openlit 1.10.0 → 1.12.0

package/dist/instrumentation/bedrock/wrapper.js

@@ -1,15 +1,46 @@
 "use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
 var __importDefault = (this && this.__importDefault) || function (mod) {
     return (mod && mod.__esModule) ? mod : { "default": mod };
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 const api_1 = require("@opentelemetry/api");
 const config_1 = __importDefault(require("../../config"));
-const helpers_1 = __importDefault(require("../../helpers"));
+const helpers_1 = __importStar(require("../../helpers"));
 const semantic_convention_1 = __importDefault(require("../../semantic-convention"));
 const base_wrapper_1 = __importDefault(require("../base-wrapper"));
-const BEDROCK_SERVER_ADDRESS = 'bedrock-runtime.amazonaws.com';
-const BEDROCK_SERVER_PORT = 443;
 function mapFinishReason(stopReason) {
     const map = {
         end_turn: 'stop',
@@ -21,22 +52,58 @@ function mapFinishReason(stopReason) {
     };
     return map[stopReason] || stopReason;
 }
-function applyInferenceConfigAttributes(span, inferenceConfig) {
-    if (inferenceConfig.maxTokens !== undefined)
-        span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_MAX_TOKENS, inferenceConfig.maxTokens);
-    if (inferenceConfig.temperature !== undefined)
-        span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_TEMPERATURE, inferenceConfig.temperature);
-    if (inferenceConfig.topP !== undefined)
-        span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_TOP_P, inferenceConfig.topP);
-    if (inferenceConfig.topK !== undefined)
-        span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_TOP_K, inferenceConfig.topK);
-    if (inferenceConfig.stopSequences !== undefined)
-        span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_STOP_SEQUENCES, inferenceConfig.stopSequences);
+function spanCreationAttrs(operationName, requestModel) {
+    return {
+        [semantic_convention_1.default.GEN_AI_OPERATION]: operationName,
+        [semantic_convention_1.default.GEN_AI_PROVIDER_NAME_OTEL]: semantic_convention_1.default.GEN_AI_SYSTEM_AWS_BEDROCK,
+        [semantic_convention_1.default.GEN_AI_REQUEST_MODEL]: requestModel,
+        [semantic_convention_1.default.SERVER_ADDRESS]: BedrockWrapper.serverAddress,
+        [semantic_convention_1.default.SERVER_PORT]: BedrockWrapper.serverPort,
+    };
+}
+/**
+ * Convert Bedrock message content blocks ({text: "..."}) to the format
+ * expected by OpenLitHelper.buildInputMessages ({type: "text", text: "..."}).
+ */
+function convertBedrockMessages(messages) {
+    return (messages || []).map((m) => {
+        const role = m.role || 'user';
+        const content = m.content;
+        if (!Array.isArray(content)) {
+            return { role, content: typeof content === 'string' ? content : '' };
+        }
+        return {
+            role,
+            content: content.map((c) => {
+                if (c.text !== undefined)
+                    return { type: 'text', text: c.text };
+                if (c.toolUse) {
+                    return {
+                        type: 'tool_use',
+                        id: c.toolUse.toolUseId || '',
+                        name: c.toolUse.name || '',
+                        input: c.toolUse.input || {},
+                    };
+                }
+                if (c.toolResult) {
+                    const rc = c.toolResult.content;
+                    return {
+                        type: 'tool_result',
+                        tool_use_id: c.toolResult.toolUseId || '',
+                        content: typeof rc === 'string' ? rc : JSON.stringify(rc || ''),
+                    };
+                }
+                return c;
+            }),
+        };
+    });
 }
 class BedrockWrapper extends base_wrapper_1.default {
     static _patchSend(tracer) {
         return (originalMethod) => {
             return async function (...args) {
+                if ((0, helpers_1.isFrameworkLlmActive)())
+                    return originalMethod.apply(this, args);
                 const command = args[0];
                 if (!command)
                     return originalMethod.apply(this, args);
@@ -47,9 +114,6 @@ class BedrockWrapper extends base_wrapper_1.default {
                 if (commandName === 'ConverseStreamCommand') {
                     return BedrockWrapper._handleConverseStreamCommand(tracer, originalMethod, this, args);
