Uh oh!
There was an error while loading. Please reload this page.
feat(coder): Gemini as bridge-default provider - #21
Merged
Conversation
Generalize the env-fallback (company-free) coder path so the bridge- default provider is configurable and prefers Gemini when a Google key is present, falling back to OpenRouter (qwen free) / Anthropic / OpenAI. Pin explicitly with MIND_DEFAULT_PROVIDER. Per-provider default model (google -> gemini-2.0-flash, openrouter -> qwen/qwen3-coder:free). BYOK (user-pref) stays the optional override and runs unmetered; bridge per-run MIND metering is unchanged. Also align env.ts MIND_AGENT_MODEL default to qwen/qwen3-coder:free (matched store.ts). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The workflow runner was hardcoded to node:22-alpine (musl). Vite/Tailwind v4 pull in lightningcss, whose musl prebuilt binary npm skips on a fresh install (optional-deps bug) — every `vite build` died with `Cannot find module lightningcss.linux-x64-musl.node` (exit 1). Default to node:22-bookworm-slim (glibc) where the gnu binary resolves cleanly; reproduced the scaffold build green on glibc. Make it overridable via MIND_WORKFLOW_IMAGE. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…rned When the in-process session cache is cold (after a restart) and several publish-path callers (post_receive + reconciler + tracker + publisher) hit loadAuthedFetchForWebId for the same WebID at once, each spent the SAME stored single-use refresh token against CSS. The first rotated it; the rest got `invalid_grant (unknown refresh_token)`, and the clobbered rotation left the stored token permanently dead — forcing a /connect re-auth (observed in prod: a refresh storm of 5+ failed refreshes on one dead session). Collapse concurrent refreshes per WebID onto one in-flight promise: the token is spent exactly once and every caller shares the resulting live session. Complements the MC-176 cache (which only helps once a session is warm). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… doesn't wipe the publish The after-upload prune compared each pod child against `kept` using the raw `ldp:contains` object. CSS can emit that object as an absolute URL, so `childRel` became a full URL that never matched the relative keys in `kept` — the prune then DELETEd every file it had just uploaded and the published site served 404s despite `publisher.done uploaded:N`. Resolve each child to absolute, re-derive the immediate segment relative to the container, and compare that. Works for both relative-slug and absolute-URL servers. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
huhn511force-pushed
the
feat/free-llm-allotment
branch
from
June 20, 2026 23:39
90663c7 to
5bccc5bCompareThe biome ci gate (#20) checks formatting too. My hand-edits to the four publish-pipeline files differed from biome's 100-col formatter, and the tracker-pod test used forbidden non-null assertions — both fail `biome ci .`. Reformat the four files and switch the test asserts to optional chaining / guarded gets. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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 freeto 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.
Generalizes the env-fallback (company-free) coder path so the bridge-default provider is configurable and prefers Gemini when a Google key is present, falling back to OpenRouter (qwen free) / Anthropic / OpenAI.
Changes
store.ts: newresolveEnvFallback()— picks the bridge-default provider (pin withMIND_DEFAULT_PROVIDER), per-provider default model (google → gemini-2.0-flash,openrouter → qwen/qwen3-coder:free, …). Widened the env-fallback summary type to any provider.env.ts: alignMIND_AGENT_MODELdefault toqwen/qwen3-coder:free(matchedstore.ts).user-pref) stays the optional override and runs unmetered; bridge per-run MIND metering unchanged.Verified
tscclean + 102 tests pass. Deployed to prod and verified in Builder: coder runsprovider=google source=env-fallback model=gemini-2.5-flash, Gemini wrote + vite-built an app end-to-end.Companion infra PR: MIND-Studio/mindpods-infra#9.
🤖 Generated with Claude Code