MnrGreg/hermes-venice-web
Hermes plugin that uses Venice.AI web_search and web_scrape tools
This plugin serves as a privacy-focused replacement for the built-in web search and extraction tools in Hermes Agent. It integrates Venice AI's /augment/search and /augment/scrape endpoints to provide the agent with explicit control over when to fetch web data. Because it uses standalone endpoints, these tools can be paired with any AI provider, including local models or other cloud APIs. The implementation ensures zero data retention and provides structured, LLM-friendly output.
- Privacy-first search and scraping with zero data retention
- Compatible with any LLM provider via standalone endpoints
- Low-cost pricing at $0.01 per request
full readme from github
Hermes Venice Web Plugin
Full replacement for Hermes Agent's built-in web_search / web_extract tools using Venice AI's official privacy-first API.
This plugin adds two high-quality tools:
venice_web_search→ powered by Venice/augment/searchvenice_web_scrape→ powered by Venice/augment/scrape
✨ Benefits of Venice API
- Zero Data Retention (ZDR) with Brave backend — queries are never stored or logged.
- Fully anonymized Google fallback — Venice never links your identity to requests.
- Strong privacy: No data storage, no training use, no logs.
- Very low cost: Only $0.01 per request for both search and scrape.
- Clean, LLM-friendly output — structured results and Markdown content.
- Fast and reliable performance.
- Provider-independent — uses Venice's standalone
/augmentendpoints, so you can pair Venice web tools with any AI provider (OpenAI, Anthropic, local models, etc.), not just Venice inference.
Unlike Venice's built-in "always-on" web integrations via model feature suffixes (e.g. default:enable_web_search=on&enable_web_scraping=true) or venice_parameters in the request body, this plugin gives the agent full explicit control — it only calls the tools when it actually needs them.
| Approach | How it works | When to use |
|---|---|---|
| This plugin (recommended) | Agent decides per-turn whether to search/scrape | Agentic workflows where the model should judge when web data is needed; works with any LLM provider |
| Model feature suffix | Append flags to the model ID (e.g. default:enable_web_search=on) |
Clients that don't allow modifying the request body; always-on search with Venice inference |
venice_parameters |
Set enable_web_search / enable_web_scraping in the request body |
Direct Venice API calls where you want every request to include web context |
Installation
One-command (recommended)
hermes plugins install mnrGreg/hermes-venice-web --enable
Manual
mkdir -p ~/.hermes/plugins/venice-web
cd ~/.hermes/plugins/venice-web
# Copy the files below into this folder
hermes plugins reload
Setup
- Get your Venice API key → https://venice.ai
- Add it to Hermes:
echo 'VENICE_API_KEY=vk-XXXXXXXXXXXXXXXX' >> ~/.hermes/.env
Remove Default Hermes Web Tools (Important!)
Edit ~/.hermes/config.yaml:
toolsets:
- venice_web # ← Keep this
# - web # ← Comment out or remove
# - browse # ← Comment out or remove
Then restart Hermes or run hermes plugins reload.
Pricing
Per the Venice API pricing:
| Feature | Config | Price |
|---|---|---|
| Web Search | enable_web_search: true |
$10.00 per 1K requests ($0.01/request) |
| Web Scraping | enable_web_scraping: true |
$10.00 per 1K URLs ($0.01/URL) |
These charges apply in addition to standard model token pricing. See full pricing details for text, image, audio, and video models.
API References
- Search: https://docs.venice.ai/api-reference/endpoint/augment/search
- Scrape: https://docs.venice.ai/api-reference/endpoint/augment/scrape
- Pricing: https://docs.venice.ai/overview/pricing#additional-features
- Model Feature Suffix: https://docs.venice.ai/api-reference/endpoint/chat/model_feature_suffix
- Chat Completions (
venice_parameters): https://docs.venice.ai/api-reference/endpoint/chat/completions
Made for the Hermes community by mnrGreg Star this repo if it helps you!