@fallom/trace 0.2.2 → 0.2.4

package/dist/index.js

@@ -26,7 +26,7 @@ __export(models_exports, {
   get: () => get,
   init: () => init2
 });
-function log4(msg) {
+function log3(msg) {
   if (debugMode2) {
     console.log(`[Fallom] ${msg}`);
   }
@@ -58,11 +58,11 @@ function ensureInit() {
 }
 async function fetchConfigs(timeout = SYNC_TIMEOUT) {
   if (!apiKey2) {
-    log4("_fetchConfigs: No API key, skipping");
+    log3("_fetchConfigs: No API key, skipping");
     return;
   }
   try {
-    log4(`Fetching configs from ${baseUrl2}/configs`);
+    log3(`Fetching configs from ${baseUrl2}/configs`);
     const controller = new AbortController();
     const timeoutId = setTimeout(() => controller.abort(), timeout);
     const resp = await fetch(`${baseUrl2}/configs`, {
@@ -70,15 +70,15 @@ async function fetchConfigs(timeout = SYNC_TIMEOUT) {
       signal: controller.signal
     });
     clearTimeout(timeoutId);
-    log4(`Response status: ${resp.status}`);
+    log3(`Response status: ${resp.status}`);
     if (resp.ok) {
       const data = await resp.json();
       const configs = data.configs || [];
-      log4(`Got ${configs.length} configs: ${configs.map((c) => c.key)}`);
+      log3(`Got ${configs.length} configs: ${configs.map((c) => c.key)}`);
       for (const c of configs) {
         const key = c.key;
         const version = c.version || 1;
-        log4(`Config '${key}' v${version}: ${JSON.stringify(c.variants)}`);
+        log3(`Config '${key}' v${version}: ${JSON.stringify(c.variants)}`);
         if (!configCache.has(key)) {
           configCache.set(key, { versions: /* @__PURE__ */ new Map(), latest: null });
         }
@@ -87,10 +87,10 @@ async function fetchConfigs(timeout = SYNC_TIMEOUT) {
         cached.latest = version;
       }
     } else {
-      log4(`Fetch failed: ${resp.statusText}`);
+      log3(`Fetch failed: ${resp.statusText}`);
     }
   } catch (e) {
-    log4(`Fetch exception: ${e}`);
+    log3(`Fetch exception: ${e}`);
   }
 }
 async function fetchSpecificVersion(configKey, version, timeout = SYNC_TIMEOUT) {
@@ -122,24 +122,24 @@ async function get(configKey, sessionId, options = {}) {
   const { version, fallback, debug = false } = options;
   debugMode2 = debug;
   ensureInit();
-  log4(
+  log3(
     `get() called: configKey=${configKey}, sessionId=${sessionId}, fallback=${fallback}`
   );
   try {
     let configData = configCache.get(configKey);
-    log4(
+    log3(
       `Cache lookup for '${configKey}': ${configData ? "found" : "not found"}`
     );
     if (!configData) {
-      log4("Not in cache, fetching...");
+      log3("Not in cache, fetching...");
       await fetchConfigs(SYNC_TIMEOUT);
       configData = configCache.get(configKey);
-      log4(
+      log3(
         `After fetch, cache lookup: ${configData ? "found" : "still not found"}`
       );
     }
     if (!configData) {
-      log4(`Config not found, using fallback: ${fallback}`);
+      log3(`Config not found, using fallback: ${fallback}`);
       if (fallback) {
         console.warn(
           `[Fallom WARNING] Config '${configKey}' not found, using fallback model: ${fallback}`
@@ -183,20 +183,20 @@ async function get(configKey, sessionId, options = {}) {
     const variantsRaw = config.variants;
     const configVersion = config.version || targetVersion;
     const variants = Array.isArray(variantsRaw) ? variantsRaw : Object.values(variantsRaw);
-    log4(
+    log3(
       `Config found! Version: ${configVersion}, Variants: ${JSON.stringify(
         variants
       )}`
     );
     const hashBytes = (0, import_crypto.createHash)("md5").update(sessionId).digest();
     const hashVal = hashBytes.readUInt32BE(0) % 1e6;
-    log4(`Session hash: ${hashVal} (out of 1,000,000)`);
+    log3(`Session hash: ${hashVal} (out of 1,000,000)`);
     let cumulative = 0;
     let assignedModel = variants[variants.length - 1].model;
     for (const v of variants) {
       const oldCumulative = cumulative;
       cumulative += v.weight * 1e4;
-      log4(
+      log3(
         `Variant ${v.model}: weight=${v.weight}%, range=${oldCumulative}-${cumulative}, hash=${hashVal}, match=${hashVal < cumulative}`
       );
       if (hashVal < cumulative) {
@@ -204,7 +204,7 @@ async function get(configKey, sessionId, options = {}) {
         break;
       }
     }
-    log4(`\u2705 Assigned model: ${assignedModel}`);
+    log3(`\u2705 Assigned model: ${assignedModel}`);
     return returnModel(configKey, sessionId, assignedModel, configVersion);
   } catch (e) {
     if (e instanceof Error && e.message.includes("not found")) {
@@ -1073,32 +1073,6 @@ function generateHexId(length) {
   crypto.getRandomValues(bytes);
   return Array.from(bytes).map((b) => b.toString(16).padStart(2, "0")).join("");
 }
-function messagesToOtelAttributes(messages, completion, model, responseId) {
-  const attrs = {};
-  if (model) {
-    attrs["gen_ai.request.model"] = model;
