Repository navigation
send GLM turns only to Together, Fireworks, DeepInfra or Baseten, with zero retention - #223
Merged
Merged
Conversation
… only where they keep nothing
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The built-in agent's GLM 5.3 turns no longer go to whichever of OpenRouter's ~46 GLM hosts is cheapest. They go only to Together, Fireworks, DeepInfra or Baseten, and only to their zero-retention endpoints (
zdr: true). Before, onlydata_collection: 'deny'was set, which allows hosts that keep prompts for abuse checks. Most requests went to the cheapest hosts, small 4-bit (fp4) providers such as Wafer and InferenceNet, and hosts based in China or Indonesia were allowed too. Claude's host settings are unchanged (Anthropic's own API isn't zero-retention on OpenRouter). Within the four, OpenRouter still favours the cheapest, so DeepInfra (fp4) will probably get most GLM requests.Open workspace in Conductor