Repository navigation
chore(agent): update the T3 provider runtime to upstream main - #2586
Conversation
Pin @t3tools/provider-runtime to the vite-hub/t3code sync of upstream main (4a1a008). The runtime now uses Effect 4.0.5, so the lockfile drops 4.0.1.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
Caution
This PR records an integrity hash that does not match the tarball currently served for the new @t3tools/contracts URL, so a clean frozen install fails before the workspace can install.
Reviewed changes
The dependency catalog and lockfile update for the T3 provider runtime, its contracts package, and the transitive Effect version were reviewed.
- Provider runtime pin — moves
@t3tools/provider-runtimeand@t3tools/contractsto the4a1a0087b30b479aa78871bc789ac9a766afffcapkg.pr.new build. - Effect resolution — replaces the runtime graph's
effect@4.0.1entry witheffect@4.0.5, matching the existing workspace catalog.
Note: 1 inline comment(s) dropped because they did not anchor to lines inside the PR diff:
pnpm-lock.yaml:42-44(RIGHT) — line 44 (RIGHT) is not inside a diff hunk
gpt-5.6-sol | 𝕏
Latest result Pull request merged. Final answersSession Pull request merged. Session Waiting for GitHub merge request 98bfe400-5d58-40cf-a2ef-7e55facb8fd5. Verified current HEAD and both recent Pullfrog reviews. The initial Pullfrog P1 contracts integrity finding was fixed in 5799dfc and the follow-up review confirms frozen install consistency; no valid P0/P1 findings remain. Focused lint passed with existing warnings, typecheck was attempted, and no repair is needed. |
pkg.pr.new rebuilt the fork's 4a1a008 packages when that commit reached the fork's main. The contracts tarball changed bytes, so CI rejected the old integrity.
@vite-hub/agent
@vite-hub/auth
@vite-hub/blob
@vite-hub/browser
@vite-hub/channels
@vite-hub/cli
@vite-hub/connections
@vite-hub/content
@vite-hub/database
@vite-hub/email
@vite-hub/env
@vite-hub/forge
@vite-hub/kv
@vite-hub/markdown-template
@vite-hub/presets
@vite-hub/queue
@vite-hub/rate-limit
@vite-hub/realtime
@vite-hub/runtime
@vite-hub/sandbox
@vite-hub/sandbox-task
@vite-hub/schedule
@vite-hub/shell
@vite-hub/source
@vite-hub/ui
vite-hub
@vite-hub/workflow
@vite-hub/workspace
commit: |
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
The incremental change since the prior Pullfrog review updates the contracts tarball integrity and was validated with a clean frozen install.
- Corrected contracts integrity — replaced the stale
@t3tools/contractshash with the integrity for the republished tarball at the existing4a1a0087b30b479aa78871bc789ac9a766afffcaURL. - Confirmed lockfile consistency —
corepack pnpm install --frozen-lockfilenow completes successfully without modifying the lockfile.
gpt-5.6-sol | 𝕏

The T3 provider runtime that Provider Agents use was 1082 upstream commits behind. The fork's daily sync had failed on merge conflicts since at least 2026-10-13.
The fork now has upstream main merged, with the embedded runtime ported to upstream's new
provider-corepackage (vite-hub/t3code#12). This PR moves the@t3tools/provider-runtimepin to that build. The runtime now uses Effect 4.0.5, so the lockfile drops Effect 4.0.1. I edited the lockfile entries by hand: 2 changed URLs, 2 new integrities, and the Effect version.The fork sync also fixes one problem: the runtime loaded the Claude SDK when a Codex-only host imported it.
This is a stopgap. The pin is still a pkg.pr.new URL, which pnpm 12 rejects. The real fix for #2545 is in a follow-up PR.
Verified
pnpm install --frozen-lockfileresumeCursorand recalls the earlier turn.provider-agent,box, andnuxt-runtime-outputtests give the same results as onmain.providers.test.ts(5 to 18 failures depending on machine load, a different set each run),babysitter-preset(8 to 13 per run, and it mocks the runtime), andnuxt-runtime-output(1 test).🤖 Generated with Claude Code