-    attrs["gen_ai.response.model"] = model;
-  }
-  if (responseId) {
-    attrs["gen_ai.response.id"] = responseId;
-  }
-  if (messages) {
-    messages.forEach((msg, i) => {
-      attrs[`gen_ai.prompt.${i}.role`] = msg.role;
-      attrs[`gen_ai.prompt.${i}.content`] = typeof msg.content === "string" ? msg.content : JSON.stringify(msg.content);
-    });
-  }
-  if (completion) {
-    attrs["gen_ai.completion.0.role"] = completion.role;
-    attrs["gen_ai.completion.0.content"] = typeof completion.content === "string" ? completion.content : JSON.stringify(completion.content);
-    if (completion.tool_calls) {
-      attrs["gen_ai.completion.0.tool_calls"] = JSON.stringify(
-        completion.tool_calls
-      );
-    }
-  }
-  return attrs;
-}
 
 // src/trace/wrappers/openai.ts
 function wrapOpenAI(client, sessionCtx) {
@@ -1120,18 +1094,25 @@ function wrapOpenAI(client, sessionCtx) {
     try {
       const response = await originalCreate(...args);
       const endTime = Date.now();
-      const attributes = captureContent2 ? messagesToOtelAttributes(
-        params?.messages,
-        response?.choices?.[0]?.message,
-        response?.model || params?.model,
-        response?.id
-      ) : {};
+      const attributes = {
+        "fallom.sdk_version": "2",
+        "fallom.method": "chat.completions.create"
+      };
+      if (captureContent2) {
+        attributes["fallom.raw.request"] = JSON.stringify({
+          messages: params?.messages,
+          model: params?.model
+        });
+        attributes["fallom.raw.response"] = JSON.stringify({
+          text: response?.choices?.[0]?.message?.content,
+          finishReason: response?.choices?.[0]?.finish_reason,
+          responseId: response?.id,
+          model: response?.model
+        });
+      }
       if (response?.usage) {
         attributes["fallom.raw.usage"] = JSON.stringify(response.usage);
       }
-      if (response?.choices?.[0]?.finish_reason) {
-        attributes["gen_ai.response.finish_reason"] = response.choices[0].finish_reason;
-      }
       sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
@@ -1146,24 +1127,12 @@ function wrapOpenAI(client, sessionCtx) {
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "OK",
-        prompt_tokens: response?.usage?.prompt_tokens,
-        completion_tokens: response?.usage?.completion_tokens,
-        total_tokens: response?.usage?.total_tokens,
-        attributes: Object.keys(attributes).length > 0 ? attributes : void 0
+        attributes
       }).catch(() => {
       });
       return response;
     } catch (error) {
       const endTime = Date.now();
-      const attributes = captureContent2 ? messagesToOtelAttributes(
-        params?.messages,
-        void 0,
-        params?.model,
-        void 0
-      ) : void 0;
-      if (attributes) {
-        attributes["error.message"] = error?.message;
-      }
       sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
@@ -1179,7 +1148,10 @@ function wrapOpenAI(client, sessionCtx) {
         duration_ms: endTime - startTime,
         status: "ERROR",
         error_message: error?.message,
-        attributes
+        attributes: {
+          "fallom.sdk_version": "2",
+          "fallom.method": "chat.completions.create"
+        }
       }).catch(() => {
       });
       throw error;
@@ -1206,21 +1178,26 @@ function wrapAnthropic(client, sessionCtx) {
     try {
       const response = await originalCreate(...args);
       const endTime = Date.now();
-      const attributes = captureContent2 ? messagesToOtelAttributes(
-        params?.messages,
-        { role: "assistant", content: response?.content?.[0]?.text || "" },
-        response?.model || params?.model,
-        response?.id
-      ) : {};
-      if (params?.system) {
-        attributes["gen_ai.system_prompt"] = params.system;
+      const attributes = {
+        "fallom.sdk_version": "2",
+        "fallom.method": "messages.create"
+      };
+      if (captureContent2) {
+        attributes["fallom.raw.request"] = JSON.stringify({
+          messages: params?.messages,
+          system: params?.system,
+          model: params?.model
+        });
+        attributes["fallom.raw.response"] = JSON.stringify({
+          text: response?.content?.[0]?.text,
+          finishReason: response?.stop_reason,
+          responseId: response?.id,
+          model: response?.model
+        });
       }
       if (response?.usage) {
         attributes["fallom.raw.usage"] = JSON.stringify(response.usage);
       }
-      if (response?.stop_reason) {
-        attributes["gen_ai.response.finish_reason"] = response.stop_reason;
-      }
       sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
@@ -1235,27 +1212,12 @@ function wrapAnthropic(client, sessionCtx) {
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "OK",
-        prompt_tokens: response?.usage?.input_tokens,
-        completion_tokens: response?.usage?.output_tokens,
-        total_tokens: (response?.usage?.input_tokens || 0) + (response?.usage?.output_tokens || 0),
-        attributes: Object.keys(attributes).length > 0 ? attributes : void 0
+        attributes
       }).catch(() => {
       });
       return response;
     } catch (error) {
       const endTime = Date.now();
-      const attributes = captureContent2 ? messagesToOtelAttributes(
-        params?.messages,
-        void 0,
-        params?.model,
-        void 0
-      ) : void 0;
-      if (attributes) {
-        attributes["error.message"] = error?.message;
-        if (params?.system) {
-          attributes["gen_ai.system_prompt"] = params.system;
-        }
-      }
       sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
@@ -1271,7 +1233,10 @@ function wrapAnthropic(client, sessionCtx) {
         duration_ms: endTime - startTime,
         status: "ERROR",
         error_message: error?.message,
-        attributes
+        attributes: {
+          "fallom.sdk_version": "2",
+          "fallom.method": "messages.create"
+        }
       }).catch(() => {
       });
       throw error;
@@ -1282,50 +1247,36 @@ function wrapAnthropic(client, sessionCtx) {
 
 // src/trace/wrappers/google-ai.ts
 function wrapGoogleAI(model, sessionCtx) {
-  const originalGenerate = model.generateContent.bind(model);
+  const originalGenerateContent = model.generateContent.bind(model);
   const ctx = sessionCtx;
   model.generateContent = async function(...args) {
     if (!isInitialized()) {
-      return originalGenerate(...args);
+      return originalGenerateContent(...args);
     }
     const traceCtx = getTraceContextStorage().getStore() || getFallbackTraceContext();
     const traceId = traceCtx?.traceId || generateHexId(32);
     const spanId = generateHexId(16);
     const parentSpanId = traceCtx?.parentSpanId;
+    const request = args[0];
     const startTime = Date.now();
     const captureContent2 = shouldCaptureContent();
     try {
-      const response = await originalGenerate(...args);
+      const response = await originalGenerateContent(...args);
       const endTime = Date.now();
-      const result = response?.response;
-      const usage = result?.usageMetadata;
-      const modelName = model?.model || "gemini";
-      const attributes = {};
+      const result = response?.response || response;
+      const attributes = {
+        "fallom.sdk_version": "2",
+        "fallom.method": "generateContent"
+      };
       if (captureContent2) {
-        attributes["gen_ai.request.model"] = modelName;
-        attributes["gen_ai.response.model"] = modelName;
-        const input = args[0];
-        if (typeof input === "string") {
-          attributes["gen_ai.prompt.0.role"] = "user";
-          attributes["gen_ai.prompt.0.content"] = input;
-        } else if (input?.contents) {
-          input.contents.forEach((content, i) => {
-            attributes[`gen_ai.prompt.${i}.role`] = content.role || "user";
-            attributes[`gen_ai.prompt.${i}.content`] = content.parts?.[0]?.text || JSON.stringify(content.parts);
-          });
-        }
-        const outputText = result?.text?.();
-        if (outputText) {
-          attributes["gen_ai.completion.0.role"] = "assistant";
-          attributes["gen_ai.completion.0.content"] = outputText;
-        }
-      }
-      if (usage) {
-        attributes["fallom.raw.usage"] = JSON.stringify(usage);
+        attributes["fallom.raw.request"] = JSON.stringify(request);
+        attributes["fallom.raw.response"] = JSON.stringify({
+          text: result?.text?.(),
+          candidates: result?.candidates
+        });
       }
-      const candidate = result?.candidates?.[0];
-      if (candidate?.finishReason) {
-        attributes["gen_ai.response.finish_reason"] = candidate.finishReason;
+      if (result?.usageMetadata) {
+        attributes["fallom.raw.usage"] = JSON.stringify(result.usageMetadata);
       }
       sendTrace({
         config_key: ctx.configKey,
@@ -1336,31 +1287,17 @@ function wrapGoogleAI(model, sessionCtx) {
         parent_span_id: parentSpanId,
         name: "generateContent",
         kind: "llm",
-        model: modelName,
+        model: model.model || "gemini",
         start_time: new Date(startTime).toISOString(),
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "OK",
-        prompt_tokens: usage?.promptTokenCount,
-        completion_tokens: usage?.candidatesTokenCount,
-        total_tokens: usage?.totalTokenCount,
-        attributes: Object.keys(attributes).length > 0 ? attributes : void 0
+        attributes
       }).catch(() => {
       });
       return response;
     } catch (error) {
       const endTime = Date.now();
-      const modelName = model?.model || "gemini";
-      const attributes = {};
-      if (captureContent2) {
-        attributes["gen_ai.request.model"] = modelName;
