latterworks/ollama-hosts-index

🛰️ Ollama Hosts Index (Latterworks)

A registry of reachable Ollama model endpoints scraped from the wild. This dataset contains 936 IP:PORT entries—mostly on port 11434—with raw /api/tags payloads included for inspection or parsing.

✨ Fields

• ip: Host address, typically of the form IP_PORT (e.g., 45.41.94.28_11434)
• model: Extracted model name, e.g., llama3:8b-instruct-q5_K_M (currently null, see note)
• tags: Optional user-provided tags (mostly empty)
• raw: Full JSON from /api/tags, includes model list and metadata (size, digest, quantization, etc.)

⚠️ Note on model field

The model column is currently null because parsing was deferred—but the model metadata is present inside the raw field. Most raw values are structured as:

{
  "models": [
    {
      "name": "llama3.2:latest",
      "model": "llama3.2:latest",
      "modified_at": "2025-05-17T22:58:10.122Z",
      ...
    }
  ]
}

To extract, parse raw as JSON and access .models[0].model.

📦 Example Use

import pandas as pd
import json

df = pd.read_parquet("data/train-00000-of-00001.parquet")

# Extract IP + first model name
df['parsed_model'] = df['raw'].apply(lambda r: json.loads(r).get("models", [{}])[0].get("model", None))

print(df[['ip', 'parsed_model']].dropna().head())

🔌 Connecting to Hosts

If a host is online, you can ping it with:

curl http://<IP>:11434/api/tags

Example:

curl http://45.41.94.28:11434/api/tags

Returns:

{"models":[{"name":"llama3.2:latest",...}]}

🛠️ Attribution

Collected by [latterworks] using passive scanning and broadcast Ollama introspection.

---