Repository navigation
feat: the model picker marks LiteLLM gateway models, and Pi leaves the LiteLLM launcher - #23
Merged
Merged
Conversation
Thread transfer impact✅ Thread transfer remains within every enforced ceiling.
Baseline: Scenario and decoded snapshot size10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.
Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed. |
The LiteLLM launcher now supports Claude, Codex, and OpenCode only. Pi had no request-tag channel and was the only user of homeEnvironment, so both go. A Pi instance saved with the launcher on shows as unavailable, and its settings card still shows the switch so the user can turn it off. Claude Opus 5.5 via Claude Code.
A LiteLLM Codex or OpenCode instance lists the gateway's models beside the harness's own catalog, and the picker showed no difference between them. The server now sets `gateway: "litellm"` on each model whose slug is in the gateway's model list (OpenCode: `litellm/<id>` only, because its other providers stay direct). The web and mobile model pickers and the settings model list show a LiteLLM label on those models. Claude Opus 5.5 via Claude Code.
hyrmedia-hector
force-pushed
the
feat/litellm-gateway-models
branch
from
November 5, 2026 03:38
4332d5b to
bd1d53e
Compare
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
In a LiteLLM Codex or OpenCode provider, the model picker showed gateway models and the harness's own models the same way, so you could not tell which ones the gateway serves. Admins rename gateway models, so the name does not tell you either. Pi also no longer needs to run through LiteLLM.
gateway: "litellm"on each model that is in the gateway's model list (lite models list). This is the same signal Claude Code uses for its "From gateway" rows: where the model came from, not its name. For OpenCode, onlylitellm/<id>models count, becauselite opencodeadds the gateway as providerlitellmand the other providers stay direct. The web model picker, the model list in Settings > Providers, and the mobile model picker show a LiteLLM label on these models.Proof:
vp run check:changed origin/mainpasses (format, lint, knip, typecheck, tests). New test cases cover the gateway marking. In a dev client with a stublite, the LiteLLM Claude modelsteam-sonnet(an admin alias) andclaude-sonnet-4-5show the LiteLLM label in the picker and in settings, and the built-in Claude models do not. Mobile is typecheck only: this host cannot build the iOS client (no CocoaPods) or start the Android emulator (no Java).Reviewed with sol-loop: 1 round with gpt-6.1-sol. Sol verdict: approved
Created with Claude Opus 5.5 in Claude Code (T3 Code).
Open in [code]smith · Need help on this PR? Tag
@codesmith-botwith what you need.