Skip to content

Admin Panel and Dashboard

Jump to bottom
fdanobey edited this page Sep 2, 2026 · 5 revisions

Admin Panel & Dashboard

OBEY API Gateway ships with two embedded web UIs — an Admin Panel for configuration management and a real-time Dashboard for monitoring.

Admin Panel

The Admin Panel provides a visual interface for managing all gateway configuration without editing YAML files.

URL: http://localhost:8080/admin (configurable via admin.path)

Admin Panel

Tabs

Tab Purpose
Server Settings Host, port, timeouts, TLS, admin auth, dashboard, Prometheus, Codex backend URL
Providers Add/edit/remove providers, configure API keys, timeouts, connection pools
Model Groups Define model groups, priority ordering, per-model cost and tier
Routing Settings Retry policy, Cache-Aware Routing, Reasoning Compatibility, CORS
Smart Routing Complexity-aware tier selection, classifiers, cascade, budgets, ONNX assets, simulation
Circuit Breaker Failure thresholds, backoff sequences, success thresholds
Logging Log levels, retention, body logging, excluded fields
Caching Exact cache and semantic cache settings
Compression Token compression levels, engines, pipelines, perplexity/ONNX assets
Context Context window management, truncation strategy
Streaming SSE reliability configuration, keep-alive, truncation retry
Loop Detection Session tracking, confidence thresholds, signal weights, enforcement
Virtual Keys Key creation, management, budgets, rate limits, usage tracking
Guardrails Quick Setup wizard, custom rules, Advanced pipeline configuration
Structured Output JSON Schema validation settings, retries, passthrough providers
Persistent Memory Memory store config, browser, extraction, vector search, live stats

Configuration Actions

Button Action
Save Validate and apply changes (hot-reload)
Reset Revert form to current saved state
Reload Config Re-read config.yaml from disk
Export YAML Download current config as YAML file
Import YAML Upload and apply a YAML config file

Authentication

admin:
  enabled: true
  path: "/admin"
  auth:
    enabled: true
    username_env: "ADMIN_USERNAME"    # Env var name
    password_env: "ADMIN_PASSWORD"    # Env var name

When auth is enabled, all admin endpoints require HTTP Basic Auth. Credentials are resolved from environment variables at runtime.

Dashboard

The Dashboard provides real-time metrics, provider health, and log viewing via WebSocket updates.

URL: http://localhost:8080/dashboard (configurable via dashboard.path)

Dashboard Overview

Tabs

Tab Purpose
Metrics Overview Total requests, avg response time, request rate, active requests, cost, cache hit rate, real-time charts
Provider Health Per-provider circuit breaker status, latency, error rates
Cost Breakdown Spending by provider and model
Compression Token compression events/timelines, tool compression statistics (reduction ratio, pruning, feedback levels, progressive disclosure)
Recent Errors Latest error logs with details
Memory Persistent memory events timeline (injections, extractions, evictions), event type distribution, namespace activity
Log Viewer Searchable, filterable request log table with body viewing

Real-Time Metrics

The dashboard connects via WebSocket for live updates (configurable interval):

dashboard:
  enabled: true
  path: "/dashboard"
  metrics_update_interval_seconds: 1

Metrics Cards

Metric Description
Total Requests Cumulative request count since startup
Avg Response Time Rolling average latency
Request Rate Requests per minute
Active Requests Currently in-flight requests
Cumulative Cost Total USD spend across all providers
Cache Hit Rate N/A until first eligible request, then percentage

Provider Health

Provider Health

Shows per-provider:

  • Circuit breaker state (Closed/Open/Half-Open)
  • Average latency
  • Error rate
  • Request count

Log Viewer

Log Viewer

Features:

  • Filter by date range, provider, model, status code, trace ID
  • Configurable result limit
  • Request/response body viewing (when body logging enabled)

Admin API Endpoints

All admin functionality is also available programmatically:

Configuration

# Get current config
curl http://localhost:8080/admin/config

# Update config (validate + write + apply)
curl -X PUT http://localhost:8080/admin/config \
  -H 'Content-Type: application/json' \
  -d '{ ... }'

# Validate without applying
curl -X POST http://localhost:8080/admin/config/validate \
  -H 'Content-Type: application/json' \
  -d '{ ... }'

# Hot-reload from disk
curl -X POST http://localhost:8080/admin/config/reload

# Export YAML
curl http://localhost:8080/admin/config/export

# Import YAML
curl -X POST http://localhost:8080/admin/config/import \
  -H 'Content-Type: application/yaml' \
  --data-binary @config.yaml

