fix(providers): container slots advertise hal0 registry id via --alias (cutover routing)#675
Merged
Merged
Conversation
Discovered during the #662 live cutover: hal0/* virtual names never reached container slots. The chat-normalization gate resolves e.g. hal0/chat -> "qwopus3.6-27b-v2" (registry id), but the container's llama-server advertises the raw GGUF basename ("Qwopus3.6-27B-v2-...gguf"), so dispatcher.dispatch() finds no matching upstream -> NoRouteFound -> lemonade fall-through (which then 404s trying to load the model from HuggingFace). Routing via the bare GGUF name worked, confirming the only gap is the advertised name. Fix: pass llama-server `--alias <registry-model-id>` (from model_info._model_key, falling back to [model].default) so the container advertises its hal0 id and the dispatcher matches it. resolved_command_for_slot shows --alias too. Verified on CT105: bare-GGUF route reached the container; normalization emits the exact registry id the alias now advertises. Gating fix for the container-runtime cutover (#662). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jun 9, 2026
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
Found during the #662 live cutover: hal0/ virtual model names never route to container slots.*
The chat-normalization gate (
_normalize_chat_body) resolveshal0/chat→ the registry idqwopus3.6-27b-v2, but the container's llama-server advertises the raw GGUF basename (Qwopus3.6-27B-v2-…gguf).dispatcher.dispatch()finds no upstream advertisingqwopus3.6-27b-v2→NoRouteFound→ lemonade fall-through → 404 (lemond tries to load it from HuggingFace).Confirmed live: routing via the bare GGUF name reached the container fine, so the only gap is the name the container advertises.
Fix
Pass llama-server
--alias <registry-model-id>(frommodel_info._model_key, fallback[model].default) so the container advertises its hal0 id.resolved_command_for_slotsurfaces it too.After this:
hal0/chat→qwopus3.6-27b-v2→ matches the container's advertised alias → routes to 127.0.0.1:8102. Same forhal0/agent→chadrock-35b-ace-saber.Tests
TDD:
_render_unitemits--alias,load_syncthreads_model_key,resolved_command_for_slotincludes it. 195 passed; ruff clean.Gating fix for the container-runtime cutover. Relates to #652, #662.
🤖 Generated with Claude Code