                 }
-                if (commandName === 'InvokeModelCommand' || commandName === 'InvokeModelWithResponseStreamCommand') {
-                    return BedrockWrapper._handleInvokeModelCommand(tracer, originalMethod, this, args, commandName);
-                }
                 return originalMethod.apply(this, args);
             };
         };
@@ -57,107 +121,100 @@ class BedrockWrapper extends base_wrapper_1.default {
     static async _handleConverseCommand(tracer, originalMethod, instance, args) {
         const command = args[0];
         const input = command.input || {};
-        const modelId = input.modelId || 'unknown';
+        const modelId = input.modelId || 'amazon.titan-text-express-v1';
         const genAIEndpoint = 'bedrock.converse';
-        const span = tracer.startSpan(genAIEndpoint, { kind: api_1.SpanKind.CLIENT });
-        return api_1.context.with(api_1.trace.setSpan(api_1.context.active(), span), async () => {
-            const startTime = Date.now();
-            let metricParams;
-            try {
-                const response = await originalMethod.apply(instance, args);
-                const duration = (Date.now() - startTime) / 1000;
-                const usage = response.usage || {};
-                const promptTokens = usage.inputTokens || 0;
-                const completionTokens = usage.outputTokens || 0;
-                const totalTokens = usage.totalTokens || promptTokens + completionTokens;
-                const cacheReadTokens = usage.cacheReadInputTokens || 0;
-                const cacheWriteTokens = usage.cacheWriteInputTokens || 0;
-                const finishReason = mapFinishReason(response.stopReason || 'stop');
-                const pricingInfo = await config_1.default.updatePricingJson(config_1.default.pricing_json);
-                const cost = helpers_1.default.getChatModelCost(modelId, pricingInfo, promptTokens, completionTokens);
-                BedrockWrapper.setBaseSpanAttributes(span, {
-                    genAIEndpoint,
-                    model: modelId,
-                    cost,
-                    aiSystem: BedrockWrapper.aiSystem,
-                    serverAddress: BEDROCK_SERVER_ADDRESS,
-                    serverPort: BEDROCK_SERVER_PORT,
-                });
-                span.setAttribute(semantic_convention_1.default.GEN_AI_OPERATION, semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_IS_STREAM, false);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_MODEL, modelId);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_INPUT_TOKENS, promptTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_OUTPUT_TOKENS, completionTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_TOTAL_TOKENS, totalTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_CLIENT_TOKEN_USAGE, totalTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_FINISH_REASON, [finishReason]);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_TYPE, semantic_convention_1.default.GEN_AI_OUTPUT_TYPE_TEXT);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_CLIENT_OPERATION_DURATION, duration);
-                span.setAttribute(semantic_convention_1.default.SERVER_ADDRESS, BEDROCK_SERVER_ADDRESS);
-                span.setAttribute(semantic_convention_1.default.SERVER_PORT, BEDROCK_SERVER_PORT);
-                if (cacheReadTokens > 0)
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_CACHE_READ_INPUT_TOKENS, cacheReadTokens);
-                if (cacheWriteTokens > 0)
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_CACHE_CREATION_INPUT_TOKENS, cacheWriteTokens);
-                const requestId = response.$metadata?.requestId;
-                if (requestId)
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_ID, requestId);
-                applyInferenceConfigAttributes(span, input.inferenceConfig || {});
-                if (config_1.default.traceContent) {
-                    const messages = (input.messages || []).map((m) => ({
-                        role: m.role,
-                        content: m.content?.map((c) => c.text || '').join('') || '',
-                    }));
-                    const systemText = input.system?.[0]?.text;
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_INPUT_MESSAGES, helpers_1.default.buildInputMessages(messages, systemText));
-                    if (systemText) {
