Merge pull request #453 from voideditor/model-selection

Fix gemini OpenRouter
This commit is contained in:
Andrew Pareles 2025-05-05 17:17:57 -07:00 committed by GitHub
commit fcd445d875
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 23 additions and 11 deletions

View file

@ -276,6 +276,8 @@ const SimpleModelSettingsDialog = ({
onClose();
};
const sourcecodeOverridesLink = `https://github.com/voideditor/void/blob/main/src/vs/workbench/contrib/void/common/modelCapabilities.ts#L144-L168`
return (
<div // Backdrop
className="fixed inset-0 bg-black/50 flex items-center justify-center z-[9999999]"
@ -322,12 +324,12 @@ const SimpleModelSettingsDialog = ({
{/* override toggle */}
<div className="flex items-center gap-2 mb-4">
<VoidSwitch size='xs' value={overrideEnabled} onChange={setOverrideEnabled} />
<span className="text-void-fg-3 text-sm">Override model defaults</span>
<span className="text-void-fg-3 text-sm">Override model defaults</span>
</div>
{/* Informational link */}
{overrideEnabled && <div className="text-sm text-void-fg-3 mb-4">
<ChatMarkdownRender string={"See the [sourcecode](https://github.com/voideditor/void/blob/d125d8698bf6ccd46c9367c1445e4adfe9aa2c1c/src/vs/workbench/contrib/void/common/modelCapabilities.ts#L144C1-L168C1) for a reference on how to set this JSON (advanced)."} chatMessageLocation={undefined} />
<ChatMarkdownRender string={`See the [sourcecode](${sourcecodeOverridesLink}) for a reference on how to set this JSON (advanced).`} chatMessageLocation={undefined} />
</div>}
<textarea

View file

@ -141,13 +141,15 @@ export const defaultModelsOfProvider = {
export type VoidStaticModelInfo = { // not stateful
// for some examples, see openAIModelOptions and anthropicModelOptions below.
// Void uses the information below to know how to handle each model.
// for some examples, see openAIModelOptions and anthropicModelOptions (below).
contextWindow: number; // input tokens
reservedOutputTokenSpace: number | null; // reserve this much space in the context window for output, defaults to 4096 if null
supportsSystemMessage: false | 'system-role' | 'developer-role' | 'separated'; // typically you should use 'system-role'. 'separated' means the system message is passed as a separate field (e.g. anthropic)
specialToolFormat?: 'openai-style' | 'anthropic-style' | 'gemini-style', // typically you should use 'openai-style'. null means "can't call tools by default", and asks the LLM to output XML in agent mode
supportsFIM: boolean;
supportsFIM: boolean; // whether the model was specifically designed for autocomplete or "FIM" ("fill-in-middle" format)
// reasoning options
reasoningCapabilities: false | {
@ -166,6 +168,7 @@ export type VoidStaticModelInfo = { // not stateful
// the only other field related to reasoning is "providerReasoningIOSettings", which varies by provider.
};
// --- below is just informative, not used in sending / receiving, cannot be customized in settings ---
cost: {
input: number;
@ -349,7 +352,7 @@ const openSourceModelOptions_assumingOAICompat = {
// keep modelName, but use the fallback's defaults
const extensiveModelOptionsFallback: VoidStaticProviderInfo['modelOptionsFallback'] = (modelName, fallbackKnownValues) => {
const extensiveOAICompatModelOptionsFallback: VoidStaticProviderInfo['modelOptionsFallback'] = (modelName, fallbackKnownValues) => {
const lower = modelName.toLowerCase()
@ -1015,32 +1018,32 @@ export const ollamaRecommendedModels = ['qwen2.5-coder:1.5b', 'llama3.1', 'qwq',
const vLLMSettings: VoidStaticProviderInfo = {
// reasoning: OAICompat + response.choices[0].delta.reasoning_content // https://docs.vllm.ai/en/stable/features/reasoning_outputs.html#streaming-chat-completions
providerReasoningIOSettings: { output: { nameOfFieldInDelta: 'reasoning_content' }, },
modelOptionsFallback: (modelName) => extensiveModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' } }),
modelOptionsFallback: (modelName) => extensiveOAICompatModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' } }),
modelOptions: {}, // TODO
}
const lmStudioSettings: VoidStaticProviderInfo = {
providerReasoningIOSettings: { output: { needsManualParse: true }, },
modelOptionsFallback: (modelName) => extensiveModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' }, contextWindow: 4_096 }),
modelOptionsFallback: (modelName) => extensiveOAICompatModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' }, contextWindow: 4_096 }),
modelOptions: {}, // TODO
}
const ollamaSettings: VoidStaticProviderInfo = {
// reasoning: we need to filter out reasoning <think> tags manually
providerReasoningIOSettings: { output: { needsManualParse: true }, },
modelOptionsFallback: (modelName) => extensiveModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' } }),
modelOptionsFallback: (modelName) => extensiveOAICompatModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' } }),
modelOptions: ollamaModelOptions,
}
const openaiCompatible: VoidStaticProviderInfo = {
// reasoning: we have no idea what endpoint they used, so we can't consistently parse out reasoning
modelOptionsFallback: (modelName) => extensiveModelOptionsFallback(modelName),
modelOptionsFallback: (modelName) => extensiveOAICompatModelOptionsFallback(modelName),
modelOptions: {},
}
const liteLLMSettings: VoidStaticProviderInfo = { // https://docs.litellm.ai/docs/reasoning_content
providerReasoningIOSettings: { output: { nameOfFieldInDelta: 'reasoning_content' } },
modelOptionsFallback: (modelName) => extensiveModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' } }),
modelOptionsFallback: (modelName) => extensiveOAICompatModelOptionsFallback(modelName, { downloadable: { sizeGb: 'not-known' } }),
modelOptions: {}, // TODO
}
@ -1193,7 +1196,14 @@ const openRouterSettings: VoidStaticProviderInfo = {
},
modelOptions: openRouterModelOptions_assumingOpenAICompat,
// TODO!!! send a query to openrouter to get the price, etc.
modelOptionsFallback: (modelName) => extensiveModelOptionsFallback(modelName),
modelOptionsFallback: (modelName) => {
const res = extensiveOAICompatModelOptionsFallback(modelName)
// openRouter does not support gemini-style, use openai-style instead
if (res?.specialToolFormat === 'gemini-style') {
res.specialToolFormat = 'openai-style'
}
return res
},
}