Skip to content

CHAT_COMMAND.md

Latest commit

 

History

History
171 lines (118 loc) · 5.01 KB

CHAT_COMMAND.md

File metadata and controls

171 lines (118 loc) · 5.01 KB

Chat Command Guide

Overview

The chat:: command is a conversational AI interface that lets you interact naturally with the AI agent. Unlike generate:: which is optimized for file-based code generation, chat:: is perfect for:

  • Multi-line conversations
  • Pasting code examples and asking for improvements
  • Asking questions about code patterns
  • Iterative discussions about implementation

Basic Usage

blink> chat:: <your initial request>

When you run the command, the AI will:

  1. Show [PROCESSING] message
  2. Wait for you to type your full message or paste code
  3. Accept multi-line input
  4. Show end to finish (type just end on a new line)
  5. Display the AI's response
  6. Offer to save the code if desired

Examples

Example 1: Create a Temperature Sensor Simulation

blink> chat:: Create a simulation of how temperature sensor works in python
[INFO] Type your message or paste code (type 'end' on new line when done):

[Paste Python code or additional instructions]

end

Output:

[PROCESSING] Analyzing your request...

─────────────────────────────────────────────────────────
[Generated code or explanation]
─────────────────────────────────────────────────────────

[SAVE?] Save generated code to file? (y/n):

Example 2: Improve Existing Code

blink> chat:: Improve this function

[Paste the code you want improved]

end

Example 3: Ask About Code Patterns

blink> chat:: How would you handle error handling in this REST API?

[Paste your API code]

end

How It Works Internally

  1. Input Collection

    • Initial request from command line
    • Multi-line input follows (until you type end)
    • All text is combined into a single message

  2. Processing

    • Message sent to Claude with conversational context
    • AI understands this is a chat interaction (vs file-based operations)
    • Full context from your workspace is available

  3. Output

    • AI response displayed with visual separators
    • Option to save code to a file
    • Conversation logged to session history

  4. Memory

    • Entire chat logged to workspace/.agent_history/current_session.json
    • Accessible via history:: command
    • Persists across sessions

Command Dispatch

In src/simplified_cli.py, the chat command is handled by:

elif command == "chat":
    if not args:
        print("[ERROR] Usage: chat:: <request>\n")
        continue
    self.handle_chat_command(args)

The handle_chat_command() method:

  • Accepts initial request
  • Collects multi-line input
  • Calls self.agent.client.generate() with chat context
  • Handles file saving
  • Logs to conversation memory

Difference from generate::

Feature generate:: chat::
Input Single-line instruction Multi-line, can paste code
Focus File operations Conversation
Context File-based Free-form
Save Option Built-in with prompts Available after generation
Use Case Quick code generation Discussion & iteration
Memory Logged per command Full conversation logged

Tips

  1. Multi-line Pasting: Paste your entire code block, then type end on a new line
  2. File Context: You can mention files from your workspace in the chat
  3. Save Generated Code: When asked, save generated code to easily add it to your project
  4. Conversation Flow: Each chat is logged separately in your session history
  5. No File Required: Unlike generate::, you don't need to reference specific files

Files Modified for This Feature

  • src/simplified_cli.py - Added handle_chat_command() method and command dispatch
  • README.md - Updated with chat command documentation
  • src/simplified_cli.py - Updated print_help() with chat examples

Testing the Command

# Activate your venv
source venv/bin/activate  # or venv\Scripts\activate on Windows

# Run Blink
python main.py

# Try the chat command
blink> chat:: Create a simple Python class for managing user profiles
[Type more instructions or paste code]
end

# Should see AI response and save option

Troubleshooting

Q: Chat seems to hang after I type the initial command? A: The command waits for you to type your full message. Either type more lines or type end to finish.

Q: My code wasn't analyzed properly? A: Make sure to paste the complete code block and type end on a new line after.

Q: Where is my chat history saved? A: In workspace/.agent_history/current_session.json. View it with history:: command.

Q: Can I chat with code examples from external files? A: Yes! You can paste code from external files, or mention file paths in your chat request.