Add Cloudflare AI Gateway and Workers AI providers (v1.12.1) - #3015
Draft
pridemusvaire wants to merge 145 commits into
Draft
pridemusvaire wants to merge 145 commits into
pridemusvaire wants to merge 145 commits into
Conversation
Promote v0.0.2 to production
Promote v0.0.3 to production
Promote v0.0.4 to production
Promote v0.1.0 to production
Promote v0.1.1 to production
Promote v0.2.0 to production
Promote v0.3.0 to production
Promote v0.3.1 to production
Promote v0.4.0 to production
Promote v0.4.1 to production
…CodeInc#299) SlackSetupExperience's intro screen owns the step action buttons, and it is skipped when the Slack config is already saved (savedSatisfied). But StepAuthEnvVars's providerOwnsActions check was missing the matching !savedSatisfied clause, so revisiting the auth step with saved Slack credentials rendered the value form with no action button at all, stranding the user on the page. Align providerOwnsActions with the intro guard and add a regression test for the saved-Slack revisit path. Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Hotfix-only cut from main with the RooCodeInc#299 Slack setup Continue fix.
…-continue [Chore] Release Roomote 0.4.2 hotfix for saved Slack continue
Drop the reverted analytics Model group-by line and document notification destination and homepage dump-flash fixes now included on release/v0.5.0.
Merge latest develop (RooCodeInc#370, RooCodeInc#371), document the Blaxel Compose --wait fix on the existing 0.5.0 notes, remove the shipping pending changeset, and quiet residual oxlint/ESLint friction from that merge.
Promote v0.5.0 to production
Promote v0.6.0 to production
Promote v0.7.0 to production
Promote v0.7.1 to production
Promote v0.8.0 to production
Promote v0.8.1 to production
Promote v0.9.0 to production
Promote v1.5.0 to production
Promote v1.5.1 to production
Promote v1.6.0 to production
Promote v1.7.0 to production
Promote v1.8.0 to production
Co-authored-by: Roomote <roomote@roomote.dev>
…odeInc#2559) OpenAI gpt-5.x models fill every optional list_skills argument, so an exact-name lookup arrived with both scope IDs set and was rejected by the bridge schema. The bare catch then reported 'The requested skill catalog is unavailable.' on every call, which made models conclude the catalog was down. - Treat null as absent for list_skills and load_skill on every Fast host, not only inside task sandboxes. - Prefer environmentId when both scope IDs are given, and ignore a continuation offset without an exact name, instead of rejecting the call. - Report the zod issues for invalid arguments and log other failures. - Degrade a failing Settings or repository source to a catalog warning so packaged and instance skills still load. Co-authored-by: Matt Rubens <2600+mrubens@users.noreply.github.com>
…ence-auth-1.8.1 [Fix] Release Roomote 1.8.1 for Azure inference and Fast skills
…el-history-1.8.2 Hotfix Roomote 1.8.2: restore Slack channel history reads
Reviewed resolution: 614b6ad Review PR: RooCodeInc#2715
Promote v1.9.0 to production
Promote v1.9.1 to production
Promote v1.9.2 to production
Promote v1.9.3 to production
Promote v1.9.4 to production
Promote v1.10.0 to production
Promote v1.11.0 to production
Promote v1.12.0 to production
Promote v1.12.1 to production
Keeps the two providers independent on the existing inference-gateway path, and maps GLM 5.3 / 5.3 Flash on the Cloudflare catalogs now that Workers AI hosts them.
Contributor
|
1 issue outstanding. See task
Reviewed 4088626 |
4 tasks
![roomote-community[bot]](https://avatars.githubusercontent.com/u/211522643?s=60&v=4){kind=small}
Helper inference skipped the Cloudflare overlay whenever OPENCODE_CONFIG_CONTENT was set, so Fast/custom-config deployments kept OpenCode's default AI Gateway catalog instead of Roomote's /ai/v1 credentials.
| ) | ||
| .map(toBedrockMantleRuntimeModelId); | ||
| .map((modelId) => | ||
| rewriteCloudflareOpenCodeModelId(toBedrockMantleRuntimeModelId(modelId)), |
Contributor
There was a problem hiding this comment.
mergeBedrockRegistrationsIntoConfigContent rewrites this role ID to cloudflare-ai-gateway/@cf/..., but the subsequent mergeReasoningIntoConfigContent still reads the raw R_*_MODEL values. For native Fast sessions with custom config and a cloudflare-ai-gateway/workers-ai/@cf/... model, it writes reasoningEffort under models["workers-ai/@cf/..."] while both the registered provider and prompt use @cf/...; the configured role reasoning is therefore ignored. Apply the same Cloudflare rewrite in mergeReasoningIntoConfigContent (and add the custom-config Fast case).
This branch has not been deployed
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
4 participants
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.
Summary
Rebases the Cloudflare AI Gateway and Workers AI inference providers onto Roomote v1.12.1.
This is the same overlay as #2809 / #1276, now on the current release so self-hosted deploys can pull 1.10–1.12 (including #2952 JOB_AUTH on bullmq) without dropping Cloudflare.
The two providers stay independent on the existing inference-gateway path. GLM 5.3 / 5.3 Flash stay mapped on the Cloudflare catalogs.
Test plan
@roomote/typesinference-gateway, opencode-provider-config, model-provider-config, task-models (170 tests)@roomote/cloud-agentsopencode-runtime + non-task-provider-usage (118 tests)@roomote/workeragent-home (90 tests)@roomote/apiinference-gateway (79 tests)v1.12.1-cf1276, keep the Cloudflare Tunnel override, confirm/healthand Workers AI credentials still resolveSupersedes the v1.9.4 overlay in #2809.