Dapto Prism API: From Questions to Execution - In Your Apps
Use one governed API to chat with any model, power agent workflows, and keep work organized across projects and teams. Prism brings Dapto’s routing, controls, and audit trail into your product-without vendor lock-in.
Build in the Console. Ship via API.
Your team can prototype prompts, agents, and policies in Dapto AI Console, then reuse the same building blocks in your application through Prism API.
Production AI Integration Without the Chaos
Standardize multi-model AI in your product with routing, agent execution, and governance built in.
One API for Any Model
Call OpenAI, Gemini, Claude, Mistral, Llama, and self-hosted models through one endpoint.
Smart Routing + Override
Route by cost, quality, latency, or policy-and override when you need explicit control.
Governance Built In
Apply policies, redact sensitive data, and capture audit signals before calls ever hit the model.
Agent-Ready Execution
Turn prompt chains into reusable workflows your app can run reliably at scale.
Cost + Usage Visibility
Track spend per key, team, workspace, or project-set caps and get alerted early.
Future-Proof Integration
Add or switch models without rewriting your app. Keep your interface stable.
From Simple Calls to Full Execution
Prism API isn’t just model access-it’s the execution layer behind Dapto’s AI workspace.
1. Universal Multi-Model Endpoint
Single endpoint for OpenAI, Anthropic, Google, Mistral, open-source, and self
BYOK (bring your own keys) or managed keys with quotas, limits, and usage controls.
Switch, combine, or compare model outputs programmatically-without vendor lock
2. Routing That Matches the Task
Route by cost, policy, latency, or quality (and always keep manual override).
Task-based routing: match summarization vs analysis vs extraction to the best model.
Fallbacks and guardrails to keep results reliable when a model is unavailable or restricted.
3. Governance That Travels With the Call
Central policy enforcement across models (prompt + response).
Redaction of PII/PHI/sensitive content before dispatch and before returning to users.
Audit signals per call (who/what/where/which model), exportable to SIEM/compliance systems.
4. Developer-First Experience
Predictable request/response format across providers, with clean error handling.
Versioning, environments (sandbox), and async-friendly patterns for high-volume workloads.
SDK-ready surface area so teams can integrate fast (and keep maintenance low).
5. Agents & Workflow Execution
Run multi-step workflows: draft → refine → extract → format.
Reuse agent instructions across apps and teams for consistent results.
Trigger workflows from events (webhooks) or scheduled jobs.
6. Usage & Cost Intelligence
Track spend per project, workspace, key, model, or route strategy.
Set quotas/caps and get alerts for anomalies or runaway usage.
Build cost-aware experiences for your users with pricing predictability.
A Single Call - With Execution Signals
Same endpoint, any model. Routing, policy checks, and audit signals included in the response.
Multi-Model API Call Example
Request
"messages": [
{"role": "user", "content": "Draft a QBR summary and action plan from these notes..."}
],
"routing": {
"strategy": "quality_first",
"fallback": "cost_optimized"
},
"context": {
"workspace_id": "ws_finance",
"project_id": "proj_qbr_2026_q1"
},
"governance": {
"redact_sensitive": true,
"audit_level": "standard"
}
}
Response
"id": "chatcmpl-abc123",
"model_used": "gpt-4o-mini",
"routing_reason": "quality_first",
"context": { "workspace_id": "ws_finance", "project_id": "proj_qbr_2026_q1" },
"governance": {
"sensitive_redacted": 2,
"policy_violations": 0,
"audit_id": "audit_xyz789"
},
"choices": [{ "message": { "role": "assistant", "content": "Here’s the QBR summary..." } }],
"usage": { "total_tokens": 512, "cost_usd": 0.0061 }
}
What Prism did automatically:
Governance
Redacted sensitive data + applied policies
Routing
Selected the best model for the strategy
Audit Signals
Returned audit_id + usage for traceability