Repository navigation
test(P0): enforce canonical router identity and local fallback in CI - #145
Merged
Rafa-Innerchispa merged 5 commits intoOct 12, 2026
Merged
Conversation
…and gate regression tests
…gh canonical package shim
…ime module changes
Rafa-Innerchispa
marked this pull request as ready for review
October 12, 2026 05:35
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
P0: certificar que el scheduler usa el router canónico y que el fallback pasa CI
Corrección importante de diagnóstico
El paquete
platform/raphiia_openai/__init__.pyes shim de compatibilidad: establece__path__haciainneros_core_runtime. Por tanto,from raphiia_openai import local_model_router(usado pordev_swarm_scheduler.py) carga normalmenteplatform/inneros_core_runtime/local_model_router.py, donde YA existe_intel_ollama_fallback. No es correcto modificar el archivo obsoletoplatform/raphiia_openai/local_model_router.pypara «reparar» el scheduler; esa modificación se revirtió en la rama antes de fusionar.Cambios que sí quedan
platform/tests/test_local_model_router_fallback.py, prueba explícita del archivo y la identidad del módulo importado por Dev Swarm, además de dos pruebas de fallback positivo/negativo en casos de indisponibilidad real del AMD.critical-runtime-contractsahora incluye el test en pytest/compile y vigila cambios en el router canónico y en el shim. La prueba existía en main, pero NO se ejecutaba en los gates.Bloqueo vivo no resuelto por este PR
Los fallos anteriores
amd_vllm_unreachablepueden depender de un endpoint/túnel de vLLM incorrecto, modelo servido diferente o ausencia de Ollama desde Intel; no queda demostrado que la causa sea que se importaba el router erróneo.Para resolverlo dentro de
ops_f46a62594c73: comprobar en hosts activosmodule.__file__, variables de entorno de ruta, GET/v1/modelsy/api/tags, resultado del fallback, y canario real con Temporal/Mongo terminal. No relanzar tareas previas sin reconcile PID/lease.CI debe pasar y la comprobación host es necesaria antes de cualquier declaración de producción.