Scenario 3 of 5

Full Cloud with Orchestrators

Cloud-Native Apps with Agent Builders & Orchestrators — Smartflow Enhances Without Replacing

End Users (External)
Global Users
Mobile Users
Web Users
USER REQUESTS
CLOUD ENVIRONMENT — AWS / Azure / GCP
AI ORCHESTRATION LAYER
Web Apps
Mobile APIs
Microservices
AGENT BUILDERS & ORCHESTRATORS
Glean · LangChain · LlamaIndex · Custom
↕ Symbiotic Integration — Enhances & Reports Back
SMARTFLOW (CLOUD)
SMARTFLOW
Cache · Route · Govern · Optimise
Enhances orchestrators — does not replace them
INTELLIGENT ROUTING
Cloud Providers
AWS Bedrock
Azure OpenAI
Google Gemini
API Providers
OpenAI GPT-4
Anthropic Claude
Cohere
Local / Hosted
Llama 3 (Hosted)
Mistral (Hosted)
Custom Models
Key Capabilities
Where Smartflow Sits
Deployed between the orchestration layer and LLM providers. Enhances existing frameworks without replacing them — your orchestrators stay intact.
Apps Inside Orchestrators
Your apps (web, mobile APIs, microservices) live inside orchestration frameworks. Smartflow transparently intercepts all LLM calls from this layer.
Symbiotic Integration
Bidirectional relationship: Smartflow enhances orchestrators with caching & routing AND reports metrics back to framework dashboards for full observability.
Non-Invasive Enhancement
Keep using Glean, LangChain, LlamaIndex as-is. Smartflow plugs in transparently via network redirect — no code changes to your orchestrators.
Multi-Provider Support
Route to AWS Bedrock, Google Gemini, OpenAI, Anthropic, and self-hosted models (Llama, Mistral) from any orchestrator or app in the platform.
60–80% Cost Savings
4-phase semantic cache reduces provider calls across all apps in the orchestration layer simultaneously. One hit benefits every user.
Framework-Native Reporting
Smartflow pushes telemetry back to LangSmith, Glean Analytics, and other framework dashboards — see Smartflow data in your existing tools.
Patent-Pending Standardisation
Proprietary normalisation layer preserves temperature, top-p, and model-specific tunings when routing between providers. Switch models without breaking behaviour.