Source Code

Eleutherios - Epistemic Analysis Infrastructure

Query knowledge graphs built from your document collections. Detect suppression patterns, coordination signatures, and get multi-perspective analysis on contested topics.

What This Skill Does

Eleutherios transforms document collections into knowledge graphs with claim-level extraction, then runs detection algorithms to surface patterns that traditional search misses:

Suppression Detection: Identifies funding cuts, career impacts, publication obstacles, and institutional marginalization patterns documented within sources
Coordination Signatures: Detects timing patterns, shared language, and citation network anomalies suggesting coordinated messaging
Multi-Perspective Clustering: Groups claims by viewpoint so you can see all sides of contested topics
Source Topology Analysis: Maps citation networks and trust relationships between sources

When to Use This Skill

Use Eleutherios when you need to:

Research topics where institutional consensus may be manufactured
Analyze historical documents for suppression patterns (e.g., declassified materials, congressional testimony)
Compare how different sources treat the same topic
Build understanding of contested scientific or historical debates
Investigate citation voids and research threads that mysteriously dead-end

Example prompts:

"What does Eleutherios show about suppression patterns for Thomas Paine?"
"Get perspectives on plasma propulsion research from my knowledge graph"
"Analyze the topic of electrogravitics - what sources exist and what patterns emerge?"
"Assess the source topology for the Smedley Butler FBI files"

Prerequisites

Eleutherios must be running locally before using this skill.

Quick Start (Docker)

# Clone the repository
git clone https://github.com/Eleutherios-project/Eleutherios-docker.git
cd Eleutherios-docker

# Start the stack
docker-compose up -d

# Verify MCP server is running
curl http://localhost:8100/health

The MCP server runs on port 8100 by default. The web UI is available at http://localhost:8080.

Importing Documents

Use the web UI or CLI to import your document collection:

# Via CLI
docker exec -it eleutherios-api python aegis_import_wizard.py /path/to/documents

# Or use the web UI at http://localhost:8080

Supported formats: PDF, DOCX, TXT, MD, HTML

Verify Connection

Once Eleutherios is running, test the MCP connection:

curl -X POST http://localhost:8100/mcp/list_domains \
  -H "Content-Type: application/json" \
  -d '{}'

You should see a JSON response with your corpus statistics.

Available Tools

analyze_topic

Run suppression and coordination detection on a topic.

Parameters:
  - topic (required): The topic to analyze
  - detail: "brief" | "standard" | "verbose" (default: "standard")
  - max_claims: Maximum claims to analyze (default: 100)

Returns:
  - Suppression score (0.0-1.0) with severity rating
  - Coordination score with pattern indicators
  - Relevant claims with source attribution
  - Detection signals (funding impacts, credential attacks, etc.)

get_perspectives

Cluster claims by viewpoint using semantic analysis.

Parameters:
  - topic (required): The topic to analyze
  - max_clusters: Number of perspective clusters (default: 5)
  - claims_per_cluster: Claims per cluster (default: 5)

Returns:
  - Clustered perspectives with representative claims
  - Cluster labels and themes
  - Cross-cluster relationships

assess_source

Analyze a specific source's position in the knowledge topology.

Parameters:
  - source_identifier (required): Document title or source file name
  - detail: "brief" | "standard" | "verbose" (default: "standard")

Returns:
  - Citation network position
  - Claim types (PRIMARY vs META)
  - Trust score aggregation
  - Related sources

get_claim_context

Deep dive on a specific claim with full context.

Parameters:
  - claim_id (required): The claim identifier
  - include_graph: Include graph relationships (default: true)

Returns:
  - Full claim text and metadata
  - Source document context
  - Related entities
  - Citation relationships

list_domains

Get corpus statistics and available domains.

Parameters: none

Returns:
  - Total claims, documents, entities
  - Domain breakdown
  - Corpus health metrics

Example Workflow

You: "What suppression patterns exist for Thomas Paine in my knowledge graph?"

[Skill calls analyze_topic with topic="Thomas Paine"]

Response: "Thomas Paine shows a CRITICAL suppression score of 0.831. 
Key indicators:
- Credential inversion: Despite being a Founding Father, systematically 
  marginalized in educational curricula
- Evidence avoidance: Critics attack character rather than engage arguments
- Career/social impacts: Documented ostracism, denied burial in consecrated ground

47 claims analyzed across 3 documents. Strongest signals from 
'Age of Reason' and 'Rights of Man' analysis."

You: "Show me the different perspectives on his work"

[Skill calls get_perspectives with topic="Thomas Paine"]

Response: "Found 4 distinct perspective clusters:
1. Revolutionary Hero (12 claims) - Focus on Common Sense, independence
2. Religious Critic (9 claims) - Age of Reason, deism arguments  
3. Institutional Dismissal (8 claims) - Character attacks, marginalization
4. Modern Rehabilitation (6 claims) - Academic reassessment

Cluster 3 shows classic suppression rhetoric: ad hominem attacks
without substantive engagement with Cluster 2's actual arguments."

Configuration

Custom MCP Port

If running Eleutherios on a different port, update your OpenClaw config:

# In your OpenClaw workspace skills config
eleutherios:
  mcp:
    server: "http://localhost:YOUR_PORT/mcp"

Remote Server

For remote Eleutherios instances (e.g., on a homelab server):

eleutherios:
  mcp:
    server: "http://192.168.1.100:8100/mcp"

Security Note: Only expose Eleutherios on trusted networks. The MCP server has no authentication by default.

Troubleshooting

"Connection refused" errors

Verify Eleutherios is running: docker ps | grep eleutherios
Check the MCP port: curl http://localhost:8100/health
Ensure no firewall blocking port 8100

Empty results

Verify documents have been imported: check web UI at localhost:8080
Run extraction if needed: documents must be processed before querying

Slow responses

Large corpus queries may take 10-30 seconds
Use max_claims parameter to limit scope
Consider running on hardware with GPU for faster embedding generation

Privacy & Security

Local-first: All data stays on your machine. No cloud dependencies.
No telemetry: Eleutherios sends no data externally.
Your documents, your analysis: Build knowledge graphs from your own curated collections.

About

Eleutherios (from Zeus Eleutherios, god of freedom) is open-source epistemic defense infrastructure. Built for researchers investigating topics where institutional gatekeepers cannot be trusted.

Created by Cedrus Strategic LLC. MIT Licensed.

eleutherios-openclaw-skill