-        attributes["error.message"] = error?.message;
-        const input = args[0];
-        if (typeof input === "string") {
-          attributes["gen_ai.prompt.0.role"] = "user";
-          attributes["gen_ai.prompt.0.content"] = input;
-        }
-      }
       sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
@@ -1370,13 +1307,16 @@ function wrapGoogleAI(model, sessionCtx) {
         parent_span_id: parentSpanId,
         name: "generateContent",
         kind: "llm",
-        model: modelName,
+        model: model.model || "gemini",
         start_time: new Date(startTime).toISOString(),
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "ERROR",
         error_message: error?.message,
-        attributes: captureContent2 ? attributes : void 0
+        attributes: {
+          "fallom.sdk_version": "2",
+          "fallom.method": "generateContent"
+        }
       }).catch(() => {
       });
       throw error;
@@ -1385,35 +1325,6 @@ function wrapGoogleAI(model, sessionCtx) {
   return model;
 }
 
-// src/trace/wrappers/vercel-ai/utils.ts
-function extractUsageFromResult(result, directUsage) {
-  let usage = directUsage ?? result?.usage;
-  const isValidNumber = (v) => v !== null && v !== void 0 && !Number.isNaN(v);
-  let promptTokens = isValidNumber(usage?.promptTokens) ? usage.promptTokens : isValidNumber(usage?.inputTokens) ? usage.inputTokens : isValidNumber(usage?.prompt_tokens) ? usage.prompt_tokens : void 0;
-  let completionTokens = isValidNumber(usage?.completionTokens) ? usage.completionTokens : isValidNumber(usage?.outputTokens) ? usage.outputTokens : isValidNumber(usage?.completion_tokens) ? usage.completion_tokens : void 0;
-  let totalTokens = isValidNumber(usage?.totalTokens) ? usage.totalTokens : isValidNumber(usage?.total_tokens) ? usage.total_tokens : void 0;
-  let cost;
-  const orUsage = result?.experimental_providerMetadata?.openrouter?.usage;
-  if (orUsage) {
-    if (promptTokens === void 0 && isValidNumber(orUsage.promptTokens)) {
-      promptTokens = orUsage.promptTokens;
-    }
-    if (completionTokens === void 0 && isValidNumber(orUsage.completionTokens)) {
-      completionTokens = orUsage.completionTokens;
-    }
-    if (totalTokens === void 0 && isValidNumber(orUsage.totalTokens)) {
-      totalTokens = orUsage.totalTokens;
-    }
-    if (isValidNumber(orUsage.cost)) {
-      cost = orUsage.cost;
-    }
-  }
-  if (totalTokens === void 0 && (promptTokens !== void 0 || completionTokens !== void 0)) {
-    totalTokens = (promptTokens ?? 0) + (completionTokens ?? 0);
-  }
-  return { promptTokens, completionTokens, totalTokens, cost };
-}
-
 // src/trace/wrappers/vercel-ai/generate-text.ts
 function createGenerateTextWrapper(aiModule, sessionCtx, debug = false) {
   const ctx = sessionCtx;
@@ -1432,54 +1343,33 @@ function createGenerateTextWrapper(aiModule, sessionCtx, debug = false) {
       const result = await aiModule.generateText(...args);
       const endTime = Date.now();
       if (debug || isDebugMode()) {
-        console.log(
-          "\n\u{1F50D} [Fallom Debug] generateText result keys:",
-          Object.keys(result || {})
-        );
-        console.log(
-          "\u{1F50D} [Fallom Debug] result.usage:",
-          JSON.stringify(result?.usage, null, 2)
-        );
-        console.log(
-          "\u{1F50D} [Fallom Debug] result.experimental_providerMetadata:",
-          JSON.stringify(result?.experimental_providerMetadata, null, 2)
-        );
+        console.log("\n\u{1F50D} [Fallom Debug] generateText raw result:", JSON.stringify(result, null, 2));
       }
       const modelId = result?.response?.modelId || params?.model?.modelId || String(params?.model || "unknown");
-      const attributes = {};
+      const attributes = {
+        "fallom.sdk_version": "2",
+        "fallom.method": "generateText"
+      };
       if (captureContent2) {
-        attributes["gen_ai.request.model"] = modelId;
-        attributes["gen_ai.response.model"] = modelId;
-        if (params?.prompt) {
-          attributes["gen_ai.prompt.0.role"] = "user";
-          attributes["gen_ai.prompt.0.content"] = params.prompt;
-        }
-        if (params?.messages) {
-          params.messages.forEach((msg, i) => {
-            attributes[`gen_ai.prompt.${i}.role`] = msg.role;
-            attributes[`gen_ai.prompt.${i}.content`] = typeof msg.content === "string" ? msg.content : JSON.stringify(msg.content);
-          });
-        }
-        if (result?.text) {
-          attributes["gen_ai.completion.0.role"] = "assistant";
-          attributes["gen_ai.completion.0.content"] = result.text;
-        }
-        if (result?.response?.id) {
-          attributes["gen_ai.response.id"] = result.response.id;
-        }
+        attributes["fallom.raw.request"] = JSON.stringify({
+          prompt: params?.prompt,
+          messages: params?.messages,
