Skip to content
| Marketplace
Sign in
Visual Studio Code>Chat>Farshid Local Folder PromptNew to Visual Studio Code? Get it now.
Farshid Local Folder Prompt

Farshid Local Folder Prompt

Dr. Farshid Pirahansiah

|
2 installs
| (0) | Free
Chat with local Ollama models, with assistants, attachments, web search, and scheduled tasks.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Farshid Local Folder Prompt — VS Code Extension

Chat with local Ollama models directly inside VS Code. Text, images, PDFs and Word docs as attachments. Optional local web search. 100% offline (except for the optional web search).

  • Package name: farshid-local-folder-prompt

Features

  • Assistants: edit/create/delete personas (name, emoji, model, system prompt). Defaults seeded on first launch.
  • Chats per assistant, persisted to extension global storage.
  • Streaming chat responses from Ollama.
  • Attachments: text files (read inline), images (base64 → vision models), PDFs (text via pdf-parse), Word docs (mammoth).
  • Web search via a local SearXNG instance — toggle the 🌐 button.
  • Scheduled tasks: schedule a prompt + attachments to run later. The scheduler runs in the extension host.
  • Drag & drop files onto the message area.

Prerequisites

1. Install Ollama

Download from ollama.com/download, then verify:

ollama --version
ollama serve   # usually started automatically by the desktop app

The extension talks to http://127.0.0.1:11434 by default (override via folderChat.ollamaUrl).

2. Pull recommended models

You need at least one chat model. The two below cover the common cases on a modest laptop:

# Multimodal — assistants that accept image / PDF / DOC / TXT inputs
ollama pull minicpm-v:latest

# Lightweight tool-calling / general chat — fast, good for agents and quick replies
ollama pull qwen3:0.6b

Suggested role mapping when creating assistants in the UI:

Assistant kind Recommended model Why
Vision / multi-input (image, PDF, DOC, TXT) minicpm-v:latest Vision-capable, accepts base64 image attachments
Tool-calling agent / quick chat qwen3:0.6b Small, fast, supports tool calls

You can pull any other Ollama model (llama3.2, qwen2.5:7b, granite4, …) and pick it from the assistant's model dropdown.

3. (Optional) SearXNG for web search

If you want the 🌐 web-search button, run a local SearXNG on http://127.0.0.1:8888 (override via folderChat.searxngUrl). See the SearXNG docs.

Install / develop

npm install
npm run build

Then in VS Code: open this folder and launch the Extension Development Host.

The default VS Code shortcut is F5, but it conflicts with many other tools. This project uses Alt + Ctrl + Shift + F11 (same combo on macOS and Windows) instead.

Add this once to your user keybindings.json (Command Palette → Preferences: Open Keyboard Shortcuts (JSON)):

[
  {
    "key": "alt+ctrl+shift+f11",
    "mac": "alt+ctrl+shift+f11",
    "command": "workbench.action.debug.start",
    "when": "debuggersAvailable && debugState == 'inactive'"
  },
  {
    "key": "alt+ctrl+shift+f11",
    "mac": "alt+ctrl+shift+f11",
    "command": "workbench.action.debug.stop",
    "when": "inDebugMode"
  }
]

After that, Alt + Ctrl + Shift + F11 launches (and stops) the Extension Development Host. Open the Folder Chat view from the activity bar (the comment-discussion icon), or run the command Folder Chat: Open Panel.

Package & install

Build a local .vsix:

npx @vscode/vsce package
# → farshid-local-folder-prompt-0.1.4.vsix

Install it into VS Code:

code --install-extension farshid-local-folder-prompt-0.1.4.vsix

Settings

Setting Default Purpose
folderChat.ollamaUrl http://127.0.0.1:11434 Local Ollama base URL
folderChat.searxngUrl http://127.0.0.1:8888 Local SearXNG base URL
folderChat.maxExtractChars 20000 Truncate per-attachment text

Storage

All extension state (assistants, chats, scheduled tasks, settings, attachments, …) is saved under a single user-level folder:

~/.farshid/local-folder-prompt/
  assistants.json     # personas
  chats.json          # chat history per assistant
  tasks.json          # scheduled tasks
  settings.json       # user-level overrides
  attachments/        # cached extracted files
  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft