🔧 refactor: Centralize Default Agent Capabilities and Better Logging (#7598)

* refactor: Simplify grid column calculation in SourcesGroup component

* refactor: Centralize default agent capabilities and simplify capability assignment

* Edge case: use defined/fallback capabilities for ephemeral agents when the "agents" endpoint is not enabled

* refactor: consolidate gemini 2 vision check

* feat: enhance capability check logging for agents

* chore: update librechat-data-provider version to 0.7.86

* refactor: import default agent capabilities for enhanced capability management

* chore: standardize quotes in error message check for consistency

* fix: improve error logging both client and api-side for mistral ocr upload errors

* ci: update error handling in MistralOCR tests to use specific error message

api/server/routes/files/files.js

@@ -283,6 +283,10 @@ router.post('/', async (req, res) => {
       message += ': ' + error.message;
     }
 
+    if (error.message?.includes('Invalid file format')) {
+      message = error.message;
+    }
+
     // TODO: delete remote file if it exists
     try {
       await fs.unlink(req.file.path);

api/server/services/Files/MistralOCR/crud.js

@@ -47,7 +47,6 @@ async function uploadDocumentToMistral({
     })
     .then((res) => res.data)
     .catch((error) => {
-      logger.error('Error uploading document to Mistral:', error.message);
       throw error;
     });
 }
@@ -217,8 +216,16 @@ const uploadMistralOCR = async ({ req, file, file_id, entity_id }) => {
       images,
     };
   } catch (error) {
-    const message = 'Error uploading document to Mistral OCR API';
-    throw new Error(logAxiosError({ error, message }));
+    let message = 'Error uploading document to Mistral OCR API';
+    const detail = error?.response?.data?.detail;
+    if (detail && detail !== '') {
+      message = detail;
+    }
+
+    const responseMessage = error?.response?.data?.message;
+    throw new Error(
+      `${logAxiosError({ error, message })}${responseMessage && responseMessage !== '' ? ` - ${responseMessage}` : ''}`,
+    );
   }
 };
 

api/server/services/Files/MistralOCR/crud.spec.js

@@ -124,13 +124,7 @@ describe('MistralOCR Service', () => {
           fileName: 'test.pdf',
           apiKey: 'test-api-key',
         }),
-      ).rejects.toThrow();
-
-      const { logger } = require('~/config');
-      expect(logger.error).toHaveBeenCalledWith(
-        expect.stringContaining('Error uploading document to Mistral:'),
-        expect.any(String),
-      );
+      ).rejects.toThrow(errorMessage);
     });
   });
 

api/server/services/ToolService.js

@@ -5,6 +5,7 @@ const { Calculator } = require('@langchain/community/tools/calculator');
 const { tool: toolFn, Tool, DynamicStructuredTool } = require('@langchain/core/tools');
 const {
   Tools,
+  Constants,
   ErrorTypes,
   ContentTypes,
   imageGenTools,
@@ -14,6 +15,7 @@ const {
   ImageVisionTool,
   openapiToFunction,
   AgentCapabilities,
+  defaultAgentCapabilities,
   validateAndParseOpenAPISpec,
 } = require('librechat-data-provider');
 const {
@@ -501,8 +503,22 @@ async function loadAgentTools({ req, res, agent, tool_resources, openAIApiKey })
   }
 
   const endpointsConfig = await getEndpointsConfig(req);
-  const enabledCapabilities = new Set(endpointsConfig?.[EModelEndpoint.agents]?.capabilities ?? []);
-  const checkCapability = (capability) => enabledCapabilities.has(capability);
+  let enabledCapabilities = new Set(endpointsConfig?.[EModelEndpoint.agents]?.capabilities ?? []);
+  /** Edge case: use defined/fallback capabilities when the "agents" endpoint is not enabled */
+  if (enabledCapabilities.size === 0 && agent.id === Constants.EPHEMERAL_AGENT_ID) {
+    enabledCapabilities = new Set(
+      req.app?.locals?.[EModelEndpoint.agents]?.capabilities ?? defaultAgentCapabilities,
+    );
+  }
+  const checkCapability = (capability) => {
+    const enabled = enabledCapabilities.has(capability);
+    if (!enabled) {
+      logger.warn(
+        `Capability "${capability}" disabled${capability === AgentCapabilities.tools ? '.' : ' despite configured tool.'} User: ${req.user.id} | Agent: ${agent.id}`,
+      );
+    }
+    return enabled;
+  };
   const areToolsEnabled = checkCapability(AgentCapabilities.tools);
 
   let includesWebSearch = false;

api/server/utils/handleText.js

@@ -4,10 +4,10 @@ const {
   Capabilities,
   EModelEndpoint,
   isAgentsEndpoint,
-  AgentCapabilities,
   isAssistantsEndpoint,
   defaultRetrievalModels,
   defaultAssistantsVersion,
+  defaultAgentCapabilities,
 } = require('librechat-data-provider');
 const { Providers } = require('@librechat/agents');
 const partialRight = require('lodash/partialRight');
@@ -197,16 +197,7 @@ function generateConfig(key, baseURL, endpoint) {
   }
 
   if (agents) {
-    config.capabilities = [
-      AgentCapabilities.execute_code,
-      AgentCapabilities.file_search,
-      AgentCapabilities.web_search,
-      AgentCapabilities.artifacts,
-      AgentCapabilities.actions,
-      AgentCapabilities.tools,
-      AgentCapabilities.chain,
-      AgentCapabilities.ocr,
-    ];
+    config.capabilities = defaultAgentCapabilities;
   }
 
   if (assistants && endpoint === EModelEndpoint.azureAssistants) {