+          system: params?.system,
+          model: modelId
+        });
+        attributes["fallom.raw.response"] = JSON.stringify({
+          text: result?.text,
+          finishReason: result?.finishReason,
+          responseId: result?.response?.id,
+          modelId: result?.response?.modelId
+        });
       }
       if (result?.usage) {
         attributes["fallom.raw.usage"] = JSON.stringify(result.usage);
       }
       if (result?.experimental_providerMetadata) {
-        attributes["fallom.raw.providerMetadata"] = JSON.stringify(
-          result.experimental_providerMetadata
-        );
-      }
-      if (result?.finishReason) {
-        attributes["gen_ai.response.finish_reason"] = result.finishReason;
+        attributes["fallom.raw.providerMetadata"] = JSON.stringify(result.experimental_providerMetadata);
       }
-      const usage = extractUsageFromResult(result);
       sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
@@ -1494,10 +1384,7 @@ function createGenerateTextWrapper(aiModule, sessionCtx, debug = false) {
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "OK",
-        prompt_tokens: usage.promptTokens,
-        completion_tokens: usage.completionTokens,
-        total_tokens: usage.totalTokens,
-        attributes: captureContent2 ? attributes : void 0
+        attributes
       }).catch(() => {
       });
       return result;
@@ -1518,7 +1405,17 @@ function createGenerateTextWrapper(aiModule, sessionCtx, debug = false) {
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "ERROR",
-        error_message: error?.message
+        error_message: error?.message,
+        attributes: {
+          "fallom.sdk_version": "2",
+          "fallom.method": "generateText",
+          "fallom.raw.request": JSON.stringify({
+            prompt: params?.prompt,
+            messages: params?.messages,
+            system: params?.system,
+            model: modelId
+          })
+        }
       }).catch(() => {
       });
       throw error;
@@ -1547,15 +1444,17 @@ function createStreamTextWrapper(aiModule, sessionCtx, debug = false) {
     let firstTokenTime = null;
     const modelId = params?.model?.modelId || String(params?.model || "unknown");
     if (result?.usage) {
-      result.usage.then(async (rawUsage) => {
+      Promise.all([
+        result.usage.catch(() => null),
+        result.text?.catch(() => null),
+        result.finishReason?.catch(() => null)
+      ]).then(async ([rawUsage, responseText, finishReason]) => {
         const endTime = Date.now();
         if (debug || isDebugMode()) {
-          console.log(
-            "\n\u{1F50D} [Fallom Debug] streamText usage:",
-            JSON.stringify(rawUsage, null, 2)
-          );
+          console.log("\n\u{1F50D} [Fallom Debug] streamText raw usage:", JSON.stringify(rawUsage, null, 2));
+          console.log("\u{1F50D} [Fallom Debug] streamText response text:", responseText?.slice(0, 100));
+          console.log("\u{1F50D} [Fallom Debug] streamText finish reason:", finishReason);
         }
-        log2("\u{1F4CA} streamText usage:", JSON.stringify(rawUsage, null, 2));
         let providerMetadata = result?.experimental_providerMetadata;
         if (providerMetadata && typeof providerMetadata.then === "function") {
           try {
@@ -1564,28 +1463,35 @@ function createStreamTextWrapper(aiModule, sessionCtx, debug = false) {
             providerMetadata = void 0;
           }
         }
-        const usage = extractUsageFromResult(
-          { experimental_providerMetadata: providerMetadata },
-          rawUsage
-        );
-        const attributes = {};
+        const attributes = {
+          "fallom.sdk_version": "2",
+          "fallom.method": "streamText",
+          "fallom.is_streaming": true
+        };
         if (captureContent2) {
-          attributes["gen_ai.request.model"] = modelId;
-          if (params?.prompt) {
-            attributes["gen_ai.prompt.0.role"] = "user";
-            attributes["gen_ai.prompt.0.content"] = params.prompt;
+          attributes["fallom.raw.request"] = JSON.stringify({
+            prompt: params?.prompt,
+            messages: params?.messages,
+            system: params?.system,
+            model: modelId
+          });
+          if (responseText || finishReason) {
+            attributes["fallom.raw.response"] = JSON.stringify({
+              text: responseText,
+              finishReason
+            });
           }
         }
-        if (firstTokenTime) {
-          attributes["gen_ai.time_to_first_token_ms"] = firstTokenTime - startTime;
-        }
         if (rawUsage) {
           attributes["fallom.raw.usage"] = JSON.stringify(rawUsage);
         }
         if (providerMetadata) {
           attributes["fallom.raw.providerMetadata"] = JSON.stringify(providerMetadata);
         }
-        const tracePayload = {
+        if (firstTokenTime) {