Virtual Keys

curl http://localhost:8080/admin/keys              # List
curl -X POST http://localhost:8080/admin/keys      # Create
curl http://localhost:8080/admin/keys/{id}         # Get
curl -X PATCH http://localhost:8080/admin/keys/{id} # Update
curl -X POST http://localhost:8080/admin/keys/{id}/revoke  # Revoke
curl -X DELETE http://localhost:8080/admin/keys/{id}        # Delete

OAuth

curl -X POST http://localhost:8080/admin/oauth/openai/login   # Start login
curl http://localhost:8080/admin/oauth/openai/status           # Check status
curl -X POST http://localhost:8080/admin/oauth/openai/logout   # Logout

Dashboard Data

curl http://localhost:8080/dashboard/metrics   # Current metrics snapshot
curl http://localhost:8080/dashboard/errors    # Recent errors
curl "http://localhost:8080/dashboard/logs?limit=100&provider=openai"  # Filtered logs

Memory

curl "http://localhost:8080/admin/memory/entries?namespace=user::project::myapp"  # List entries
curl -X POST http://localhost:8080/admin/memory/entries \
  -H 'Content-Type: application/json' \
  -d '{"namespace":"user::project::myapp","memory_type":"fact","content":"Uses React 19"}'
curl -X DELETE http://localhost:8080/admin/memory/entries/{id}            # Delete entry
curl -X DELETE http://localhost:8080/admin/memory/namespaces/{namespace}  # Clear namespace
curl http://localhost:8080/admin/memory/stats                             # Store statistics
curl http://localhost:8080/admin/memory/projects                          # Detected projects

ONNX Assets

curl http://localhost:8080/admin/onnx/status                # Model/runtime readiness
curl -X POST http://localhost:8080/admin/onnx/install \
  -H 'Content-Type: application/json' \
  -d '{"model_path":"./models/perplexity_scorer.onnx"}'    # Install assets

Tool Compression

# Get current tool compression statistics
curl http://localhost:8080/admin/tool-compression/stats

# Get feedback loop state per model group
curl http://localhost:8080/admin/tool-compression/feedback

# Test compression against a tools array without sending to a provider
curl -X POST http://localhost:8080/admin/tool-compression/test \
  -H 'Content-Type: application/json' \
  -d '{"tools": [...], "model_group": "gpt-4-group"}'

Loop Detection

# List active sessions (paginated)
curl "http://localhost:8080/admin/loop-detection/sessions?limit=50&offset=0"

# Get full session detail
curl http://localhost:8080/admin/loop-detection/sessions/{session_id}

# Reset a session's enforcement state
curl -X POST http://localhost:8080/admin/loop-detection/sessions/{session_id}/reset

# Aggregate statistics
curl http://localhost:8080/admin/loop-detection/stats

Metrics

curl -X POST http://localhost:8080/admin/metrics/reset-active  # Reset active request counter

Provider Utilities

# Test connectivity to a provider (validates URL + auth)
curl -X POST http://localhost:8080/admin/test-connection \
  -H 'Content-Type: application/json' \
  -d '{"provider_name": "openai"}'

# Proxy provider's /v1/models endpoint (for model discovery in the UI)
curl http://localhost:8080/admin/providers/models?provider=openai

Hot Reload

Configuration can be changed and applied without restarting the gateway:

  1. Via Admin Panel — edit settings and click Save
  2. Via API — POST /admin/config/reload to re-read from disk
  3. Via PUT — PUT /admin/config to validate, save, and apply in one call

On reload:

  • All circuit breakers are reset
  • New provider settings take effect immediately
  • Active requests complete with old settings

Prometheus Metrics

For integration with existing monitoring infrastructure:

prometheus:
  enabled: true
  path: "/metrics"

Exposes standard Prometheus text format at http://localhost:8080/metrics.

Key metrics include:

  • obey_api_requests_total{provider, model, status}
  • obey_api_request_duration_seconds{provider, model}
  • obey_api_cache_hits_total{tier}
  • obey_api_circuit_breaker_state{provider}
  • obey_api_guardrail_stage_executions_total{pipeline, stage, provider, action}
  • obey_api_guardrail_stage_latency_ms{pipeline, stage, provider}

Next Steps

OBEY API Gateway

Home

Getting Started

Core Features

Access Control

Operations

GitHub Repository | Releases

Clone this wiki locally