Josefalbers / Vimlm: Vimlm is a Vim Plugin that provides an assistant with LLM editing by allowing users to a local LLM command and automatic ingesting code in context.

Coding partner in LLM for Vim, inspired by Github Copilot / cursor. Allocate comprehension code understanding, meeting, and AI help directly to your workflow.

• Model agnostic – Use any MPX model equal by a configuration file
• Vim-native UX – Intuitive keybindings and answers to the window in the split-window
• Deeper context – understands the context of the code from:
  • Current file
  • Visual Choices
  • Files mentioned
  • Structure in project directory
• CONVERSATION – Itative pinution with follow-up questions
• Security gapped in air-gapped – 100% Offline – No apis, no tracking, no data leak

• Apple M-Series Chip
• Python 3.12.8

• Ctrl-l: Add current line + file to context
  Example Prompt: "Regex for removing HTML tags from item.content"

• Choose Code → Ctrl-l: Add selected block + current context file
  Example Prompt: "Convert this to async/await syntax"

• Ctrl-j: Keep current thread
  Example Follow-Up: "Use Manifest V3 instead"

• Ctrl-p: Enter the code blocks from the response to:
  • Last Visual Pelection (Normal Mode)
  • Active Choice (Visual Mode)

Example of Workflow:

1. Choose a block of code in visual mode
2. Promptings Ctrl-l: "Convert this to async/await syntax"
3. Press Ctrl-p To replace the selection with code generated

!include (PATH)  # Add files/folders to context

• !include (no path): current folder
• !include ~/projects/utils.py: Specific file
• !include ~/docs/api-specs/: Full folder
  Example: "AJAX-ify this app !include ~/scrap/hypermedia-applications.summ.md"

!deploy (DEST_DIR)  # Extract code blocks to directory

• !deploy (no path): current directory
• !deploy ./src: Specific directory
  Example: "Create REST API endpoint !deploy ./api"

!continue (MAX_TOKENS)  # Continue stopped response

• !continue: Default 2000 Tokens
• !continue 3000: Custom Token Limit
  Example: "tl;dr !include large-file.txt !continue 5000"

!followup  # Equivalent to Ctrl-j

Example:
Initial: "Create Chrome extension"
Follow-up: "Add dark mode support !followup"

Many orders of many orders in a quick:

"Create HTMX component !include ~/lib/styles.css !deploy ./components !continue 4000"

Vimlm uses a file in JSON config with the following parameters in the Configuring:

{
  "DEBUG": true,
  "LLM_MODEL": null,
  "NUM_TOKEN": 2000,
  "SEP_CMD": "!",
  "USE_LEADER": false
}

1. Browse models: MLX community models to deal with face
2. Edit config file:

{
  "LLM_MODEL": "mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit",
  "NUM_TOKEN": 9999
}

3. Save the: ~/vimlm/config.json
4. Restart Vimlm

If you want to use In the area of For Vimlm key sealing:

Villm is licensed under Apache-2.0 License.