+          attributes["fallom.time_to_first_token_ms"] = firstTokenTime - startTime;
+        }
+        sendTrace({
           config_key: ctx.configKey,
           session_id: ctx.sessionId,
           customer_id: ctx.customerId,
@@ -1599,13 +1505,10 @@ function createStreamTextWrapper(aiModule, sessionCtx, debug = false) {
           end_time: new Date(endTime).toISOString(),
           duration_ms: endTime - startTime,
           status: "OK",
-          prompt_tokens: usage.promptTokens,
-          completion_tokens: usage.completionTokens,
-          total_tokens: usage.totalTokens,
           time_to_first_token_ms: firstTokenTime ? firstTokenTime - startTime : void 0,
-          attributes: captureContent2 ? attributes : void 0
-        };
-        sendTrace(tracePayload).catch(() => {
+          is_streaming: true,
+          attributes
+        }).catch(() => {
         });
       }).catch((error) => {
         const endTime = Date.now();
@@ -1624,7 +1527,12 @@ function createStreamTextWrapper(aiModule, sessionCtx, debug = false) {
           end_time: new Date(endTime).toISOString(),
           duration_ms: endTime - startTime,
           status: "ERROR",
-          error_message: error?.message
+          error_message: error?.message,
+          attributes: {
+            "fallom.sdk_version": "2",
+            "fallom.method": "streamText",
+            "fallom.is_streaming": true
+          }
         }).catch(() => {
         });
       });
@@ -1672,25 +1580,30 @@ function createGenerateObjectWrapper(aiModule, sessionCtx, debug = false) {
       const endTime = Date.now();
       if (debug || isDebugMode()) {
         console.log(
-          "\n\u{1F50D} [Fallom Debug] generateObject result keys:",
-          Object.keys(result || {})
-        );
-        console.log(
-          "\u{1F50D} [Fallom Debug] result.usage:",
-          JSON.stringify(result?.usage, null, 2)
+          "\n\u{1F50D} [Fallom Debug] generateObject raw result:",
+          JSON.stringify(result, null, 2)
         );
       }
       const modelId = result?.response?.modelId || params?.model?.modelId || String(params?.model || "unknown");
-      const attributes = {};
+      const attributes = {
+        "fallom.sdk_version": "2",
+        "fallom.method": "generateObject"
+      };
       if (captureContent2) {
-        attributes["gen_ai.request.model"] = modelId;
-        attributes["gen_ai.response.model"] = modelId;
-        if (result?.object) {
-          attributes["gen_ai.completion.0.role"] = "assistant";
-          attributes["gen_ai.completion.0.content"] = JSON.stringify(
-            result.object
-          );
-        }
+        attributes["fallom.raw.request"] = JSON.stringify({
+          prompt: params?.prompt,
+          messages: params?.messages,
+          system: params?.system,
+          model: modelId,
+          schema: params?.schema ? "provided" : void 0
+          // Don't send full schema, just note if present
+        });
+        attributes["fallom.raw.response"] = JSON.stringify({
+          object: result?.object,
+          finishReason: result?.finishReason,
+          responseId: result?.response?.id,
+          modelId: result?.response?.modelId
+        });
       }
       if (result?.usage) {
         attributes["fallom.raw.usage"] = JSON.stringify(result.usage);
@@ -1700,10 +1613,6 @@ function createGenerateObjectWrapper(aiModule, sessionCtx, debug = false) {
           result.experimental_providerMetadata
         );
       }
-      if (result?.finishReason) {
-        attributes["gen_ai.response.finish_reason"] = result.finishReason;
-      }
-      const usage = extractUsageFromResult(result);
       sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
@@ -1718,10 +1627,7 @@ function createGenerateObjectWrapper(aiModule, sessionCtx, debug = false) {
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "OK",
-        prompt_tokens: usage.promptTokens,
-        completion_tokens: usage.completionTokens,
-        total_tokens: usage.totalTokens,
-        attributes: captureContent2 ? attributes : void 0
+        attributes
       }).catch(() => {
       });
       return result;
@@ -1742,7 +1648,11 @@ function createGenerateObjectWrapper(aiModule, sessionCtx, debug = false) {
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "ERROR",
-        error_message: error?.message
+        error_message: error?.message,
+        attributes: {
+          "fallom.sdk_version": "2",
+          "fallom.method": "generateObject"
+        }
       }).catch(() => {
       });
       throw error;
@@ -1751,9 +1661,6 @@ function createGenerateObjectWrapper(aiModule, sessionCtx, debug = false) {
 }
 
 // src/trace/wrappers/vercel-ai/stream-object.ts
-function log3(...args) {
-  if (isDebugMode()) console.log("[Fallom]", ...args);
-}
 function createStreamObjectWrapper(aiModule, sessionCtx, debug = false) {
   const ctx = sessionCtx;
