feat(neuralwatt): expose full GLM 5.2 reasoning effort scale - #2645
Merged
rekram1-node merged 1 commit intoJun 17, 2026
Merged
Conversation
GLM-5.2 accepts the OpenAI-standard reasoning_effort field and supports a wider depth range than the three levels previously advertised. Per the Neuralwatt chat-completions docs [1], the gateway accepts and normalizes the full scale: minimal -> skips the reasoning phase entirely (eq enable_thinking: false) low -> mapped to high medium -> mapped to high high -> enhanced reasoning (balanced) xhigh -> mapped to max (deepest; best for math/planning/agentic tasks) The provider's thinkingLevelMap (pi-neuralwatt-provider/patch.json) already exposes all five pi tiers, so mirror that here by adding minimal and xhigh to the effort values for glm-5.2. [1] https://portal.neuralwatt.com/docs/api/chat-completions
Contributor
Author
|
@rekram1-node i got another one papi 📦 |
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Follow-up to #2636 (comment). GLM 5.2 accepts the OpenAI-standard
reasoning_effortfield and supports a wider depth range than the three levels (low,medium,high) that were advertised in the merged model file. This exposes the full effort scale.Context
Per the Neuralwatt chat-completions docs, the gateway accepts the full OpenAI-style
reasoning_effortscale and normalizes it onto GLM-5.2's two native levels (highandmax):minimalenable_thinking: falselowhighmediumhighhighhighxhighmaxreasoning_effortis unset)The pi provider extension (
pi-neuralwatt-provider/patch.json) already exposes all five pi reasoning tiers viathinkingLevelMap(minimal,low,medium,high,xhigh, withxhighmapped tomax). Themodels.devcatalog was the side missing the extra efforts.Changes
providers/neuralwatt/models/glm-5.2.tomlreasoning_optionseffort.values:["low", "medium", "high"]→["minimal", "low", "medium", "high", "xhigh"]No other GLM-5.2 fields change. The other neuralwatt models already match the provider data and are unaffected.
Validation
bun validatepasses (exit 0); the generatedglm-5.2entry now listsminimalandxhighin itsreasoning_options.values.