-                        const systemStr = JSON.stringify([{ type: 'text', content: systemText }]);
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_SYSTEM_INSTRUCTIONS, systemStr);
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_SYSTEM_INSTRUCTIONS_OTEL, systemStr);
-                    }
-                    const outputText = response.output?.message?.content?.map((c) => c.text || '').join('') || '';
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_MESSAGES, helpers_1.default.buildOutputMessages(outputText, finishReason));
-                }
-                metricParams = { genAIEndpoint, model: modelId, cost, aiSystem: BedrockWrapper.aiSystem };
-                return response;
-            }
-            catch (e) {
-                helpers_1.default.handleException(span, e);
-                throw e;
-            }
-            finally {
-                span.end();
-                if (metricParams)
-                    base_wrapper_1.default.recordMetrics(span, metricParams);
-            }
+        const spanName = `${semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT} ${modelId}`;
+        const effectiveCtx = (0, helpers_1.getFrameworkParentContext)() ?? api_1.context.active();
+        const span = tracer.startSpan(spanName, {
+            kind: api_1.SpanKind.CLIENT,
+            attributes: spanCreationAttrs(semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT, modelId),
+        }, effectiveCtx);
+        return api_1.context
+            .with(api_1.trace.setSpan(effectiveCtx, span), async () => {
+            return originalMethod.apply(instance, args);
+        })
+            .then((response) => {
+            return BedrockWrapper._converseComplete({ input, genAIEndpoint, response, span, modelId });
+        })
+            .catch((e) => {
+            helpers_1.default.handleException(span, e);
+            base_wrapper_1.default.recordMetrics(span, {
+                genAIEndpoint,
+                model: modelId,
+                aiSystem: BedrockWrapper.aiSystem,
+                serverAddress: BedrockWrapper.serverAddress,
+                serverPort: BedrockWrapper.serverPort,
+                errorType: e?.constructor?.name || '_OTHER',
+            });
+            span.end();
+            throw e;
         });
     }
+    static async _converseComplete({ input, genAIEndpoint, response, span, modelId, }) {
+        let metricParams;
+        try {
+            metricParams = BedrockWrapper._converseCommonSetter({
+                input,
+                genAIEndpoint,
+                result: response,
+                span,
+                modelId,
+                isStream: false,
+            });
+            return response;
+        }
+        catch (e) {
+            helpers_1.default.handleException(span, e);
+            throw e;
+        }
+        finally {
+            span.end();
+            if (metricParams) {
+                base_wrapper_1.default.recordMetrics(span, metricParams);
+            }
+        }
+    }
     static async _handleConverseStreamCommand(tracer, originalMethod, instance, args) {
         const command = args[0];
         const input = command.input || {};
-        const modelId = input.modelId || 'unknown';
+        const modelId = input.modelId || 'amazon.titan-text-express-v1';
         const genAIEndpoint = 'bedrock.converse_stream';
         const startTime = Date.now();
-        const span = tracer.startSpan(genAIEndpoint, { kind: api_1.SpanKind.CLIENT });
-        // Call the original method to get the response object (with .stream async iterable)
-        const response = await api_1.context.with(api_1.trace.setSpan(api_1.context.active(), span), () => originalMethod.apply(instance, args));
-        // Accumulated state from stream events
+        const spanName = `${semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT} ${modelId}`;
+        const effectiveCtx = (0, helpers_1.getFrameworkParentContext)() ?? api_1.context.active();
+        const span = tracer.startSpan(spanName, {
+            kind: api_1.SpanKind.CLIENT,
+            attributes: spanCreationAttrs(semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT, modelId),
+        }, effectiveCtx);
+        let response;
+        try {
+            response = await api_1.context.with(api_1.trace.setSpan(effectiveCtx, span), () => originalMethod.apply(instance, args));
+        }
+        catch (e) {
