Onaro™ONARO

Implementation guides

Step-by-step guides, grouped by the team that does the work.

For finance teams

Export what Meridian needs from your accounting system, then post locked chargeback months back to the general ledger.

  • QuickBooks Online + Meridian

    Available

    Connect QuickBooks Online, map cost pools to classes, and post locked AI chargeback months as balanced journal entries. Written for controllers: what Meridian can and cannot touch in your books, and why each setup step matters.

    20 minutes · Beginner · Best for: Controllers and finance leads posting AI chargebacks to QuickBooks Online

  • Dynamics 365 Business Central + Meridian

    Early access

    Connect Business Central, set up the MERIDIAN journal batch and permission set, and post locked AI chargeback months as general journal entries tagged with a Department dimension. Written for controllers and the admins who carry out the setup.

    45 minutes · Intermediate · Best for: Controllers and finance leads posting AI chargebacks to Business Central

For platform and engineering teams

Connect AI providers, gateways and GPU telemetry so Meridian can attribute spend to agents, owners and departments.

  • LiteLLM + Meridian

    Early access

    Send LiteLLM Generic API callback logs to Meridian gateway ingest, attribute calls by virtual key in the gateway key registry, and verify usage in the ledger.

    1-2 hours · Intermediate · Best for: Teams running LiteLLM as an AI gateway

  • Kong AI Gateway + Meridian

    Early access

    Send Kong AI Gateway audit logs to Meridian with http-log, map Kong consumers to agents in the gateway key registry, and verify usage in the ledger.

    2-3 hours · Intermediate · Best for: Teams standardizing on Kong AI Gateway

  • Portkey + Meridian

    Early access

    Forward Portkey Logs Export records to Meridian gateway ingest on a schedule, attribute them by agent id metadata, and verify usage in the ledger.

    1-2 hours · Intermediate · Best for: Teams using Portkey as an AI gateway

  • GPU capacity feed + Meridian

    Early access

    Send DCGM GPU metrics from Prometheus remote write or the OpenTelemetry Collector to Meridian, map pods to endpoints, and split cost pools by GPU time.

    2-3 hours · Intermediate · Best for: Teams metering private GPU pools with Prometheus or OpenTelemetry

  • FOCUS-format cloud exports guide

    Early access

    Set up FOCUS Cost and Usage exports in AWS, Azure and Google Cloud and import them in Meridian Uploads.

    Guide not published yet.

Engineering practices for cheaper, more reliable LLM operations

General engineering patterns that sit next to Meridian, not inside it. None of them requires Meridian, and Meridian does not implement them for you.

  • Edge Proxy

    Learn how to implement an edge proxy for AI APIs: route traffic, balance load, enforce policies, and cut latency. This Onaro™ guide covers architecture patterns, provider configuration, and safe rollout for high-volume OpenAI and Anthropic workloads.

    2-4 hours · Intermediate · Best for: Organizations with >100K API calls/month

  • Circuit Breakers

    Add circuit breakers around LLM calls to stop cascading failures, shed load during outages, and avoid runaway spend when APIs degrade. Step-by-step patterns for retries, fallbacks, and observability in production AI systems.

    2-3 hours · Intermediate · Best for: Production systems with high availability requirements

  • Semantic Caching

    Implement semantic caching so similar prompts hit a cache instead of the model—often cutting API cost dramatically. Covers embeddings, similarity thresholds, invalidation, and when caching is safe for your use case.

    4-6 hours · Advanced · Best for: Applications with repetitive or similar queries

  • Model Switching

    Route tasks to the right model tier: cheap models for simple work, premium models where quality matters. Practical routing rules, evaluation tips, and examples to lower spend without surprising regressions.

    2-4 hours · Beginner · Best for: Multi-task AI applications

  • Prompt Compression

    Compress prompts and context to cut token usage 30–50% while preserving answer quality: summarization, structured extraction, trimming policies, and measurement so savings show up in your real traffic.

    3-5 hours · Intermediate · Best for: Applications with long context windows

  • Response Streaming

    Stream model responses to users for snappier UX without raising token cost. Covers SSE patterns, client handling, backpressure, and provider-specific streaming options for chat and agent interfaces.

    1-2 hours · Beginner · Best for: All user-facing AI applications

  • Batch Processing

    Batch LLM and embedding jobs to unlock provider batch discounts and simpler rate limits. When to batch, how to chunk inputs, idempotency, and monitoring so throughput goes up and per-token cost goes down.

    3-4 hours · Intermediate · Best for: Applications with bulk processing needs