38-Provider Unified API
Switch or blend frontier, regional, and on-prem open-weight models via configuration without rewriting agent code.
CEEZ Compass is the horizontal model control plane of the Ceez platform. It centralizes model routing, latency/quality SLAs, data-residency boundaries, semantic caching, and hard token budgets across 38 commercial and self-hosted AI providers—while identifying repetitive LLM paths that can be compiled into $0 deterministic scripts.
Every model request is evaluated for task tier, data classification, budget headroom, and cache hit before dispatch.
When individual engineering teams embed direct API keys to frontier models, costs spike unpredictably, provider outages halt operations, and sensitive data crosses residency boundaries. Compass puts a single policy control plane between every agent or workflow and 38 model providers.
Platform teams gain real-time tokenomics attribution down to each workflow step, automatic multi-provider failover, and a structured path to reduce inference costs by 60–80%.
Connect OpenAI, Anthropic, Google Vertex AI, AWS Bedrock, Azure OpenAI, Mistral, Cohere, DeepSeek, and self-hosted vLLM/Ollama behind one unified gateway.
Match simple classification and extraction to fast sub-cent models while reserving frontier reasoning models for complex multi-hop diagnostics.
Set strict daily/monthly dollar caps per business unit, workflow, or agent team—with sub-second failover when a provider rate-limits or degrades.
Detect high-frequency, low-variance prompt/response patterns and help FDEs compile them into zero-token deterministic rules in Flow.
Every capability in Compass operates within your existing enterprise identity, data residency, and governance boundaries.
Switch or blend frontier, regional, and on-prem open-weight models via configuration without rewriting agent code.
Automatically route confidential or regulated payloads exclusively to sovereign VPC endpoints or air-gapped local GPUs.
Eliminate redundant inference spend and cut tail latency on recurring operational queries and enrichment lookups.
Attribute exact prompt, completion, and tool-call costs to specific teams, use cases, and business outcomes.
Replay live production traffic against candidate models in FDE Workbench to compare accuracy, latency, and cost before switching.
Log exact model ID, temperature, prompt hash, and token counts into Ceez Prove for every decision.
Compass governs model choice and spend across every Ceez Crew agent, Insight query, and FDE Workbench evaluation—as well as across standalone customer applications. In FDE Workbench, teams combine Compass with Ground, Insight, Crew, Flow, Ingest, Govern, Prove and FDE Workbench to deliver repeatable enterprise use cases without building custom infrastructure glue.
Map how Compass connects to your existing systems, data models, and governance controls in a focused 30-day pilot.