> ## Documentation Index > Fetch the complete documentation index at: https://docs.domesystems.ai/llms.txt > Use this file to discover all available pages before exploring further. # Move LLM Pool > Reorders a pool relative to its workspace siblings. Exactly one of before_pool_id or after_pool_id must be set in the request; the moved pool's sort_key is rewritten to the midpoint between the chosen neighbor and its adjacent sibling. The resolver evaluates pools' match_when predicates in sort_key ascending order, so ordering controls which pool wins when multiple match_when predicates could fire for the same request. ## OpenAPI ````yaml POST /v1/models/pools/{pool_id}/move openapi: 3.0.3 info: title: Dome Platform API description: Connect and REST APIs for the Dome agent governance platform. version: 2.0.0 servers: - url: https://api.domesystems.ai description: Production security: [] tags: - name: Agents - name: Authorization - name: Guards - name: Tools - name: Tool Catalog - name: Models - name: Model Pools - name: Quotas - name: Gateways - name: Webhooks - name: Audit - name: Workspaces - name: Access Control - name: Tenants - name: Platform Keys - name: Enrollment - name: Platform - name: Identity paths: /v1/models/pools/{pool_id}/move: post: tags: - Model Pools description: |- Reorders a pool relative to its workspace siblings. Exactly one of before_pool_id or after_pool_id must be set in the request; the moved pool's sort_key is rewritten to the midpoint between the chosen neighbor and its adjacent sibling. The resolver evaluates pools' match_when predicates in sort_key ascending order, so ordering controls which pool wins when multiple match_when predicates could fire for the same request. operationId: Management_MoveLLMPool parameters: - name: pool_id in: path required: true schema: type: string requestBody: content: application/json: schema: $ref: '#/components/schemas/MoveLLMPoolRequest' required: true responses: '200': description: OK content: application/json: schema: $ref: '#/components/schemas/MoveLLMPoolResponse' components: schemas: MoveLLMPoolRequest: type: object properties: workspace_id: type: string pool_id: type: string before_pool_id: type: string description: |- Exactly one of before_pool_id / after_pool_id must be set. The moved pool's sort_key is rewritten to the midpoint between the named neighbor and its adjacent sibling. Empty/missing means "no constraint on that side" — use before_pool_id alone to place the moved pool immediately before that pool (with no upper neighbor when it is currently the first); use after_pool_id alone to place it immediately after (no lower neighbor when it is the last). The "top" of the list is the lowest sort_key. after_pool_id: type: string MoveLLMPoolResponse: type: object properties: pool: $ref: '#/components/schemas/LLMPool' LLMPool: type: object properties: id: type: string workspace_id: type: string name: type: string description: type: string match_when: type: object description: |- Optional `match_when` predicate. Null/empty Struct means match by exact name only. Grammar: prompt_tokens / requested_max_tokens / tool_count numeric comparators ({gt, lt, eq}); endpoint string equality; header. case-insensitive equality; principal.metadata. equality; principal.act_as. — the verified end-user identity (sub/email scalars, claims. scalar record, roles/groups Set with {contains|containsAny| containsAll}), the same trust boundary Cedar gates tool access on. AND across keys. See plan §9.2. is_default: type: boolean cache_ttl_secs: type: integer format: int32 cache_scope: type: string description: 'One of: workspace | caller.' failover_max: type: integer description: |- Maximum number of additional dispatch attempts after the primary member fails. null and 0 are equivalent at runtime (no failover); null preserves the "admin has not made an explicit choice" state. When N > 0, the resolver builds a Candidates slice of length min(N+1, eligible_member_count) ordered by priority ascending and strategy-within-tier. format: int32 routing_strategy: type: string description: |- One of: priority_weighted | round_robin | least_loaded | most_quota_remaining. most_quota_remaining routes to the member with the most estimated cost-quota headroom remaining (members with no model quota are unbounded and routed last); ties are broken by round-robin rotation keyed by strategy_scope. weight / priority are ignored, as with round_robin / least_loaded. created_at: type: string format: date-time strategy_scope: type: string description: |- Scope of stateful routing counters. One of: workspace | caller. workspace: a single counter/inflight bucket per pool (round_robin, most_quota_remaining tiebreak) or per connection (least_loaded). caller: keyed additionally by agent_id so each agent observes its own rotation / load. Ignored for the stateless priority_weighted strategy. Defaults to workspace. sort_key: type: number description: |- Explicit per-workspace ordering. Pools are evaluated for match_when in ascending sort_key order. Managed by MoveLLMPool; direct edits via Create/Update are not supported. format: double reachable: type: boolean description: |- Gateway: false when the resource is in no Gateway → unreachable by agents/users. Computed, not stored. gateway_ids: type: array items: type: string description: |- Gateway: the Gateways this pool is a member of (gateway_llm_pool_members), as UUID strings. Computed at read time; lets the UI render Gateway chips without a per-row GetGateway fan-out. Empty ⇔ reachable is false. description: |- LLMPool: a logical name (e.g. "default-chat") that fans out to one or more LLMModelConnections via LLMPoolMember rows. See plan §7 / §9. ````