+            helpers_1.default.handleException(span, e);
+            base_wrapper_1.default.recordMetrics(span, {
+                genAIEndpoint,
+                model: modelId,
+                aiSystem: BedrockWrapper.aiSystem,
+                serverAddress: BedrockWrapper.serverAddress,
+                serverPort: BedrockWrapper.serverPort,
+                errorType: e?.constructor?.name || '_OTHER',
+            });
+            span.end();
+            throw e;
+        }
         let llmResponse = '';
         let finishReason = 'stop';
         let inputTokens = 0;
         let outputTokens = 0;
         let cacheReadTokens = 0;
         let cacheWriteTokens = 0;
-        let firstChunkTime = null;
-        const chunkTimestamps = [];
+        const timestamps = [];
         const originalStream = response.stream;
         async function* wrappedStream() {
             try {
                 for await (const event of originalStream) {
-                    const now = Date.now();
-                    if (firstChunkTime === null)
-                        firstChunkTime = now;
-                    chunkTimestamps.push(now);
+                    timestamps.push(Date.now());
                     if (event.contentBlockDelta?.delta?.text)
                         llmResponse += event.contentBlockDelta.delta.text;
                     if (event.messageStop?.stopReason)
@@ -172,63 +229,34 @@ class BedrockWrapper extends base_wrapper_1.default {
                 }
             }
             finally {
-                // Record telemetry once the stream is fully consumed
                 try {
-                    const duration = (Date.now() - startTime) / 1000;
-                    const ttft = firstChunkTime !== null ? (firstChunkTime - startTime) / 1000 : 0;
+                    const ttft = timestamps.length > 0 ? (timestamps[0] - startTime) / 1000 : 0;
                     let tbt = 0;
-                    if (chunkTimestamps.length > 1) {
-                        const timeDiffs = chunkTimestamps.slice(1).map((t, i) => t - chunkTimestamps[i]);
+                    if (timestamps.length > 1) {
+                        const timeDiffs = timestamps.slice(1).map((t, i) => t - timestamps[i]);
                         tbt = timeDiffs.reduce((a, b) => a + b, 0) / timeDiffs.length / 1000;
                     }
-                    const totalTokens = inputTokens + outputTokens;
-                    const pricingInfo = await config_1.default.updatePricingJson(config_1.default.pricing_json);
-                    const cost = helpers_1.default.getChatModelCost(modelId, pricingInfo, inputTokens, outputTokens);
-                    BedrockWrapper.setBaseSpanAttributes(span, {
+                    const result = {
+                        output: { message: { content: [{ text: llmResponse }] } },
+                        stopReason: finishReason,
+                        usage: {
+                            inputTokens,
+                            outputTokens,
+                            cacheReadInputTokens: cacheReadTokens,
+                            cacheWriteInputTokens: cacheWriteTokens,
+                        },
+                        $metadata: response.$metadata,
+                    };
+                    const metricParams = BedrockWrapper._converseCommonSetter({
+                        input,
                         genAIEndpoint,
-                        model: modelId,
-                        cost,
-                        aiSystem: BedrockWrapper.aiSystem,
+                        result,
+                        span,
+                        modelId,
+                        isStream: true,
+                        ttft,
+                        tbt,
                     });
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_OPERATION, semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_IS_STREAM, true);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_MODEL, modelId);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_INPUT_TOKENS, inputTokens);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_OUTPUT_TOKENS, outputTokens);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_TOTAL_TOKENS, totalTokens);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_CLIENT_TOKEN_USAGE, totalTokens);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_FINISH_REASON, [finishReason]);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_TYPE, semantic_convention_1.default.GEN_AI_OUTPUT_TYPE_TEXT);
