Repository navigation
fix(autocomplete): switch default FIM model to qwen3.6-plus (qwen3.5-plus removed from Go) — release 0.7.10 - #267
Merged
Merged
Conversation
…plus removed from Go) The inline-suggestions default qwen3.5-plus was removed from the Go gateway, so every FIM request returned HTTP 400 'Model is unavailable' which the engine logs silently — ghost text never appeared and the budget looked like the culprit. Verified live (dummy-key probe, gateway validates session -> model -> credential): qwen3.5-plus rejected before auth, qwen3.6-plus passes model validation. Default + package.json enum + README updated; enum also drops deepseek-v4-flash (removed too) for deepseek-v4.1-flash. Probe script refreshed and now sends the required x-opencode-session. Existing installs must switch manually once: "opencodego.inlineSuggestionsModel": "qwen3.6-plus"
… synced Real-key probe confirms qwen3.6-plus: HTTP 200, 1459ms TTFB, reasoningChars=0 (on par with the removed default's 1518ms) and the suffix field tolerated — measurements added to doc 15, doc 113, devlog, and the CHANGELOG entry. Doc sync per documentation protocol: feature doc 15 (canonical inline-suggestions reference) updated to the new default with the old probe table preserved as history; doc 100's open caveat annotated as resolved; devlog session entry + refreshed header. Version bumped to 0.7.10 and [Unreleased] merged into the 0.7.10 changelog section.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
Inline suggestions went completely silent — reported via LinkedIn by Yannick Huebner: "not working or suggesting anything" with
opencodego.inlineSuggestions: true, even with a valid Go budget. First hypothesis (budget exhausted, no fallback path) was wrong.Root cause
DEFAULT_INLINE_MODELwasqwen3.5-plus. OpenCode removed that model from the Go gateway, so every FIM request returned HTTP 400Model is unavailable— which the completion engine logs to theOpenCode Completionsoutput channel and renders as nothing. The registry's deprecation filter never applied here: FIM sends the raw setting value straight to/chat/completions, bypassing the picker. This was the landmine doc 100 flagged on 2026-09-08 ("flagged, unchanged") — upstream finally dropped the model.Evidence (live, 2026-10-20)
Gateway validates in the order session → model → credential, so a dummy-key probe distinguishes the failure classes without a paid key:
GET zen/go/v1/modelsqwen3.5-plusabsent (Qwen on Go:qwen3.6-plus,qwen3.7-plus,qwen3.7-max,qwen3.8-flash,qwen3.8-max)GET zen/v1/models(Zen)qwen3.5-plusstill listed — removal is Go-only (suggestions are Go-only too)qwen3.5-plus, dummy keyModel is unavailable— before authqwen3.6-plus, dummy keyInvalid credential— passes model validationqwen3.6-plus, real keyreasoningChars=0— on par with the old default's 1518msChanges
src/config.ts:DEFAULT_INLINE_MODEL→qwen3.6-plus(+ comment naming why)package.json: settings default/enum updated — also drops the equally-removeddeepseek-v4-flashfordeepseek-v4.1-flashsrc/test/config.test.ts: default assertion updatedREADME.md: inline-suggestions section + manual-switch caveatscripts/probe-completion-latency.ts: candidates refreshed + requiredx-opencode-sessionheaderdocs/issues/113, feature doc 15 updated (old probe table kept as history), doc 100 caveat annotated resolved, devlog entry, CHANGELOG[Unreleased]merged into the 0.7.10 changelog sectionNo settings auto-migration (same call as doc 112) — users on the old default switch manually once:
"opencodego.inlineSuggestionsModel": "qwen3.6-plus".Verification
npm run compile— pass; unit tests 0 failvsce package→ 0.7.10 VSIX built and installed locally (ltmoerdani.opencode-copilot-chat@0.7.10)