fix(geminiGoogle): Unable to submit request because it has a topK value of 64 but the supported range is from 1 (inclusive) to 41 (exclusive).

Add option to set topK and topP in client settings.

Author: GanymedeNil

CHANGELOG.md

@@ -2,6 +2,14 @@
 
 ## [Unreleased]
 
+### Added
+
+- Option to set top K and top P in Gemini Google client settings.
+
+### Fixed
+
+- Unable to submit request to Gemini Google because it has a topK value of 64 but the supported range is from 1 (inclusive) to 41 (exclusive).
+
 ## [2.7.0] - 2024-11-01
 
 ### Added

gradle.properties

@@ -3,7 +3,7 @@ pluginGroup = com.github.blarc
 pluginName = AICommits
 pluginRepositoryUrl = https://github.com/Blarc/ai-commits-intellij-plugin
 # SemVer format -> https://semver.org
-pluginVersion = 2.7.0
+pluginVersion = 2.7.1
 
 # https://plugins.jetbrains.com/docs/intellij/build-number-ranges.html
 pluginSinceBuild = 233

src/main/kotlin/com/github/blarc/ai/commits/intellij/plugin/AICommitsExtensions.kt

@@ -50,6 +50,20 @@ fun ValidationInfoBuilder.isInt(value: String): ValidationInfo? {
     }
 }
 
+fun ValidationInfoBuilder.isDouble(value: String): ValidationInfo? {
+    if (value.isBlank()){
+        return null
+    }
+
+    value.toDoubleOrNull().let {
+        if (it == null) {
+            return error(message("validation.double"))
+        } else {
+            return null
+        }
+    }
+}
+
 // Adds emptyText method to all cells that contain a component that implements ComponentWithEmptyText class
 fun <T>Cell<T>.emptyText(emptyText: String) : Cell<T> where T : JComponent, T : ComponentWithEmptyText {
     this.component.emptyText.text = emptyText

src/main/kotlin/com/github/blarc/ai/commits/intellij/plugin/settings/clients/geminiGoogle/GeminiGoogleClientConfiguration.kt

@@ -20,6 +20,8 @@ class GeminiGoogleClientConfiguration : LLMClientConfiguration(
     var tokenIsStored: Boolean = false
     @Transient
     var token: String? = null
+    var topK = 40
+    var topP = 0.95
 
     companion object {
         const val CLIENT_NAME = "Gemini Google"
@@ -52,6 +54,8 @@ class GeminiGoogleClientConfiguration : LLMClientConfiguration(
         copy.temperature = temperature
         copy.tokenIsStored = tokenIsStored
         copy.token = token
+        copy.topK = topK
+        copy.topP = topP
         return copy
     }
 

src/main/kotlin/com/github/blarc/ai/commits/intellij/plugin/settings/clients/geminiGoogle/GeminiGoogleClientPanel.kt

@@ -3,19 +3,21 @@ package com.github.blarc.ai.commits.intellij.plugin.settings.clients.geminiGoogl
 import GeminiGoogleClientService
 import com.github.blarc.ai.commits.intellij.plugin.AICommitsBundle.message
 import com.github.blarc.ai.commits.intellij.plugin.emptyText
+import com.github.blarc.ai.commits.intellij.plugin.isDouble
+import com.github.blarc.ai.commits.intellij.plugin.isInt
 import com.github.blarc.ai.commits.intellij.plugin.settings.clients.LLMClientPanel
 import com.intellij.ui.components.JBPasswordField
-import com.intellij.ui.dsl.builder.Align
-import com.intellij.ui.dsl.builder.Panel
-import com.intellij.ui.dsl.builder.bindText
-import com.intellij.ui.dsl.builder.panel
+import com.intellij.ui.components.JBTextField
+import com.intellij.ui.dsl.builder.*
 
 class GeminiGoogleClientPanel private constructor(
     private val clientConfiguration: GeminiGoogleClientConfiguration,
     val service: GeminiGoogleClientService
 ) : LLMClientPanel(clientConfiguration) {
 
     private val tokenPasswordField = JBPasswordField()
+    private val topKTextField = JBTextField()
+    private val topPTextField = JBTextField()
 
     constructor(configuration: GeminiGoogleClientConfiguration): this(configuration, GeminiGoogleClientService.getInstance())
 
