Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    pdf rag graph

    1

    High-precision RAG skill for searching PDF collections with page-level citations and knowledge graph traversal.

    Free

    Security scanned
    pdf rag graph

    pdf rag graph

    Example session with this skill installed

    Search the QED thesis for the definition of the Ward identity and give me the page number.

    • Read your context and instructions
    • Compiled the pdf rag graph

    I have retrieved the following evidence from the document index regarding the Ward identity.

    Search Results: Ward Identity

    DocumentPage(s)SectionSnippet
    qed_thesis.pdf18, 19INTRODUCTION > 5. WARD-TAKAHASHI IDENTITY IN QED"This identity implies that there exists a relation between the fermion two point function (self-energy) and the vertex function in QED. Specifically, the renormalization constants Z1 and Z2 must be equal..."
    qed_thesis.pdf22RENORMALIZATION > 2. GAUGE INVARIANCE"The Ward-Takahashi identity is a consequence of gauge symmetry in quantum electrodynamics, ensuring that the longitudinal part of the vertex function vanishes."

    The Ward identity (specifically the Ward-Takahashi identity) defines a relationship between the fermion self-energy and the vertex function, implying that the renormalization constants $Z_1$ and $Z_2$ are equal (qed_thesis.pdf, p. 18).

    Next steps

    1. Run graph_walk if you need to see how this identity relates to other renormalization constants.
    2. Use search with mode: "bm25" to find specific mentions of $Z_1$ or $Z_2$ values.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Git Hub link: https://github.com/gopalsharma1615-ai/rag-mcp-server

    Finding specific facts in a mountain of PDFs is slow and prone to hallucination. This skill provides a high-precision retrieval system that extracts evidence with page-level citations from your local documents.

    What it does

    • Hybrid search combines semantic dense retrieval with BM25 keyword matching for finding exact terms and codes.
    • Knowledge graph walking resolves multi-hop queries that require chaining facts across different documents.
    • Structural ingestion preserves document hierarchies, headings, and table headers during the indexing process.
    • Reranking support utilizes a cross-encoder to verify the relevance of retrieved passages before answering.
    • Entity exploration allows inspecting specific terms, their relationships, and source document clusters.

    How it works

    1. Ingest your PDF collection using layout-rag to create a local vector index of text and tables.
    2. Build an optional knowledge graph using an LLM to map relationships for complex, relational questions.
    3. Query the system via the search, graph_walk, or guided_walk tools to retrieve raw evidence chunks.
    4. Cite results automatically with document names, page numbers, and section paths provided in the tool output.

    Frameworks & tools

    Built on Python 3.10+ using rag-mcp-server, layout-rag, and graphwalk-rag. It utilizes Hugging Face models for local embedding and optionally integrates with OpenAI-compatible APIs for graph construction.

    Why this beats prompting it yourself

    Standard RAG often loses context in tables or fails when answers span multiple documents. This skill uses specialized layout-aware chunking and graph-traversal logic to maintain document structure and follow relational links that simple vector search misses.

    Use cases

    • Audit complex legal contracts or insurance policies for conflicting clauses.
    • Extract specific data points from technical manuals or scientific papers with page citations.
    • Perform due diligence across a large folder of corporate reports and filings.
    • Build a structured knowledge base from internal documentation for team support.

    Known limitations

    Optimized for English by default. It does not support image or chart analysis. Knowledge graph quality depends on the LLM used during the build phase.

    Optimized for English-language PDFs. No support for interpreting images, charts, or diagrams. Requires Python 3.10+ and Tesseract for OCR tasks.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 6 days ago

    • Free with an Agensi account

    Requires

    Rag MCP
    Python

    Permissions

    Terminal / Shell
    Read Files
    Network Access
    Environment Variables
    Listed6 days ago

    What's inside

    Frequently Asked Questions

    Standard RAG relies on simple text proximity, whereas this skill uses 'layout-aware' chunking to preserve table structures and a knowledge graph to connect related facts across different documents, drastically reducing hallucinations.

    This skill is designed for any platform supporting Python 3.10+ and the Model Context Protocol (MCP), making it compatible with major AI agents and orchestration frameworks that utilize tool-calling.

    No, your documents are indexed locally; while you can optionally use external APIs for graph construction, the core vector index and retrieval logic run on your local infrastructure to ensure data privacy.

    You receive the complete Python-based skill package, including the 'layout-rag' and 'graphwalk-rag' modules, configuration templates for indexing, and the tool definitions for search and graph traversal.

    Yes, the system is designed to provide specific document names, page numbers, and section headers for every piece of evidence it retrieves, allowing for instant human verification.

    The skill currently supports text and tables within PDFs, but it does not perform OCR on images or analyze visual charts and diagrams at this time.

    New in LLM Ops