Admin
Monitor usage, evaluate model quality, and communicate with your users, all from the Admin Panel.
Hrida.ai gives administrators a full operational toolkit. Track which models your team uses and how much they cost. Run blind evaluations to find the best model for your workload. Push announcements to every user with customizable banners. Wire up webhooks to get notified when users sign up or when long-running tasks finish.
What's In This Section
| 📊 Analytics | Usage dashboards with message volume, token consumption, model breakdowns, and per-user activity |
| 🏆 Evaluation | Model arena, blind A/B testing, ELO-based leaderboards, and chat snapshots for fine-tuning |
| 📢 Banners | Customizable system-wide announcements with dismissible notifications |
| 🔔 Webhooks | Automated notifications for user sign-ups, chat completions, and external service integration |
| 🚨 Incident Management | Plugin-based integration with Grafana OnCall, PagerDuty, and Opsgenie — live incident feed + agent triggers |
| 💰 Model Pricing | Configure per-token USD costs for every model so analytics and usage limits can track spend |
| ⚡ Usage Limits | Set per-user or per-group daily/monthly token and cost quotas with hard or soft enforcement |
| 🤖 Digital Employees | Plugin-based multi-agent system — configure HR, IT, Documentation agents with per-agent access control |
Analytics
Get a bird's-eye view of how your instance is being used. The analytics dashboard shows message volume, token consumption, model popularity, and user activity, filterable by time period and user group.
Use it for: capacity planning, cost estimation, identifying popular models, tracking adoption trends, and per-department usage reporting.
Evaluation
Not sure which model fits your team best? The evaluation system lets users rate model responses with thumbs up/down during normal conversations. Ratings feed into an ELO-based leaderboard, and rated chats are snapshotted for future model fine-tuning.
Two evaluation modes:
- Arena Mode: Randomly selects models for blind comparison, removing bias
- Normal interaction: Rate responses during regular use, then compare models by swapping them out
Banners
Push high-visibility messages to every logged-in user. Use banners for maintenance windows, incident alerts, policy reminders, and feature announcements. Configure via the Admin Panel or the HRIDAAI_BANNERS environment variable for GitOps workflows.
| Type | Color | Best for |
|---|---|---|
info | Blue | General announcements |
success | Green | Resolved incidents, completed changes |
warning | Yellow | Planned maintenance, partial degradation |
error | Red | Active incidents, outages |
Webhooks
Three types of webhook integrations keep your external services in sync:
| Webhook | Purpose |
|---|---|
| Admin Webhook | Notifies a Slack/ channel when a new user signs up |
| User Webhook | Notifies individual users when a long-running chat response is ready |
| Channel Webhooks | Let external services (CI/CD, monitoring, scripts) post messages into Hrida.ai channels |
Model Pricing
Record the input and output token costs (in USD per million tokens) for every model in your instance. Once set, Hrida AI Studio automatically calculates the cost of each message and stores it alongside the conversation. Cost data powers the Analytics spend columns and feeds the Usage Limits enforcement engine.
Common models (OpenAI, Anthropic, Google) are seeded with default prices on first startup — update any entry to match your negotiated rates or custom deployments.
Usage Limits
Cap how much any user or group can consume per day or per month, measured in tokens or USD. Limits can be hard (requests are blocked at the threshold) or soft (users are warned but not blocked). An alert threshold (default 80%) triggers an in-app banner before the limit is reached.
Requires Model Pricing to be configured for cost-based (cost_usd) limits.