⚗️ Distil

Weekly distil generator for drug discovery and AI research. Aggregates content from RSS feeds and YouTube transcripts, filters by relevance, and generates executive summaries using local or cloud LLMs.

Quick Start

Requirements: Python 3.8+ (usually pre-installed)

Step 1: Install uv (Python package manager)

Windows:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

macOS/Linux:

curl -LsSf https://astral.sh/uv/install.sh | sh

Note: Restart your terminal after installation to update your PATH.

Step 2: Run distil (auto-installs Ollama and models)

git clone https://github.com/ai-mindset/distil.git && cd distil
uv run distil run

Distil automatically installs Ollama and downloads models as needed.

Prerequisites

• Python 3.11+ (usually pre-installed)
• Internet connection for initial setup
• Everything else (uv, Ollama, models) is automatically installed

Installation

From source (development):

git clone https://github.com/ai-mindset/distil.git
cd distil
uv run distil run  # Automatically sets up everything on first run

From package (when published):

uv tool install distil
distil run  # Automatically sets up Ollama and models

Configuration

Copy config.toml to your working directory and edit:

[llm]
model = "ollama/mistral:latest"  # Local (free) — or "anthropic/claude-sonnet-4-20250514" (requires API key)

[domain]
focus = "drug discovery, pharmacology, AI/ML for therapeutics"

[[feeds]]
url = "https://rss.arxiv.org/rss/cs.ai"
keywords = ["drug", "molecule", "protein", "binding"]  # Only items matching these
max_items = 20

For cloud LLMs, set your API key:

macOS/Linux:

export ANTHROPIC_API_KEY="sk-..."

Windows (PowerShell):

$env:ANTHROPIC_API_KEY="sk-..."

Windows (Command Prompt):

set ANTHROPIC_API_KEY=sk-...

Usage

Web UI (recommended):

uv run distil serve  # Auto-installs dependencies if needed

Opens browser at http://localhost:5001. Click "Fetch Items" to preview, then "Generate Distil".

CLI:

uv run distil run              # Generate distil with defaults (auto-installs Ollama if needed)
uv run distil run --days 3     # Last 3 days only

Output

Distils are saved to history/distil-YYYY-MM-DD_HHMM.md in your current directory. View past distils at http://localhost:5001/history.

Adding Sources

Edit config.toml to add feeds:

[[feeds]]
url = "https://example.com/rss"
name = "My Feed"
keywords = ["relevant", "terms"]  # Optional: filter by keywords
max_items = 25                     # Optional: limit items

Features

• ⚡ Zero-Config Setup: Automatically installs Ollama and downloads models on first run
• 🌍 Cross-Platform: Works on Linux, macOS, and Windows
• 🌙 Dark Mode: Toggle between light/dark themes in the web UI
• 📊 Feed Health Monitoring: Real-time status of each RSS feed
• 🔄 Batch Processing: Handles large content volumes without timeout
• 📈 Streaming Progress: Live updates during distil generation
• 📁 History Management: View and manage past distils

Troubleshooting

Issue	Solution
"Connection refused" from Ollama	Distil will automatically start Ollama server; if issues persist, check system logs
Slow generation	Reduce max_items per feed, or use fewer feeds
Missing items	Check keywords aren't too restrictive
Web app stuck at "Fetching..."	Check feed URLs are accessible; see feed health report
Timeout errors	System now uses batch processing to prevent this
Windows installation	Distil will prompt you to download Ollama manually from https://ollama.com/download