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_CLIENT_OPERATION_DURATION, duration);
-                    if (ttft > 0)
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_SERVER_TTFT, ttft);
-                    if (tbt > 0)
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_SERVER_TBT, tbt);
-                    span.setAttribute(semantic_convention_1.default.SERVER_ADDRESS, BEDROCK_SERVER_ADDRESS);
-                    span.setAttribute(semantic_convention_1.default.SERVER_PORT, BEDROCK_SERVER_PORT);
-                    if (cacheReadTokens > 0)
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_CACHE_READ_INPUT_TOKENS, cacheReadTokens);
-                    if (cacheWriteTokens > 0)
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_CACHE_CREATION_INPUT_TOKENS, cacheWriteTokens);
-                    const requestId = response.$metadata?.requestId;
-                    if (requestId)
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_ID, requestId);
-                    applyInferenceConfigAttributes(span, input.inferenceConfig || {});
-                    if (config_1.default.traceContent) {
-                        const messages = (input.messages || []).map((m) => ({
-                            role: m.role,
-                            content: m.content?.map((c) => c.text || '').join('') || '',
-                        }));
-                        const systemText = input.system?.[0]?.text;
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_INPUT_MESSAGES, helpers_1.default.buildInputMessages(messages, systemText));
-                        if (systemText) {
-                            const systemStr = JSON.stringify([{ type: 'text', content: systemText }]);
-                            span.setAttribute(semantic_convention_1.default.GEN_AI_SYSTEM_INSTRUCTIONS, systemStr);
-                            span.setAttribute(semantic_convention_1.default.GEN_AI_SYSTEM_INSTRUCTIONS_OTEL, systemStr);
-                        }
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_MESSAGES, helpers_1.default.buildOutputMessages(llmResponse, finishReason));
-                    }
-                    const metricParams = { genAIEndpoint, model: modelId, cost, aiSystem: BedrockWrapper.aiSystem };
                     base_wrapper_1.default.recordMetrics(span, metricParams);
                 }
                 catch { /* ignore telemetry errors in finally */ }
@@ -239,126 +267,151 @@ class BedrockWrapper extends base_wrapper_1.default {
         }
         return { ...response, stream: wrappedStream() };
     }
-    static async _handleInvokeModelCommand(tracer, originalMethod, instance, args, commandName) {
-        const command = args[0];
-        const input = command.input || {};
-        const modelId = input.modelId || 'unknown';
-        const isStream = commandName === 'InvokeModelWithResponseStreamCommand';
-        const genAIEndpoint = isStream ? 'bedrock.invoke_model_stream' : 'bedrock.invoke_model';
-        const span = tracer.startSpan(genAIEndpoint, { kind: api_1.SpanKind.CLIENT });
-        return api_1.context.with(api_1.trace.setSpan(api_1.context.active(), span), async () => {
-            const startTime = Date.now();
-            let metricParams;
-            try {
-                const response = await originalMethod.apply(instance, args);
-                const duration = (Date.now() - startTime) / 1000;
-                // Parse response body
-                let parsedBody = {};
-                let promptTokens = 0;
-                let completionTokens = 0;
-                let outputText = '';
-                let rawFinishReason = 'stop';
-                try {
-                    const bodyBytes = response.body;
-                    if (bodyBytes) {
-                        const bodyStr = typeof bodyBytes === 'string'
-                            ? bodyBytes
-                            : Buffer.from(bodyBytes).toString('utf-8');
-                        parsedBody = JSON.parse(bodyStr);
-                    }
-                }
-                catch { /* ignore parse errors */ }
-                // Handle different provider response formats
-                if (modelId.startsWith('anthropic')) {
-                    promptTokens = parsedBody.usage?.input_tokens || 0;