@@ -24,9 +26,39 @@ class GeminiGoogleClientPanel private constructor(
         tokenRow()
         modelIdRow()
         temperatureRow()
+        topKRow()
+        topPRow()
         verifyRow()
     }
 
+    private fun Panel.topKRow() {
+        row {
+            label(message("settings.geminiGoogle.topK"))
+                .widthGroup("label")
+            cell(topKTextField)
+                .bindIntText(clientConfiguration::topK)
+                .align(Align.FILL)
+                .validationOnInput { isInt(it.text) }
+                .resizableColumn()
+            contextHelp(message("settings.geminiGoogle.topK.comment"))
+                .align(AlignX.RIGHT)
+        }
+    }
+
+    private fun Panel.topPRow() {
+        row {
+            label(message("settings.geminiGoogle.topP"))
+                .widthGroup("label")
+            cell(topPTextField)
+                .bindText({ clientConfiguration.topP.toString() }, { s -> clientConfiguration.topP = s.toDouble() })
+                .align(Align.FILL)
+                .validationOnInput { isDouble(it.text) }
+                .resizableColumn()
+            contextHelp(message("settings.geminiGoogle.topP.comment"))
+                .align(AlignX.RIGHT)
+        }
+    }
+
     private fun Panel.tokenRow() {
         row {
             label(message("settings.geminiGoogle.token"))

src/main/kotlin/com/github/blarc/ai/commits/intellij/plugin/settings/clients/geminiGoogle/GeminiGoogleClientService.kt

@@ -29,6 +29,8 @@ class GeminiGoogleClientService(private val cs: CoroutineScope) : LLMClientServi
             .apiKey(token)
             .modelName(client.modelId)
             .temperature(client.temperature.toDouble())
+            .topK(client.topK)
+            .topP(client.topP)
             .build()
     }
 

src/main/resources/messages/AiCommitsBundle.properties

@@ -38,6 +38,7 @@ settings.prompt.name=Name
 settings.prompt.content=Content
 validation.required=This value is required.
 validation.integer=Value should be an integer.
+validation.double=Value should be double.
 validation.temperature=Temperature should be between 0 and 2.
 settings.prompt.comment=<ul>\
   <li>Customize your prompt with variables: {locale}, {diff}, {branch}, {hint}, {taskId}, {taskSummary} and {taskDescription}.</li>\
@@ -94,6 +95,17 @@ settings.gemini.location=Location
 settings.geminiGoogle.token=API key
 settings.geminiGoogle.token.comment=You can get your token <a href="https://aistudio.google.com/app/u/2/apikey">here.</a>.
 settings.geminiGoogle.token.example=BZcxxx-xxxx-xxxxx-xxxxxxxxxxxxxxxxxxxxx
+settings.geminiGoogle.topK=Top K
+settings.geminiGoogle.topK.comment=The Top K parameter changes how the model selects tokens for output.\
+  A Top K of 1 means the selected token is the most probable among all the tokens in the model's vocabulary (also called greedy decoding),\
+  while a Top K of 3 means that the next token is selected from among the 3 most probable using the temperature.\
+  For each token selection step, the Top K tokens with the highest probabilities are sampled.\
+  Tokens are then further filtered based on topP with the final token selected using temperature sampling.
+settings.geminiGoogle.topP=Top P
+settings.geminiGoogle.topP.comment=The Top P parameter changes how the model selects tokens for output.\
+  Tokens are selected from the most to least probable until the sum of their probabilities equals the Top P value.\
+  For example, if tokens A, B, and C have a probability of 0.3, 0.2, and 0.1 and the Top P value is 0.5,\
+  then the model will select either A or B as the next token by using the temperature and exclude C as a candidate.
 
 settings.anthropic.token.example=sk-ant-api03-TTz_qsxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
 settings.anthropic.token.comment=You can get your token <a href="https://console.anthropic.com/settings/keys">here.</a>
@@ -113,3 +125,4 @@ settings.huggingface.token.example=hf_fKASPPYLkasgjasKwpSnAASRdasdCdAsddsASSDF
 settings.huggingface.maxNewTokens=Max new tokens
 settings.huggingface.waitForModel=Wait for model
 settings.huggingface.waitModel.comment=When a model is warm, it is ready to be used, and you will get a response relatively quickly. However, some models are cold and need to be loaded before they can be used. In that case, you will get a 503 error. Rather than doing many requests until it is loaded, you can wait for the model to be loaded.
+