Filechat

A file-based AI chat tool designed to integrate with Visual Studio Code (VS Code), enabling you to engage with a Large Language Model (LLM) directly within a Markdown file.

Requirements

Python 3.12.2

https://www.python.org/downloads/release/python-3122/

Python packages

See requirements.txt.

Visual Studio Code

https://code.visualstudio.com/

Usage

Cloning the repository and installing dependencies

1. Clone the repository:

   git clone https://github.com/noveky/filechat.git

2. Ensure that Python and pip are installed on your system.

3. Create a Python virtual environment in the repository. (Optional)

4. Install project dependencies:

   pip install -r requirements.txt

Building a workflow in VS Code

1. Open the repository with VS Code.

2. Install the Code Runner extension (formulahendry.code-runner) for VS Code.

3. Specify the run code command for Markdown files in .vscode/settings.json.

   • For Windows:

     // .vscode/settings.json
     {
         "code-runner.runInTerminal": true,
         "code-runner.executorMap": {
             "markdown": "cd $dir && ..\\run.ps1 $fullFileName"
         }
     }

   • For macOS/Linux:

     // .vscode/settings.json
     {
         "code-runner.runInTerminal": true,
         "code-runner.executorMap": {
             "markdown": "cd $dir && bash ../run.sh $fullFileName"
         }
     }

4. Install the Markdown All in One extension (yzhang.markdown-all-in-one).

   This extension enables you to open a preview panel for a Markdown file (default keyboard shortcut is Ctrl + K V), where you can read the rendered Markdown content more easily.

Configuring the app

Specifying OpenAI API key and base URL

To access a chat model, you need to specify the OpenAI API key and base URL.

There are three ways to do that:

1. Set system environment variables OPENAI_API_KEY and OPENAI_BASE_URL .

2. Create a .env file in the root directory of the repository and set the two environment variables in the file.

   OPENAI_API_KEY=<your_api_key>
   OPENAI_BASE_URL=<your_base_url>

3. Specify the API key and the base URL in the app configuration file introduced below.

Customizing app configurations

You can customize the app configurations in config.yaml.

Here is a template, and you can choose what you want to specify:

# ---- Chat Completion ----
model: # Name of your preferred completion model
temperature: # Your preferred `temperature` parameter for the model (optional)
max_tokens: # Your preferred `max_tokens` parameter for the model (optional)

# ---- App Behaviors ----
max_retries: # How many times to retry (default is 3)
print_response: # Whether to print the response in standard output (default is true)
stream_for_file: # Whether to append the response to the file token by token or as a whole (default is true)

# ---- OpenAI API ----
api_key: # OpenAI API key (overrides the environment variable `OPENAI_API_KEY` if specified)
base_url: # OpenAI base URL (overrides the environment variable `OPENAI_BASE_URL` if specified)

Chatting with an LLM in a file

Here is a quick guide to get yourself started:

1. Make a chats directory in the repository and create a Markdown file in it (e.g. New Chat.md).

2. Type a message in the Markdown file.

   Tell me a joke.

3. Save the file, and run Filechat on the current file by hitting "Run Code" (default keyboard shortcut is Ctrl + Alt + N).

   Non-structured file content is regarded as a user message, so the file gets formatted into a # User heading followed by the original content.

   After a few seconds, hopefully you will see the LLM response get appended to the end of the file. The new # User heading prompts you continue the conversation.

   # User
   
   Tell me a joke.
   
   # Assistant
   
   Why don't skeletons fight each other?
   
   They don't have the guts!
   
   # User
   
   |

4. Type your reply message and follow step 3 to continue chatting.

Adding a system prompt

Besides user messages, you can optionally include a system prompt, labeled with a # System heading. This is a special section that sets the context or provides initial instructions for the model before processing user inputs.

# System

This is a system prompt.

# User

This is a user message.

Adding file-wise configurations

At the very beginning of your chat file, you can include a front matter section. This section is enclosed within triple dashes (---) and is written in YAML format. It allows you to specify file-wise configurations that override the app configurations.

---
model: gpt-4o
temperature: 0.7
---