Verifying authentication...
Back

Users

Username Email Role Status Last Login Actions
Loading users...

Feedback Overview

0
Total Feedback
0
Positive (šŸ‘)
0
Negative (šŸ‘Ž)
0
Pending Review
0
In Training Set

Feedback Entries

Rating Question Explanation User Status Date Actions
Loading feedback...

Fine-tuning Guide

How feedback improves the model:

  • Positive feedback (šŸ‘) reinforces good responses. The model learns to generate similar answers for similar questions.
  • Negative feedback (šŸ‘Ž) with explanations helps identify errors. The correction reason teaches the model what NOT to do.

Best practices for feedback review:

  • Review negative feedback to ensure corrections are accurate before including in training
  • Exclude feedback that contains sensitive information or is factually incorrect
  • Use admin notes to document why certain feedback was excluded

Export formats:

  • JSONL - One JSON object per line, ideal for most fine-tuning frameworks
  • JSON - Array format, useful for data analysis and processing
ChromaDB Ollama --:--
-- files (-- chunks)

Knowledge Graph ETL

Idle

Extract entities and relationships from CMMC documents to build the knowledge graph. Select source folders and monitor the extraction progress in real-time.

Select Source Folders:

Loading available folders...
ETL Console Output 0 lines
[Ready] Waiting to start...

Graph Versions

Version Status Nodes Relationships Created Actions
Loading versions...

Graph Validation & Cleanup

Idle

Post-ETL validation to clean up the knowledge graph: merge duplicates, normalize IDs, remove orphan nodes, and prune low-confidence relationships.

60%
Validation Console Output 0 lines
[Ready] Select a graph and start validation...

Add New Company

Companies

Manage companies, view documents, monitor chat sessions, and track user interactions. Click on a company card to expand and see details.

Loading companies...

CMMC Dashboard ETL

Idle

Import company compliance data from the synthetic_data folder into the CMMC Dashboard database. Monitor agent activities and extraction progress in real-time.

Select Companies to Import:

Loading available companies...

ETL Console

0 lines
[Ready] Select companies and click Start ETL

Agent Status

17 Agents
Pipeline
File Scanner Idle
Parser Idle
Identifier Idle
Extraction
Company Profile Idle
Controls Idle
Evidence Idle
POA&M Idle
Assets Idle
Personnel & Training Idle
Incidents Idle
Policies Idle
Audits Idle
Contracts Idle
Vulnerabilities Idle
Vendors Idle
Risk Assessment Idle
Analysis
SPRS Calculator Idle
Gap Analyzer Idle
Assessment Readiness Idle
Continuous Monitoring Idle

Model Benchmark Comparison

Advanced Benchmark Options

Gold Set: 20 rigorous CMMC questions covering AC, IA, IR, MP, RA, CM, SC domains | Judge: GPT-5.2 grades responses 1-5 like a real C3PAO auditor | RAG: Retrieves relevant source documentation to ground responses and verify accuracy

Model Management

Click "Scan for GGUF Models" to find models in trained_models/
Loaded in Ollama: -

Training Data Statistics

-
Total Q&A Pairs
-
STIG Questions
-
CVE Questions
-
NIST Questions

Application Model Mapping

Configure which models are used for different tasks in the application.
Used for general chat conversations
Used for knowledge graph extraction
Used for document embeddings
Fallback for quick responses
Used for document analysis
Default model for benchmarking

Available Models

Models available in your LiteLLM proxy configuration. These are sourced from the running LiteLLM instance.
Model Name Provider API Base Status Actions
Loading models...

LiteLLM Proxy Status

Checking LiteLLM status...

Langfuse Observability

Open Langfuse Dashboard
Generate keys in your Langfuse dashboard. These keys enable LLM call logging and observability.