Admin Dashboard

The ContextView Dashboard is a Django-based admin UI that provides real-time visibility into the ContextUnity service mesh — agent configurations, execution traces, episodic memory, and system health.

Dashboard Views

View	URL	Description
Overview	/	System KPIs, service health probes, request/error charts
Traces	/traces/	Searchable list of agent execution traces
Trace Detail	/traces/<id>/	Step-by-step inspection with chain-of-thought
Agents	/agents/	Agent configuration list with status indicators
Agent Detail	/agents/<id>/	Permissions, tools, hierarchy, and activity
Memory	/memory/	Episodic and entity memory browser
Events	/events/	System event log with filtering

Architecture

The dashboard reads data from two sources:

                    ┌─────────────┐
                    │  Dashboard  │
                    │   (Django)  │
                    └──────┬──────┘
                           │
              ┌────────────┼────────────┐
              ▼                         ▼
      ┌──────────────┐         ┌──────────────┐
      │  Django DB   │         │   Brain DB   │
      │  (SQLite)    │         │  (Postgres)  │
      ├──────────────┤         ├──────────────┤
      │ Agent configs│         │ Traces       │
      │ Permissions  │         │ Episodes     │
      │ Tool policies│         │ Knowledge    │
      └──────────────┘         │ Analytics    │
                               └──────────────┘

• Agent configurations → Django DB (local to View, managed via Django Admin)
• Traces, episodes, facts → Brain PostgreSQL via brain_db queries
• Health checks → Direct TCP probes to each service endpoint

KPI Dashboard

The overview page displays:

• Total / Active agents — from Django Agent model
• Trace volume — 24h/7d/30d request counts from Brain
• Error rates — aggregated from trace error classifications
• Service health — TCP probe status for Router, Brain, Shield, Worker
• Request/error charts — time-series rendered via Chart.js

Time range is selectable via ?range=24h|7d|30d query parameter.

Trace Inspector

The trace detail view provides:

1. Step timeline — ordered execution steps with timestamps
2. Chain of thought — LLM reasoning for each step
3. Tool calls — tool names, arguments, and results
4. Token usage — input/output/total tokens per step
5. Langfuse link — direct cross-reference to Langfuse trace (when configured)

Agent Hierarchy

The agent management view shows:

• Permission tree — visual hierarchy of agent capabilities
• Tool policies — which tools each agent can invoke
• Activity log — recent agent execution history
• Configuration editor — update agent settings via Django Admin

Data Layer

All Brain data access goes through the data/brain_db/ package, which queries Brain’s PostgreSQL schema directly:

Module	Queries
traces.py	Trace search, detail, chain-of-thought
episodes.py	Episode listing, stats, memory browsing
knowledge.py	Knowledge graph queries
agents.py	Agent activity from trace data
analytics.py	System and error analytics aggregations
schemas.py	Typed row schemas (TraceRow, EpisodeRow, etc.)

Langfuse Integration

When configured (LANGFUSE_HOST + LANGFUSE_PROJECT_ID), the dashboard constructs direct links to Langfuse traces for deep LLM debugging — token-level analysis, prompt inspection, and cost tracking.