-                    completionTokens = parsedBody.usage?.output_tokens || 0;
-                    outputText = parsedBody.content?.[0]?.text || '';
-                    rawFinishReason = parsedBody.stop_reason || 'stop';
-                }
-                else if (modelId.startsWith('amazon')) {
-                    promptTokens = parsedBody.inputTextTokenCount || 0;
-                    completionTokens = parsedBody.results?.[0]?.tokenCount || 0;
-                    outputText = parsedBody.results?.[0]?.outputText || '';
-                    rawFinishReason = parsedBody.results?.[0]?.completionReason || 'stop';
-                }
-                else if (modelId.startsWith('meta')) {
-                    promptTokens = parsedBody.prompt_token_count || 0;
-                    completionTokens = parsedBody.generation_token_count || 0;
-                    outputText = parsedBody.generation || '';
-                    rawFinishReason = parsedBody.stop_reason || 'stop';
-                }
-                else if (modelId.startsWith('mistral') || modelId.startsWith('mixtral')) {
-                    promptTokens = parsedBody.usage?.prompt_tokens || 0;
-                    completionTokens = parsedBody.usage?.completion_tokens || 0;
-                    outputText = parsedBody.outputs?.[0]?.text || '';
-                    rawFinishReason = parsedBody.outputs?.[0]?.stop_reason || 'stop';
-                }
-                else if (modelId.startsWith('ai21')) {
-                    promptTokens = parsedBody.prompt?.tokens?.length || 0;
-                    completionTokens = parsedBody.completions?.[0]?.data?.tokens?.length || 0;
-                    outputText = parsedBody.completions?.[0]?.data?.text || '';
-                    rawFinishReason = parsedBody.completions?.[0]?.finishReason?.reason || 'stop';
-                }
-                else {
-                    outputText = parsedBody.output || parsedBody.generation || parsedBody.text || '';
-                }
-                const finishReason = mapFinishReason(rawFinishReason);
-                const totalTokens = promptTokens + completionTokens;
-                const pricingInfo = await config_1.default.updatePricingJson(config_1.default.pricing_json);
-                const cost = helpers_1.default.getChatModelCost(modelId, pricingInfo, promptTokens, completionTokens);
-                BedrockWrapper.setBaseSpanAttributes(span, {
-                    genAIEndpoint,
-                    model: modelId,
-                    cost,
-                    aiSystem: BedrockWrapper.aiSystem,
-                    serverAddress: BEDROCK_SERVER_ADDRESS,
-                    serverPort: BEDROCK_SERVER_PORT,
-                });
-                span.setAttribute(semantic_convention_1.default.GEN_AI_OPERATION, semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_IS_STREAM, isStream);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_MODEL, modelId);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_INPUT_TOKENS, promptTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_OUTPUT_TOKENS, completionTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_TOTAL_TOKENS, totalTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_CLIENT_TOKEN_USAGE, totalTokens);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_FINISH_REASON, [finishReason]);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_TYPE, semantic_convention_1.default.GEN_AI_OUTPUT_TYPE_TEXT);
-                span.setAttribute(semantic_convention_1.default.GEN_AI_CLIENT_OPERATION_DURATION, duration);
-                span.setAttribute(semantic_convention_1.default.SERVER_ADDRESS, BEDROCK_SERVER_ADDRESS);
-                span.setAttribute(semantic_convention_1.default.SERVER_PORT, BEDROCK_SERVER_PORT);
-                const requestId = response.$metadata?.requestId;
-                if (requestId)
-                    span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_ID, requestId);
-                if (config_1.default.traceContent) {
-                    try {
-                        const reqBody = input.body
-                            ? JSON.parse(typeof input.body === 'string' ? input.body : Buffer.from(input.body).toString())
-                            : {};
-                        const prompt = reqBody.prompt || reqBody.inputText || '';
-                        if (prompt) {
