nicobailon/pi-web-access

Web search and content extraction extension for Pi coding agent

What it solves

Pi Web Access provides a comprehensive web-interaction layer for the Pi agent, enabling it to perform web searches, extract content from various sources, and understand video content. It solves the problem of giving an AI agent a reliable, multi-provider search and content retrieval system with smart fallbacks and specialized handling for complex formats like GitHub repositories, YouTube videos, and PDFs.

How it works

The project implements a set of tools (web_search, fetch_content, get_search_content, and source_check) that the Pi agent can use to interact with the web. It supports a vast array of search providers (e.g., OpenAI, Brave, Exa, Tavily, Gemini, Mistral) and content extraction methods.

Key mechanisms include:

  • Smart Fallbacks: A sequential chain of providers is used to ensure search and fetch requests succeed even if one provider fails.
  • Specialized Extraction: Instead of simple scraping, it clones GitHub repositories locally for full file access and uses Gemini for visual and transcript-based video understanding.
  • PDF Processing: It converts PDFs to Markdown using a tiered system (Datalab, Gemini, or local unpdf) to maintain document structure.
  • Caching: Fetched content is stored in a local cache to allow the agent to retrieve specific passages without reloading the entire page into the context window.

Who it’s for

This tool is designed for users of the Pi agent who want to extend its capabilities with real-time web access, deep technical research (via GitHub cloning), and multimodal analysis of video and PDF documents.

Highlights

  • Extensive Provider Support: Integrates with dozens of search APIs and supports zero-config search via Exa MCP.
  • Video Understanding: Extracts transcripts, visual descriptions, and specific frames from YouTube and local videos using Gemini and ffmpeg.
  • get_search_content: Allows the agent to efficiently navigate large pages using offsets and fuzzy text searching.
  • GitHub Integration: Clones repositories locally for genuine file exploration rather than scraping HTML.
  • Source Checking: A dedicated source_check tool for verifying claims with machine-readable citations and SHA-256 hashes.

Related

  • Project
    Amm1rr/WebAI-to-API

    WebAI‑to‑API is a self‑hosted bridge that runs Google Gemini (via web API or a Playwright‑driven browser) and exposes it through OpenAI‑compatible endpoints, with a local dashboard and Docker support.

  • Project
    Sophomoresty/gemini-web2api

    gemini‑web2api is a Python server that proxies Google Gemini’s private web API and presents it as an OpenAI‑compatible endpoint (`/v1/chat/completions`, `/v1/models`, etc.). It supports optional API‑key auth, streaming, function calling, image input, multiple Gemini model aliases, adjustable reasoning depth, and can run directly or inside Docker.

  • Project
    Aas-ee/open-webSearch

    Open‑WebSearch is a Node‑JS service that lets AI agents perform live web searches and fetch article content without any API keys. It offers an MCP‑compatible server, a CLI, and a long‑running HTTP daemon, aggregates results from many public search engines, supports proxy use, and can optionally use Playwright‑driven headless browsers for sites that need JavaScript rendering.

  • Project
    narumiruna/pi-extensions

    A collection of modular extensions and libraries for the Pi Coding Agent that add capabilities for coding, browser automation, and workflow observability.

  • Project
    agegr/pi-web

    A local browser UI for the pi coding agent that allows users to manage conversations, configure models, and inspect project files via a web interface.