BigContext

Provides a server that segments large documents and enables TF-IDF search for efficient retrieval.

python

0

GitHub Stars

python

Language

4 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "rixmerz-bigcontext_mcp": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/path/to/bigcontext-mcp",
        "bigcontext-mcp"
      ]
    }
  }
}

BigContext MCP is a server that helps you work with very large documents by automatically segmenting content and performing TF-IDF searches to retrieve only the most relevant fragments for each query. This keeps you from saturating model context windows while you explore books, manuals, papers, and other extensive texts.

How to use

You use BigContext MCP by running it through an MCP client that supports standard MCP endpoints. Start the server with the included runtime command, then connect your client to the local or remote endpoint to ingest documents, search segments, and validate extraction results. The server exposes tooling for ingestion, semantic search, metadata access, and numerous domain-agnostic validators to help you work with large documents efficiently.

How to install

Prerequisites you need before you begin:

Python 3.11+ must be installed
uvx runtime must be installed to run MCP as a stdio-based server
A compatible MCP client to interact with this server (for example, a client that supports standard MCP protocols)

Step-by-step commands to run BigContext MCP via the recommended runtime (stdio):

uvx --from git+https://github.com/Rixmerz/bigcontext_mcp.git bigcontext-mcp

Configuration and operation notes

Configure your MCP client to connect to the stdio server you launch with the command above. The server uses a dynamic vocabulary provided at runtime, so you do not need to hardcode domain terms in advance.

Usage patterns you can perform

Ingest documents: load a document, let the server segment it, and index the content for fast retrieval.

Search: query for specific ideas or phrases and receive the most relevant segments with associated metadata and context.

Validate extractions: apply the available validators to ensure quoted material and data extraction stay grounded in the source content.

Security and best practices

Treat the domain vocabulary you provide as the primary source of truth for semantic interpretation. Do not rely on hardcoded terms. Use the validators to ensure compliance with extraction rules and avoid pattern contamination.

Examples and workflows

Workflow example: ingest a large PDF, search for a specific concept using TF-IDF ranking, then pull the exact passage that matches the claim and validate its provenance.

Troubleshooting

If the MCP server fails to start, verify you can access the Git URL from your environment and that uvx is installed correctly. Check for Python 3.11+ compatibility and that the runtime command you use is exactly as shown in the installation steps.

Notes

This MCP focuses on domain-agnostic extraction validators and relies on dynamic vocabulary provided at runtime to tailor analyses to the content you work with.

Available tools

ingest_document

Load, segment, and index a document.

search_segment

Search for relevant segments using TF-IDF.

get_metadata

Retrieve metadata, structure, and top terms.

list_documents

List all indexed documents.

compare_segments

Compare two segments for themes and similarity.

get_source_capabilities

Analyze what a document can and cannot support.

validate_claim

Check if a claim can be grounded in the source.

get_epistemological_report

Complete analysis before scholarly claims.

check_language_operation

Validate linguistic operations.

detect_semantic_frames

Identify conceptual frameworks like causal or performative frames.

analyze_subdetermination

Distinguish indeterminacy from subdeterminacy.

detect_performatives

Identify performative speech acts.

check_anachronisms

Detect imported post-biblical concepts.

audit_cognitive_operations

Validate query and output compliance.

detect_inference_violations

Scan for unauthorized connectors in reasoning.

get_permitted_operations

Get allowed operations per text type.

generate_safe_fallback

Produce a compliant response when violations are detected.

validate_literal_quote

Verify quoted text exists exactly in the source.

validate_proximity

Check if segments are adjacent within proximity constraints.

get_adjacent_segments

Retrieve segments within a defined proximity.

identify_speaker

Detect who is speaking within a segment.

detect_pattern_contamination

Find pattern completions not present in the source.

validate_extraction_schema

Validate that data extractions follow a strict schema.

detect_narrative_voice

Distinguish voice types in text.

validate_agency_execution

Differentiate executed versus referenced actions.

detect_text_genre

Identify the genre based on document structure.

detect_divine_agency_without_speech

Find actions described without explicit speech verbs.

detect_weak_quantifiers

Detect unsupported generalizations.

validate_existential_response

Validate yes/no responses against the source.

build_document_vocabulary

Create a closed vocabulary from the document.

validate_output_vocabulary

Ensure output uses only source vocabulary.