genai-lite 0.3.0 → 0.3.2

package/README.md

@@ -240,7 +240,7 @@ const response = await llmService.sendMessage({
 // 
 // The response will have:
 // - response.choices[0].message.content = "The answer is 36."
-// - response.choices[0].reasoning = "<!-- Extracted by genai-lite from <thinking> tag -->\n15% means 15/100 = 0.15. So 15% of 240 = 0.15 × 240 = 36."
+// - response.choices[0].reasoning = "15% means 15/100 = 0.15. So 15% of 240 = 0.15 × 240 = 36."
 
 // If the model doesn't include the <thinking> tag, you'll get an error (with default 'auto' mode)
 ```

package/dist/llm/LLMService.js

@@ -155,7 +155,8 @@ class LLMService {
                         // Check if native reasoning is active
                         const isNativeReasoningActive = modelInfo.reasoning?.supported === true &&
                             (internalRequest.settings.reasoning?.enabled === true ||
-                                modelInfo.reasoning?.enabledByDefault === true ||
+                                (modelInfo.reasoning?.enabledByDefault === true &&
+                                    internalRequest.settings.reasoning?.enabled !== false) || // Only if not explicitly disabled
                                 modelInfo.reasoning?.canDisable === false); // Always-on models
                         effectiveOnMissing = isNativeReasoningActive ? 'ignore' : 'error';
                     }
@@ -167,11 +168,15 @@ class LLMService {
                             console.log(`Extracted <${tagName}> block from response.`);
                             // Handle the edge case: append to existing reasoning if present.
                             const existingReasoning = choice.reasoning || '';
-                            const separator = existingReasoning ? '\n\n' : '';
-                            // Add a comment to indicate the source of this reasoning block.
-                            const newReasoning = `<!-- Extracted by genai-lite from <${tagName}> tag -->\n${extracted}`;
-                            // Update the choice object
-                            choice.reasoning = `${existingReasoning}${separator}${newReasoning}`;
+                            // Only add a separator when appending to existing reasoning
+                            if (existingReasoning) {
+                                // Use a neutral markdown header that works for any consumer (human or AI)
+                                choice.reasoning = `${existingReasoning}\n\n#### Additional Reasoning\n\n${extracted}`;
+                            }
+                            else {
+                                // No existing reasoning, just use the extracted content directly
+                                choice.reasoning = extracted;
+                            }
                             choice.message.content = remaining;
                         }
                         else {

package/dist/llm/LLMService.test.js

@@ -376,8 +376,7 @@ describe('LLMService', () => {
             const response = await service.sendMessage(request);
             expect(response.object).toBe('chat.completion');
             const successResponse = response;
-            expect(successResponse.choices[0].reasoning).toContain('I am thinking about this problem.');
-            expect(successResponse.choices[0].reasoning).toContain('<!-- Extracted by genai-lite from <thinking> tag -->');
+            expect(successResponse.choices[0].reasoning).toBe('I am thinking about this problem.');
             expect(successResponse.choices[0].message.content).toBe('Here is the answer.');
         });
         it('should not extract thinking tag when disabled', async () => {
@@ -413,19 +412,15 @@ describe('LLMService', () => {
             const response = await service.sendMessage(request);
             expect(response.object).toBe('chat.completion');
             const successResponse = response;
-            expect(successResponse.choices[0].reasoning).toContain('Working through the logic...');
-            expect(successResponse.choices[0].reasoning).toContain('<!-- Extracted by genai-lite from <scratchpad> tag -->');
+            expect(successResponse.choices[0].reasoning).toBe('Working through the logic...');
             expect(successResponse.choices[0].message.content).toBe('Final answer is 42.');
         });
         it('should append to existing reasoning', async () => {
-            // For this test, we first create a response with reasoning by using a reasoning-enabled model
-            // Then test that thinking extraction appends to it
-            // Since MockClientAdapter doesn't generate reasoning, we'll skip this complex test
-            // and just test the simple case
+            // Use test_reasoning to get a response with existing reasoning, then test extraction appends to it
             const request = {
                 providerId: 'mistral',
                 modelId: 'codestral-2501',
-                messages: [{ role: 'user', content: 'test_thinking:<thinking>Additional thoughts here.</thinking>The analysis is complete.' }],
+                messages: [{ role: 'user', content: 'test_reasoning:<thinking>Additional thoughts here.</thinking>The analysis is complete.' }],
                 settings: {
                     thinkingExtraction: {
                         enabled: true,
@@ -436,8 +431,8 @@ describe('LLMService', () => {
             const response = await service.sendMessage(request);
             expect(response.object).toBe('chat.completion');
             const successResponse = response;
-            expect(successResponse.choices[0].reasoning).toContain('<!-- Extracted by genai-lite from <thinking> tag -->');
