📊 refactor: use Parameters from Side Panel for OpenAI, Anthropic, and Custom endpoints (#4092)

* feat: openai parameters

* refactor: anthropic/bedrock params, add preset params for openai, and add azure params

* refactor: use 'compact' schemas for anthropic/openai

* refactor: ensure custom endpoints are properly recognized as valid param endpoints

* refactor: update paramEndpoints check in BaseClient.js

* chore: optimize logging by omitting modelsConfig

* refactor: update label casing in baseDefinitions combobox items

* fix: remove 'stop' model options when using o1 series models

* refactor(AnthropicClient): remove default `stop` value

* refactor: reset params on parameters change

* refactor: remove unused default parameter value map introduced in prior commit

* fix: 'min' typo for 'max' value

* refactor: preset settings

* refactor: replace dropdown for image detail with slider; remove `preventDelayedUpdate` condition from DynamicSlider

* fix: localizations for freq./pres. penalty

* Refactor maxOutputTokens to use coerceNumber in tConversationSchema

* refactor(AnthropicClient): use `getModelMaxOutputTokens`

api/app/clients/AnthropicClient.js

@@ -17,8 +17,8 @@ const {
   parseParamFromPrompt,
   createContextHandlers,
 } = require('./prompts');
+const { getModelMaxTokens, getModelMaxOutputTokens, matchModelName } = require('~/utils');
 const { spendTokens, spendStructuredTokens } = require('~/models/spendTokens');
-const { getModelMaxTokens, matchModelName } = require('~/utils');
 const { sleep } = require('~/server/utils');
 const BaseClient = require('./BaseClient');
 const { logger } = require('~/config');
@@ -120,7 +120,14 @@ class AnthropicClient extends BaseClient {
       this.options.maxContextTokens ??
       getModelMaxTokens(this.modelOptions.model, EModelEndpoint.anthropic) ??
       100000;
-    this.maxResponseTokens = this.modelOptions.maxOutputTokens || 1500;
+    this.maxResponseTokens =
+      this.modelOptions.maxOutputTokens ??
+      getModelMaxOutputTokens(
+        this.modelOptions.model,
+        this.options.endpointType ?? this.options.endpoint,
+        this.options.endpointTokenConfig,
+      ) ??
+      1500;
     this.maxPromptTokens =
       this.options.maxPromptTokens || this.maxContextTokens - this.maxResponseTokens;
 
@@ -144,17 +151,6 @@ class AnthropicClient extends BaseClient {
     this.endToken = '';
     this.gptEncoder = this.constructor.getTokenizer('cl100k_base');
 
-    if (!this.modelOptions.stop) {
-      const stopTokens = [this.startToken];
-      if (this.endToken && this.endToken !== this.startToken) {
-        stopTokens.push(this.endToken);
-      }
-      stopTokens.push(`${this.userLabel}`);
-      stopTokens.push('<|diff_marker|>');
-
-      this.modelOptions.stop = stopTokens;
-    }
-
     return this;
   }
 

api/app/clients/BaseClient.js

@@ -3,7 +3,7 @@ const fetch = require('node-fetch');
 const {
   supportsBalanceCheck,
   isAgentsEndpoint,
-  paramEndpoints,
+  isParamEndpoint,
   ErrorTypes,
   Constants,
   CacheKeys,
@@ -588,7 +588,10 @@ class BaseClient {
 
     if (typeof completion === 'string') {
       responseMessage.text = addSpaceIfNeeded(generation) + completion;
-    } else if (Array.isArray(completion) && paramEndpoints.has(this.options.endpoint)) {
+    } else if (
+      Array.isArray(completion) &&
+      isParamEndpoint(this.options.endpoint, this.options.endpointType)
+    ) {
       responseMessage.text = '';
       responseMessage.content = completion;
     } else if (Array.isArray(completion)) {

api/app/clients/OpenAIClient.js

@@ -1295,6 +1295,7 @@ ${convo}
 
       if (modelOptions.stream && /\bo1\b/i.test(modelOptions.model)) {
         delete modelOptions.stream;
+        delete modelOptions.stop;
       }
 
       if (modelOptions.stream) {

api/server/controllers/AskController.js

@@ -16,7 +16,12 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
     overrideParentMessageId = null,
   } = req.body;
 
-  logger.debug('[AskController]', { text, conversationId, ...endpointOption });
+  logger.debug('[AskController]', {
+    text,
+    conversationId,
+    ...endpointOption,
+    modelsConfig: endpointOption.modelsConfig ? 'exists' : '',
+  });
 
   let userMessage;
   let userMessagePromise;

api/server/controllers/EditController.js

@@ -25,6 +25,7 @@ const EditController = async (req, res, next, initializeClient) => {
     isContinued,
     conversationId,
     ...endpointOption,
+    modelsConfig: endpointOption.modelsConfig ? 'exists' : '',
   });
 
   let userMessage;

api/utils/tokens.js

@@ -123,7 +123,16 @@ const modelMaxOutputs = {
   system_default: 1024,
 };
 
+const anthropicMaxOutputs = {
+  'claude-3-haiku': 4096,
+  'claude-3-sonnet': 4096,
+  'claude-3-opus': 4096,
+  'claude-3.5-sonnet': 8192,
+  'claude-3-5-sonnet': 8192,
+};
+
 const maxOutputTokensMap = {
+  [EModelEndpoint.anthropic]: anthropicMaxOutputs,
   [EModelEndpoint.azureOpenAI]: modelMaxOutputs,
   [EModelEndpoint.openAI]: modelMaxOutputs,
   [EModelEndpoint.custom]: modelMaxOutputs,