Alt+Enter = send)
  • Emacs-style shortcuts (e.g., Ctrl+A, Ctrl+E, Ctrl+K, etc.)
• 💾 Conversation history saved to ./history
• 📂 File input via lines starting with ~␣ (tilde + space):
  • Text-based files (.txt, .csv, .py, .md, etc.)
  • Excel files (.xlsx, automatically converted to JSON)
  • PDF files (.pdf, supports both local files and URLs)
  • Images (.jpg, .jpeg, .png, supports both local files and URLs)
• 🌐 Web search using the --search flag for up-to-date answers
• ⚡ Instant save using !save command

---

⚙️ Requirements

• Python 3.10+
• Install dependencies:

  pip install -r requirements.txt

Optional (Windows):
Use Windows Terminal for better rendering and fewer artifacts.

---

🔑 Setup

1. Set API Key

   setx OPENAI_API_KEY "your_api_key_here"   # Windows
   export OPENAI_API_KEY="your_api_key_here" # Linux / macOS

2. Configure Settings

   Copy settings_example.py → settings.py and edit:

   PROMPT = "You are a helpful assistant. Respond in well-structured Markdown."
   MODEL = "gpt-5.1-chat-latest"  # gpt-5.1-chat-latest | gpt-5.1 | gpt-5.1-codex | gpt-5.1-codex-mini | gpt-5-chat-latest | gpt-5 | gpt-5-mini | gpt-4.1 | gpt-4.1-mini | o4-mini
   TEMPERATURE = 0.35
   STREAM = False  # True | False
   REASONING_EFFORT = "low"  # low | medium | high | minimal
   COUNTRY = "JP"  # A two-letter ISO country code, like US

3. Run

   python clichatbot.py

   (Windows users can use script/wt-openai.bat to launch in Windows Terminal.)

---

💬 Usage

Basic Chat

Type your question.

• Enter → new line
• Alt+Enter (or Esc then Enter) → send message

user: Your question

File Analysis

Multiple files supported (each file path on a new line starting with ~␣):

user: Explain these files
~ /path/to/example.py
~ /path/to/document.pdf
~ /path/to/image.png

Image / PDF via URL

user: Are these files related?
~ https://example.com/image.jpg
~ https://example.com/document.pdf

Web Search

Add the --search flag at the end of your question. gpt-5-chat-latest is not supported.

user: What are the latest developments in quantum computing? --search

Exit Session

Press Alt+Enter (or Esc then Enter) on empty input → history is saved to ./history.

Instant Save

user: !save

---

⌨️ Keyboard Shortcuts (while editing a prompt, Emacs-style)

• Ctrl + A → Move cursor to beginning of line
• Ctrl + E → Move cursor to end of line
• Ctrl + U → Delete to beginning of line
• Ctrl + K → Delete to end of line
• Ctrl + D → Delete character under cursor
• Ctrl + H → Backspace

---

🖼️ Example

user: Explain these files
~ clichatbot.py
~ settings_example.py
Completed loading the file: 'clichatbot.py'
Completed loading the file: 'settings_example.py'
gpt-5-medium assistant:
Here’s a clear walkthrough of what each file does, how to use the CLI chatbot, and how the pieces fit together.

Overview

- clichatbot.py: A command-line chatbot that talks to OpenAI’s Responses API. It supports:
  - Streaming and non-streaming outputs
...

user: Analyze the letter and provide a summary of the key points.
~ https://www.berkshirehathaway.com/qtrly/2ndqtr25.pdf
Processing the file: 'https://www.berkshirehathaway.com/qtrly/2ndqtr25.pdf'
gpt-5-medium assistant:
Here’s a concise, decision‑oriented summary of Berkshire Hathaway’s Q2 2025 Form 10‑Q.

Headline results
- Net earnings attributable to Berkshire shareholders: $12.37B for Q2 (vs. $30.35B in Q2’24); $16.97B for 1H’25 (vs. $43.05B in 1H’24).
- Revenues: $92.5B in Q2 (−1% YoY); $182.2B in 1H (−1% YoY).
...

---

⚙️ Configuration

• PROMPT → System prompt for the assistant (e.g., "You are a helpful assistant.")
• MODEL → OpenAI model (gpt-5.1, gpt-4.1, o4-mini, etc.)
• TEMPERATURE → Higher = more creative, Lower = more deterministic
• STREAM → Stream output (True/False)
• REASONING_EFFORT → Controls reasoning depth (low, medium, high, minimal)
• COUNTRY → ISO country code (e.g., US, JP) for localized web search

---

📝 Notes

• Reasoning usage:
  • Valid only for reasoning models (gpt-5, o1–o9), excluding gpt-5-chat-latest.
  • minimal → intended for gpt-5 family.
  • For o1–o9, minimal falls back to low.
• STREAM=False → single Markdown-rendered reply per turn.
• If an error occurs with an HTTP-linked PDF, try downloading the file first.
• Surrogate pair characters are not supported due to a limitation in prompt_toolkit.

---

📜 License

MIT