   return async (...args) => {
@@ -1761,7 +1668,6 @@ function createStreamObjectWrapper(aiModule, sessionCtx, debug = false) {
     const startTime = Date.now();
     const captureContent2 = shouldCaptureContent();
     const result = await aiModule.streamObject(...args);
-    log3("\u{1F50D} streamObject result keys:", Object.keys(result || {}));
     if (!isInitialized()) {
       return result;
     }
@@ -1769,18 +1675,19 @@ function createStreamObjectWrapper(aiModule, sessionCtx, debug = false) {
     const traceId = traceCtx?.traceId || generateHexId(32);
     const spanId = generateHexId(16);
     const parentSpanId = traceCtx?.parentSpanId;
-    let firstTokenTime = null;
     const modelId = params?.model?.modelId || String(params?.model || "unknown");
     if (result?.usage) {
-      result.usage.then(async (rawUsage) => {
+      Promise.all([
+        result.usage.catch(() => null),
+        result.object?.catch(() => null),
+        result.finishReason?.catch(() => null)
+      ]).then(async ([rawUsage, responseObject, finishReason]) => {
         const endTime = Date.now();
         if (debug || isDebugMode()) {
-          console.log(
-            "\n\u{1F50D} [Fallom Debug] streamObject usage:",
-            JSON.stringify(rawUsage, null, 2)
-          );
+          console.log("\n\u{1F50D} [Fallom Debug] streamObject raw usage:", JSON.stringify(rawUsage, null, 2));
+          console.log("\u{1F50D} [Fallom Debug] streamObject response object:", JSON.stringify(responseObject)?.slice(0, 100));
+          console.log("\u{1F50D} [Fallom Debug] streamObject finish reason:", finishReason);
         }
-        log3("\u{1F4CA} streamObject usage:", JSON.stringify(rawUsage, null, 2));
         let providerMetadata = result?.experimental_providerMetadata;
         if (providerMetadata && typeof providerMetadata.then === "function") {
           try {
@@ -1789,16 +1696,25 @@ function createStreamObjectWrapper(aiModule, sessionCtx, debug = false) {
             providerMetadata = void 0;
           }
         }
-        const usage = extractUsageFromResult(
-          { experimental_providerMetadata: providerMetadata },
-          rawUsage
-        );
-        const attributes = {};
+        const attributes = {
+          "fallom.sdk_version": "2",
+          "fallom.method": "streamObject",
+          "fallom.is_streaming": true
+        };
         if (captureContent2) {
-          attributes["gen_ai.request.model"] = modelId;
-        }
-        if (firstTokenTime) {
-          attributes["gen_ai.time_to_first_token_ms"] = firstTokenTime - startTime;
+          attributes["fallom.raw.request"] = JSON.stringify({
+            prompt: params?.prompt,
+            messages: params?.messages,
+            system: params?.system,
+            model: modelId,
+            schema: params?.schema ? "provided" : void 0
+          });
+          if (responseObject || finishReason) {
+            attributes["fallom.raw.response"] = JSON.stringify({
+              object: responseObject,
+              finishReason
+            });
+          }
         }
         if (rawUsage) {
           attributes["fallom.raw.usage"] = JSON.stringify(rawUsage);
@@ -1820,10 +1736,8 @@ function createStreamObjectWrapper(aiModule, sessionCtx, debug = false) {
           end_time: new Date(endTime).toISOString(),
           duration_ms: endTime - startTime,
           status: "OK",
-          prompt_tokens: usage.promptTokens,
-          completion_tokens: usage.completionTokens,
-          total_tokens: usage.totalTokens,
-          attributes: captureContent2 ? attributes : void 0
+          is_streaming: true,
+          attributes
         }).catch(() => {
         });
       }).catch((error) => {
@@ -1842,31 +1756,16 @@ function createStreamObjectWrapper(aiModule, sessionCtx, debug = false) {
           end_time: new Date(endTime).toISOString(),
           duration_ms: endTime - startTime,
           status: "ERROR",
-          error_message: error?.message
+          error_message: error?.message,
+          attributes: {
+            "fallom.sdk_version": "2",
+            "fallom.method": "streamObject",
+            "fallom.is_streaming": true
+          }
         }).catch(() => {
         });
       });
     }
-    if (result?.partialObjectStream) {
-      const originalStream = result.partialObjectStream;
-      const wrappedStream = (async function* () {
-        for await (const chunk of originalStream) {
-          if (!firstTokenTime) {
-            firstTokenTime = Date.now();
-            log3("\u23F1\uFE0F Time to first token:", firstTokenTime - startTime, "ms");
-          }
-          yield chunk;
-        }
-      })();
-      return new Proxy(result, {
-        get(target, prop) {
-          if (prop === "partialObjectStream") {
-            return wrappedStream;
-          }
-          return target[prop];
-        }
-      });
-    }
