How to Test Bonito

Create a fresh account to test the full onboarding flow.

• Go to /register and create an account with a valid email + password
• You should be redirected to the dashboard after registration
• Check /auth/me returns your user profile

Verify session management works correctly.

• Log out, then log back in with the same credentials
• Try logging in with a wrong password — should show 'Invalid credentials'
• After login, refreshing the page should keep you logged in (JWT stored)

Connect your AWS account to pull Bedrock models.

• Go to Providers page → click Add Provider → select AWS
• Enter Access Key ID, Secret Access Key, and Region (e.g. us-east-1)
• Provider should validate (STS get-caller-identity) and show as 'Active'
• Models page should now show AWS Bedrock models (Claude, Llama, Titan, etc.)

Connect your Azure subscription for OpenAI and Cognitive Services models.

• Add Provider → Azure → Enter Tenant ID, Client ID, Client Secret, Subscription ID
• Provider should validate (OAuth2 token acquisition) and show as 'Active'
• Models page should show Azure OpenAI models (GPT-4, GPT-4o, etc.)

Connect GCP for Gemini and other Vertex AI models.

• Add Provider → GCP → Upload or paste your Service Account JSON
• Provider should validate (project lookup) and show as 'Active'
• Models page should show Vertex AI models (Gemini, PaLM, etc.)

All synced models should appear with correct provider tabs.

• Models page shows all synced models from connected providers
• Filter tabs show ALL connected providers (AWS, Azure, GCP)
• If a provider tab shows ⚠️, click Sync to re-fetch models
• Search bar filters by model name or ID
• Click a model card to see details (pricing, capabilities, context window)

Test models live from the browser.

• Click a model → open the Playground tab
• Send a test message — response should stream back
• Check that token usage and cost appear after the response
• Try with different temperature / max token settings

Compare responses from multiple models side-by-side.

• Select 2-4 models for comparison
• Send the same prompt to all — responses appear side by side
• Compare latency, token usage, and cost across models

Create a gateway key to route requests through Bonito.

• Go to Gateway page → click 'Create Key'
• Copy the generated key (bn-xxx format)
• Key should appear in the keys list with creation date

Test the OpenAI-compatible proxy endpoint.

• Use curl or any OpenAI SDK pointed at your Bonito gateway URL
• Send a chat completion request with your bn-xxx key
• Response should come back in OpenAI format
• Check Gateway → Logs to see the request logged with cost + tokens

The gateway page shows ready-to-use code snippets.

• Copy the Python snippet and run it locally
• Copy the curl snippet and run it in your terminal
• Both should return a valid chat completion response

Set up intelligent routing between models/providers.

• Go to Routing → Create Policy
• Select a strategy: cost-optimized, latency-optimized, balanced, failover, or A/B test
• Assign primary and fallback models
• For A/B testing: set percentage weights (must sum to 100)
• Save the policy — it should appear in the list

Dry-run model selection to verify routing logic.

• Click 'Test' on a policy — it should show which model would be selected
• For failover: verify the fallback model is picked when primary is down
• For A/B: run multiple tests and verify the distribution matches weights

Deploy a model directly into your cloud from the Bonito UI.

• Go to Deployments page → click 'Create Deployment'
• Select a provider and model
• AWS: choose On-demand or Provisioned Throughput (PT requires commitment)
• Azure: set TPM capacity for the deployment (Standard or GlobalStandard tier)
• GCP: serverless by default — verify access
• Deployment should appear in the list with status updates

Check that deployment lifecycle notifications work.

• Deployment status should update: Creating → Active (or Failed)
• In-app notification should appear for deployment status changes
• Deployment details page shows provider, model, capacity, and status

Check real cost data from your cloud providers.

• Costs page shows aggregated spending across all providers
• Breakdown by provider (AWS, Azure, GCP) with charts
• Cost forecast shows projected spending with confidence bounds

Get optimization suggestions.

• Recommendations endpoint returns cheaper model alternatives
• Cross-provider routing recommendations appear if applicable

Verify security posture across all connected providers.

• Compliance page shows check results by provider
• AWS: Bedrock logging, IAM permissions, EBS encryption, CloudTrail
• Azure: Network rules, RBAC roles, diagnostic settings
• GCP: SA permissions, audit logging, VPC Service Controls
• Framework mapping: SOC2, HIPAA, GDPR, ISO 27001

All sensitive actions are logged.

• Audit page shows a timeline of actions (logins, provider connects, invocations)
• Each entry has timestamp, user, action type, and details

Track API usage and trends.

• Analytics page shows overview cards (requests, tokens, cost, avg latency)
• Usage charts show daily/weekly/monthly trends
• Cost breakdown by provider and model

Check that in-app notifications work.

• Bell icon in the header shows unread count
• Click to see notification list with read/unread states
• Mark notifications as read

Set up budget and compliance alerts.

• Create an alert rule with a budget threshold
• Set notification preferences (email, in-app, weekly digest)

Test the Groq-powered AI assistant.

• Click the AI Copilot panel (or Cmd+K)
• Ask: 'What are my costs this month?'
• Ask: 'Which models are available?'
• Ask: 'Run a compliance check'
• Responses should be context-aware (knows your providers, models, costs)
• Quick action buttons should work (Cost Summary, Compliance Check, etc.)