Per-virtual-model repetition_limit override field #69

Activity

1. weselben

   added

   wayfinder:taskWayfinder ticketWayfinder ticket

   on Sep 1, 2026
2. weselben

   added a parent issue on Sep 1, 2026

   • Stream repetition canceller — wayfinder map #53
3. weselben commented on Sep 1, 2026

   

   weselben

   on Sep 1, 2026

   OwnerAuthor

   More actions

   Scope update from #73: this task now ships two virtual-model fields, not one.

   • repetition_limit — tri-state *int (unset inherit / 0 explicit off / N on), schema pattern: RawResilienceConfig pointer style.
   • repetition_max_pattern — int, unset = inherit global (default 8), accepted range 1–64.
     Both render in the Models view next to slowdown. Validation rejects limit < 0 and pattern outside 1–64.
4. weselben commented on Sep 1, 2026

   

   weselben

   on Sep 1, 2026

   OwnerAuthor

   More actions

   Repo-conformant notes:

   • Add RepetitionLimit *int and RepetitionMaxPattern *int to the virtual-model config struct in internal/virtualmodels/. Pointer type matches RawResilienceConfig (config/resilience.go) for the unset/inherit semantics.
   • Validation in internal/virtualmodels/validation.go (existing pattern): reject RepetitionLimit < 0; reject RepetitionMaxPattern outside 1..64; accept nil = inherit.
   • Wire through the orchestrator: gateway.InferenceConfig.StreamRepetitionLimit and StreamRepetitionMaxPattern (new fields). StreamChatCompletion and StreamResponses resolve the effective value per request: per-model field > virtual-model field > workflow field > global default > guard off.
   • Models-view form: two number inputs next to slowdown in the existing virtual-model form renderer. Same labels, same disabled-state handling.

   Tests: extend internal/virtualmodels/config_overlay_test.go and internal/gateway/request_model_resolution_test.go with the new fields. Pattern: table-driven, assert resolution order. Repo norm: no testify.

   Minimal-diff: one struct + two fields, one validator, two wiring sites, two form inputs.
5. weselben commented on Sep 2, 2026

   

   weselben

   on Sep 2, 2026

   OwnerAuthor

   More actions

   Resolved on feat/stream-repetition-canceller (commits c98d41f, 7567c64).

   What changed:

   • internal/virtualmodels: RepetitionLimit and RepetitionMaxPattern added to types, config, validation, Mongo/SQL stores, and the admin handler form (next to slowdown).
   • config/virtualmodels.go + test: schema support for the two fields.
   • internal/virtualmodels/repetition.go: ResolveRepetitionLimit honors alias redirect -> policy -> inherit, with each field independently nullable so an explicit 0 disables the guard.
   • internal/core/request_model_resolution.go: added RepetitionLimit / RepetitionMaxPattern to RequestModelResolution.
   • internal/gateway: request_model_resolution.go resolves the override, inference_execute.go resolveEffectiveRepetition merges it (nil inherits orchestrator defaults), and StreamResult carries the final pair to WrapDeliveryStream.
   • Tests: internal/virtualmodels/repetition_test.go, internal/gateway/inference_execute_repetition_test.go, internal/gateway/request_model_resolution_test.go.

   Evidence: go test ./internal/virtualmodels/... ./internal/gateway/... green; a virtual model with repetition_limit: 4 triggers on a 4-repeat loop; no value falls back to the global default.

   🤖 Written by Kimi Code (AI agent) from the stream-repetition-canceller worktree.
6. weselben

   closed this as completedon Sep 2, 2026
7. weselben

   added a commit that references this issue on Sep 3, 2026

   feat(virtualmodels): per-model repetition_limit and repetition_max_pa…

   ...

   This issue will close once commit d92dcdc is merged into the 'main' branch.

   d92dcdc