-                            span.setAttribute(semantic_convention_1.default.GEN_AI_INPUT_MESSAGES, helpers_1.default.buildInputMessages([{ role: 'user', content: prompt }]));
-                        }
-                    }
-                    catch { /* ignore */ }
-                    if (outputText) {
-                        span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_MESSAGES, helpers_1.default.buildOutputMessages(outputText, finishReason));
-                    }
+    static _converseCommonSetter({ input, genAIEndpoint, result, span, modelId, isStream, ttft = 0, tbt = 0, }) {
+        const captureContent = config_1.default.captureMessageContent;
+        const inferenceConfig = input.inferenceConfig || {};
+        if (inferenceConfig.temperature !== undefined) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_TEMPERATURE, inferenceConfig.temperature);
+        }
+        if (inferenceConfig.topP !== undefined) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_TOP_P, inferenceConfig.topP);
+        }
+        if (inferenceConfig.topK !== undefined) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_TOP_K, inferenceConfig.topK);
+        }
+        if (inferenceConfig.maxTokens != null) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_MAX_TOKENS, inferenceConfig.maxTokens);
+        }
+        if (inferenceConfig.stopSequences) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_STOP_SEQUENCES, inferenceConfig.stopSequences);
+        }
+        if (inferenceConfig.frequencyPenalty) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_FREQUENCY_PENALTY, inferenceConfig.frequencyPenalty);
+        }
+        if (inferenceConfig.presencePenalty) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_PRESENCE_PENALTY, inferenceConfig.presencePenalty);
+        }
+        span.setAttribute(semantic_convention_1.default.GEN_AI_REQUEST_IS_STREAM, isStream);
+        const usage = result.usage || {};
+        const inputTokens = usage.inputTokens || 0;
+        const outputTokens = usage.outputTokens || 0;
+        const cacheReadTokens = usage.cacheReadInputTokens || 0;
+        const cacheWriteTokens = usage.cacheWriteInputTokens || 0;
+        const responseModel = modelId;
+        const finishReason = mapFinishReason(result.stopReason || 'stop');
+        const pricingInfo = config_1.default.pricingInfo || {};
+        const cost = helpers_1.default.getChatModelCost(modelId, pricingInfo, inputTokens, outputTokens);
+        BedrockWrapper.setBaseSpanAttributes(span, {
+            genAIEndpoint,
+            model: modelId,
+            cost,
+            aiSystem: BedrockWrapper.aiSystem,
+            serverAddress: BedrockWrapper.serverAddress,
+            serverPort: BedrockWrapper.serverPort,
+        });
+        span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_MODEL, responseModel);
+        const requestId = result.$metadata?.requestId;
+        if (requestId) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_ID, requestId);
+        }
+        span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_INPUT_TOKENS, inputTokens);
+        span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_OUTPUT_TOKENS, outputTokens);
+        if (cacheReadTokens > 0) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_CACHE_READ_INPUT_TOKENS, cacheReadTokens);
+        }
+        if (cacheWriteTokens > 0) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_USAGE_CACHE_CREATION_INPUT_TOKENS, cacheWriteTokens);
+        }
+        if (ttft > 0) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_SERVER_TTFT, ttft);
+        }
+        if (tbt > 0) {
+            span.setAttribute(semantic_convention_1.default.GEN_AI_SERVER_TBT, tbt);
+        }
+        span.setAttribute(semantic_convention_1.default.GEN_AI_RESPONSE_FINISH_REASON, [finishReason]);
+        const outputText = result.output?.message?.content?.map((c) => c.text || '').join('') || '';
+        const outputType = typeof outputText === 'string'
+            ? semantic_convention_1.default.GEN_AI_OUTPUT_TYPE_TEXT
+            : semantic_convention_1.default.GEN_AI_OUTPUT_TYPE_JSON;
+        span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_TYPE, outputType);
+        const contentBlocks = result.output?.message?.content || [];
+        const toolCalls = contentBlocks
+            .filter((c) => c.toolUse)
+            .map((c) => ({
+            id: c.toolUse.toolUseId || '',
