AI providers module
Choose where agent workloads run without hardcoding the integration.
Current OpenAI-compatible and Ollama profiles sit behind common provider contracts, so additional hosted or local drivers can be added without redesigning agents.
Inside the module
See the working surface, step by step.
Use the controls to inspect the current application interface. Any arrow or step selection stops the automatic tour and leaves your chosen screenshot in place.
What it covers
Focused functions. Shared project truth.
The module has a clear operating boundary while remaining connected to the people, agents, automation, and evidence around it.
-
Current drivers, extensible contract
Configure OpenAI-compatible and Ollama chat profiles now; common provider and capability contracts allow more hosted, private, or local drivers to be added.
-
Capabilities before brand
Record streaming, tool calling, structured output, background responses, image generation, hosted tools, and local MCP support per profile.
-
Health and behavior checks
Test provider connectivity and chat behavior before binding a profile to an agent.
-
Credential boundaries
Bind managed secret references while keeping credential values out of provider records, agent definitions, and diagnostics.
-
Pricing and observation
Maintain standard and long-context pricing per model, then combine it with provider-reported or calculated runtime usage. Unknown pricing remains visible instead of being guessed.
-
Runtime resilience
Use pooled handles, provider/model concurrency gates, retry classification, and explicit unsupported-capability failures.
How it connects
Provider choice stays replaceable. Cost stays attributable.
Agents bind to profiles while projects, processes, and workflows depend on the agent contract. Observed usage and known pricing can be grouped by provider, model, agent, workflow or process, and the project that authorized the work—without hiding observations whose price is still unknown.