-            expect(successResponse.choices[0].reasoning).toContain('Additional thoughts here.');
+            // Should contain both the initial reasoning and the extracted thinking with separator
+            expect(successResponse.choices[0].reasoning).toBe('Initial model reasoning from native capabilities.\n\n#### Additional Reasoning\n\nAdditional thoughts here.');
             expect(successResponse.choices[0].message.content).toBe('The analysis is complete.');
         });
         it('should handle missing tag with explicit ignore', async () => {
@@ -549,6 +544,48 @@ describe('LLMService', () => {
                 const errorResponse = response;
                 expect(errorResponse.error.message).toContain('expected to start with a <reasoning> tag');
             });
+            describe('auto mode with native reasoning detection', () => {
+                it('should enforce thinking tags for non-reasoning models by default', async () => {
+                    // Mistral model doesn't have reasoning support
+                    const request = {
+                        providerId: 'mistral',
+                        modelId: 'codestral-2501',
+                        messages: [{ role: 'user', content: 'test_thinking:Response without thinking tag.' }],
+                        settings: {
+                            thinkingExtraction: {
+                                enabled: true,
+                                onMissing: 'auto'
+                            }
+                        }
+                    };
+                    const response = await service.sendMessage(request);
+                    // Should error because model doesn't have native reasoning
+                    expect(response.object).toBe('error');
+                    const errorResponse = response;
+                    expect(errorResponse.error.code).toBe('MISSING_EXPECTED_TAG');
+                    expect(errorResponse.error.message).toContain('does not have native reasoning active');
+                });
+                it('should respect explicit reasoning.enabled: false even for models with enabledByDefault', async () => {
+                    // This is the key test for the fix
+                    const request = {
+                        providerId: 'mistral',
+                        modelId: 'codestral-2501',
+                        messages: [{ role: 'user', content: 'test_thinking:Response without thinking tag.' }],
+                        settings: {
+                            reasoning: { enabled: false }, // Explicitly disabled
+                            thinkingExtraction: {
+                                enabled: true,
+                                onMissing: 'auto'
+                            }
+                        }
+                    };
+                    const response = await service.sendMessage(request);
+                    // Should error because reasoning is explicitly disabled
+                    expect(response.object).toBe('error');
+                    const errorResponse = response;
+                    expect(errorResponse.error.code).toBe('MISSING_EXPECTED_TAG');
+                });
+            });
         });
     });
 });

package/dist/llm/clients/MockClientAdapter.js

@@ -100,6 +100,11 @@ class MockClientAdapter {
             const startIndex = originalContent.indexOf("test_thinking:") + "test_thinking:".length;
             responseContent = originalContent.substring(startIndex).trim();
         }
+        else if (userContent.includes("test_reasoning:")) {
+            // Extract content after "test_reasoning:" and return it as both content and reasoning
+            const startIndex = originalContent.indexOf("test_reasoning:") + "test_reasoning:".length;
+            responseContent = originalContent.substring(startIndex).trim();
+        }
         else if (userContent.includes("hello") || userContent.includes("hi")) {
             responseContent =
                 "Hello! I'm a mock LLM assistant. How can I help you today?";
@@ -153,21 +158,26 @@ class MockClientAdapter {
         else if (request.settings.stopSequences.some((seq) => responseContent.includes(seq))) {
             finishReason = "stop";
         }
+        // Check if we need to add reasoning to the response
+        const isReasoningTest = userContent.includes("test_reasoning:");
+        const choice = {
+            message: {
+                role: "assistant",
+                content: responseContent,
+            },
+            finish_reason: finishReason,
+            index: 0,
+        };
+        // Add reasoning field for test_reasoning pattern
+        if (isReasoningTest) {
+            choice.reasoning = "Initial model reasoning from native capabilities.";
+        }
         return {
             id: `mock-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`,
             provider: request.providerId,
             model: request.modelId,
             created: Math.floor(Date.now() / 1000),
-            choices: [
-                {
-                    message: {
-                        role: "assistant",
-                        content: responseContent,
-                    },
-                    finish_reason: finishReason,
-                    index: 0,
-                },
-            ],
+            choices: [choice],
             usage: {
                 prompt_tokens: promptTokenCount,
                 completion_tokens: mockTokenCount,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "genai-lite",
-  "version": "0.3.0",
+  "version": "0.3.2",
   "description": "A lightweight, portable toolkit for interacting with various Generative AI APIs.",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",