The whole routing chain for this world
const url = 'http://localhost:11044/api/v1/config/router';const options = {method: 'GET', headers: {'X-Embabel-Api-Key': '<X-Embabel-Api-Key>'}};
try { const response = await fetch(url, options); const data = await response.json(); console.log(data);} catch (error) { console.error(error);}curl --request GET \ --url http://localhost:11044/api/v1/config/router \ --header 'X-Embabel-Api-Key: <X-Embabel-Api-Key>'Each context with the role or model it uses and the model that resolves; every role with its deployment and world model and the contexts that use it; the chat models available; and the deployment’s own model settings, which no world can change yet.
Authorizations
Section titled “Authorizations”Responses
Section titled “Responses”The chain, resolved
The whole model-routing chain of the signed-in user’s world, resolved: contexts, roles, the models available, and the deployment’s own model settings. Answered by GET /config/router
object
Every context, with the role or model it uses and the model that resolves
One context, a job this world uses a model for, with the role or model it is set to and the model that answers it. Part of GET /config/router
object
The model the deployment’s default for this context names; null when that default selects by role, or there is no default
The role the deployment’s default for this context selects by; null when that default names a model, or there is no default (the planner, which follows chat, or a context the deployment leaves unset)
True for the planner when it has no role or model of its own, so it runs on whatever chat runs on, through chat’s role; false for every other context
The context: chat, plannerLlm, codingLlm, fallbackCodingLlm, rankerLlm or lensLlm; the {id} in PUT /config/router/contexts/{id}
The model this world pinned the context to; null when it pinned none
The model it runs on now; null when nothing can serve it, or when the context names no model at all (a fallback coder with neither a setting nor a default)
object
True when its provider runs locally: LM Studio, Ollama or Docker
The model’s name, as the appliance’s model provider knows it
Who serves it, as the model provider names it (for example LM Studio); null when the model is not among those the provider lists
The role this world chose for the context; null when it chose none
The role the context reaches its model through, from this world’s setting or else the deployment’s default; null when it reaches a model without one
Model settings the deployment makes that no world can change yet
A model setting the deployment makes under assistant.infrastructure, which no world can change yet; listed in GET /config/router as deploymentUses
object
The model it names; null when it selects by role rather than by name
The LLM role it selects its model through; null when it does not select by role
Where it is configured: the property path under assistant.infrastructure, by Kotlin property name (camelCase)
The language models the appliance can serve, local ones first, then by name
A model as the router resolves it: its name, who serves it, and whether it runs on this machine
object
True when its provider runs locally: LM Studio, Ollama or Docker
The model’s name, as the appliance’s model provider knows it
Who serves it, as the model provider names it (for example LM Studio); null when the model is not among those the provider lists
Standard roles first, then the deployment’s, then this world’s custom roles
One LLM role, standard, mapped by the deployment, or added by this world, with the model it resolves to and the contexts that use it. Part of GET /config/router
object
True for a role this world added, the only kind that can be deleted
The model the deployment maps this role to in embabel.models.llms; null when it maps none
What the role is for; null for a role only the deployment declares, which carries no description
Where a standard role is shown: GENERAL or VIRTUAL_CYPHER; null for deployment and custom roles
The role’s id, such as chat_best or vc_execution
The model the role reaches now: this world’s choice when it made one, else what the platform selects for the role; null when nothing can serve it
object
True when its provider runs locally: LM Studio, Ollama or Docker
The model’s name, as the appliance’s model provider knows it
Who serves it, as the model provider names it (for example LM Studio); null when the model is not among those the provider lists
Ids of the contexts that reach their model through this role
The model this world chose for the role, overriding the deployment’s; null when it chose none