PromptCtl Quick Capture

Version: 1.0.0
Status: Production Ready

1. Abstract

PromptCtl Quick Capture is a browser extension for capturing, organizing, and optimizing AI prompts. It connects to a local backend server that stores prompts and optionally optimizes them using DSPy.

2. Architecture

┌─────────────────┐     HTTP/JSON      ┌─────────────────┐
│  Browser Ext.   │ ◄──────────────►   │  promptctl.py   │
│  (Chrome/FF)    │    localhost:8741  │  FastAPI Server │
└─────────────────┘                    └────────┬────────┘
                                                │
                                       ┌────────▼────────┐
                                       │   Local Ollama  │
                                       │   (Optional)    │
                                       └─────────────────┘

Components:

• extension/ — Browser extension (Manifest V3)
• promptctl.py — FastAPI backend server
• core/ — Backend modules (storage, DSPy optimization)

3. User Interface Reference

3.1 Capture Tab

Header:

• Application title with connection status indicator (green = connected)
• Status shows active features: "Connected + Pipeline + DSPy"

Prompt Text:

• Large text area for entering or pasting prompt text
• Supports selection capture via keyboard shortcut (Ctrl+Shift+S / Cmd+Shift+S)

3.2 Capture Form (Intent & Goals)

Intent & Goals Section:

• Analyze button — Auto-detect prompt type and metadata
• Type — Prompt category (auto-detect or manual selection)
• Target Audience — Intended users (developers, students, general)
• Desired Outcome — Expected result from the prompt
• Constraints — Limitations (word count, tone, format)

Metadata:

• Name — Optional identifier for the prompt
• Tags — Comma-separated labels for organization
• Link to previous prompt — Chain prompts into conversations

Options:

• Auto-tag from domain — Extract tags from current website
• Auto-optimize with DSPy — Enable automatic optimization

Actions:

• Save — Store prompt without optimization
• Save & Optimize — Store and queue for DSPy optimization

3.3 Prompts Tab

Search:

• Filter by content, name, tag, or hash
• Refresh button to reload from server

Prompt List:

• Each prompt displays UUID and preview text
• Click to view/edit prompt details
• Supports multilingual content (Persian, English, etc.)

3.4 Jobs Tab

Background Jobs:

• Lists active DSPy optimization jobs
• Refresh button to check job status
• Shows "No jobs" when queue is empty

3.5 Settings Tab

AI Provider:

• Provider — Select backend (Local Ollama, OpenAI, etc.)
• Ollama URL — Endpoint for local Ollama instance
• Model — LLM model name (e.g., phi3.5:latest)

Optimizer:

• Enable DSPy Optimization — Toggle DSPy pipeline
• Optimization Rounds — Number of optimization iterations
  • 1 round (fast)
  • 3 rounds (balanced)
  • 5 rounds (thorough)

Actions:

• Save Settings — Persist configuration
• Test Connection — Verify backend connectivity

4. Installation

4.1 Docker (Recommended)

git clone https://github.com/OmidZamani/promptctl.git
cd promptctl
docker-compose up -d

Server runs on http://localhost:8741. Done.

4.2 Manual (Alternative)

git clone https://github.com/OmidZamani/promptctl.git
cd promptctl
pip install -r requirements.txt
python promptctl.py

4.3 Browser Extension

Chrome:

1. Navigate to chrome://extensions/
2. Enable "Developer mode"
3. Click "Load unpacked"
4. Select the extension/ directory

Firefox:

1. Navigate to about:debugging#/runtime/this-firefox
2. Click "Load Temporary Add-on"
3. Select extension/manifest.json

5. Configuration

Variable	Default	Description
PROMPTCTL_HOST	0.0.0.0	Server bind address
PROMPTCTL_PORT	8741	Server port
OLLAMA_URL	http://localhost:11434	Ollama API endpoint

Prompts stored in ~/.promptctl/prompts/ as JSON files.

6. Keyboard Shortcuts

Shortcut	Action
Ctrl+Shift+S (Win/Linux)	Save selected text as prompt
Cmd+Shift+S (Mac)	Save selected text as prompt

7. Security Considerations

1. Local-only by default — Server binds to localhost
2. No telemetry — No data sent to external services
3. API tokens — Store in environment variables, never in code
4. Ollama — Runs locally; no cloud dependency

8. License

MIT License