+            name: c.toolUse.name || '',
+            arguments: c.toolUse.input || {},
+        }));
+        if (toolCalls.length > 0) {
+            const toolNames = toolCalls.map((t) => t.name).filter(Boolean);
+            const toolIds = toolCalls.map((t) => t.id).filter(Boolean);
+            const toolArgs = toolCalls
+                .map((t) => (typeof t.arguments === 'string' ? t.arguments : JSON.stringify(t.arguments)))
+                .filter(Boolean);
+            if (toolNames.length > 0) {
+                span.setAttribute(semantic_convention_1.default.GEN_AI_TOOL_NAME, toolNames.join(', '));
+            }
+            if (toolIds.length > 0) {
+                span.setAttribute(semantic_convention_1.default.GEN_AI_TOOL_CALL_ID, toolIds.join(', '));
+            }
+            if (toolArgs.length > 0) {
+                span.setAttribute(semantic_convention_1.default.GEN_AI_TOOL_ARGS, toolArgs.join(', '));
+            }
+        }
+        const messages = convertBedrockMessages(input.messages || []);
+        const systemBlock = input.system || [];
+        const systemParts = [];
+        if (Array.isArray(systemBlock)) {
+            for (const item of systemBlock) {
+                if (item?.text) {
+                    systemParts.push({ type: 'text', content: item.text });
                 }
-                metricParams = { genAIEndpoint, model: modelId, cost, aiSystem: BedrockWrapper.aiSystem };
-                return response;
             }
-            catch (e) {
-                helpers_1.default.handleException(span, e);
-                throw e;
+        }
+        let inputMessagesJson;
+        let outputMessagesJson;
+        if (captureContent) {
+            inputMessagesJson = helpers_1.default.buildInputMessages(messages);
+            span.setAttribute(semantic_convention_1.default.GEN_AI_INPUT_MESSAGES, inputMessagesJson);
+            if (systemParts.length > 0) {
+                span.setAttribute(semantic_convention_1.default.GEN_AI_SYSTEM_INSTRUCTIONS, JSON.stringify(systemParts));
             }
-            finally {
-                span.end();
-                if (metricParams)
-                    base_wrapper_1.default.recordMetrics(span, metricParams);
+            outputMessagesJson = helpers_1.default.buildOutputMessages(outputText, finishReason, toolCalls.length > 0 ? toolCalls : undefined);
+            span.setAttribute(semantic_convention_1.default.GEN_AI_OUTPUT_MESSAGES, outputMessagesJson);
+        }
+        if (!config_1.default.disableEvents) {
+            const eventAttrs = {
+                [semantic_convention_1.default.GEN_AI_OPERATION]: semantic_convention_1.default.GEN_AI_OPERATION_TYPE_CHAT,
+                [semantic_convention_1.default.GEN_AI_REQUEST_MODEL]: modelId,
+                [semantic_convention_1.default.GEN_AI_RESPONSE_MODEL]: responseModel,
+                [semantic_convention_1.default.SERVER_ADDRESS]: BedrockWrapper.serverAddress,
+                [semantic_convention_1.default.SERVER_PORT]: BedrockWrapper.serverPort,
+                [semantic_convention_1.default.GEN_AI_RESPONSE_FINISH_REASON]: [finishReason],
+                [semantic_convention_1.default.GEN_AI_OUTPUT_TYPE]: outputType,
+                [semantic_convention_1.default.GEN_AI_USAGE_INPUT_TOKENS]: inputTokens,
+                [semantic_convention_1.default.GEN_AI_USAGE_OUTPUT_TOKENS]: outputTokens,
+            };
+            if (requestId) {
+                eventAttrs[semantic_convention_1.default.GEN_AI_RESPONSE_ID] = requestId;
             }
-        });
+            if (captureContent) {
+                if (inputMessagesJson)
+                    eventAttrs[semantic_convention_1.default.GEN_AI_INPUT_MESSAGES] = inputMessagesJson;
+                if (outputMessagesJson)
+                    eventAttrs[semantic_convention_1.default.GEN_AI_OUTPUT_MESSAGES] = outputMessagesJson;
+            }
+            helpers_1.default.emitInferenceEvent(span, eventAttrs);
+        }
+        return {
+            genAIEndpoint,
+            model: modelId,
+            cost,
+            aiSystem: BedrockWrapper.aiSystem,
+        };
     }
 }
 BedrockWrapper.aiSystem = semantic_convention_1.default.GEN_AI_SYSTEM_AWS_BEDROCK;
+BedrockWrapper.serverAddress = 'bedrock-runtime.amazonaws.com';
+BedrockWrapper.serverPort = 443;
 exports.default = BedrockWrapper;
 //# sourceMappingURL=wrapper.js.map