     return result;
   };
 }
@@ -1885,105 +1784,69 @@ function wrapAISDK(ai, sessionCtx, options) {
 // src/trace/wrappers/mastra.ts
 function wrapMastraAgent(agent, sessionCtx) {
   const originalGenerate = agent.generate.bind(agent);
-  const agentName = agent.name || "MastraAgent";
   const ctx = sessionCtx;
   agent.generate = async function(...args) {
     if (!isInitialized()) {
       return originalGenerate(...args);
     }
-    const traceId = generateHexId(32);
+    const traceCtx = getTraceContextStorage().getStore() || getFallbackTraceContext();
+    const traceId = traceCtx?.traceId || generateHexId(32);
     const spanId = generateHexId(16);
+    const parentSpanId = traceCtx?.parentSpanId;
+    const input = args[0];
     const startTime = Date.now();
-    const messages = args[0] || [];
+    const captureContent2 = shouldCaptureContent();
     try {
       const result = await originalGenerate(...args);
       const endTime = Date.now();
-      const model = result?.model?.modelId || "unknown";
-      const toolCalls = [];
-      if (result?.steps?.length) {
-        for (const step of result.steps) {
-          if (step.toolCalls?.length) {
-            for (let i = 0; i < step.toolCalls.length; i++) {
-              const tc = step.toolCalls[i];
-              const tr = step.toolResults?.[i];
-              toolCalls.push({
-                name: tc.toolName,
-                arguments: tc.args,
-                result: tr?.result
-              });
-            }
-          }
-        }
-      }
       const attributes = {
-        "gen_ai.system": "Mastra",
-        "gen_ai.request.model": model,
-        "gen_ai.response.model": model,
-        "fallom.source": "mastra-agent",
-        "llm.request.type": "chat"
+        "fallom.sdk_version": "2",
+        "fallom.method": "agent.generate",
+        "fallom.agent_name": agent.name || "unknown"
       };
-      if (Array.isArray(messages)) {
-        messages.forEach((msg, i) => {
-          attributes[`gen_ai.prompt.${i}.role`] = msg.role || "user";
-          attributes[`gen_ai.prompt.${i}.content`] = typeof msg.content === "string" ? msg.content : JSON.stringify(msg.content);
-        });
-      }
-      if (result?.text) {
-        attributes["gen_ai.completion.0.role"] = "assistant";
-        attributes["gen_ai.completion.0.content"] = result.text;
-        attributes["gen_ai.completion.0.finish_reason"] = "stop";
-      }
-      if (toolCalls.length > 0) {
-        attributes["fallom.tool_calls"] = JSON.stringify(toolCalls);
-        toolCalls.forEach((tc, i) => {
-          attributes[`gen_ai.completion.0.tool_calls.${i}.name`] = tc.name;
-          attributes[`gen_ai.completion.0.tool_calls.${i}.type`] = "function";
-          attributes[`gen_ai.completion.0.tool_calls.${i}.arguments`] = JSON.stringify(tc.arguments);
-        });
-      }
-      if (result?.usage) {
-        attributes["gen_ai.usage.prompt_tokens"] = result.usage.promptTokens;
-        attributes["gen_ai.usage.completion_tokens"] = result.usage.completionTokens;
-        attributes["llm.usage.total_tokens"] = result.usage.totalTokens;
+      if (captureContent2) {
+        attributes["fallom.raw.request"] = JSON.stringify(input);
+        attributes["fallom.raw.response"] = JSON.stringify(result);
       }
-      const traceData = {
+      sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
         customer_id: ctx.customerId,
         trace_id: traceId,
         span_id: spanId,
-        name: `mastra.${agentName}.generate`,
-        kind: "client",
-        model,
+        parent_span_id: parentSpanId,
+        name: `agent.${agent.name || "unknown"}.generate`,
+        kind: "agent",
         start_time: new Date(startTime).toISOString(),
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "OK",
-        prompt_tokens: result?.usage?.promptTokens,
-        completion_tokens: result?.usage?.completionTokens,
-        total_tokens: result?.usage?.totalTokens,
         attributes
-      };
-      sendTrace(traceData).catch(() => {
+      }).catch(() => {
       });
       return result;
     } catch (error) {
       const endTime = Date.now();
-      const traceData = {
+      sendTrace({
         config_key: ctx.configKey,
         session_id: ctx.sessionId,
         customer_id: ctx.customerId,
         trace_id: traceId,
         span_id: spanId,
-        name: `mastra.${agentName}.generate`,
-        kind: "client",
+        parent_span_id: parentSpanId,
+        name: `agent.${agent.name || "unknown"}.generate`,
+        kind: "agent",
         start_time: new Date(startTime).toISOString(),
         end_time: new Date(endTime).toISOString(),
         duration_ms: endTime - startTime,
         status: "ERROR",
-        error_message: error instanceof Error ? error.message : String(error)
-      };
-      sendTrace(traceData).catch(() => {
+        error_message: error?.message,
+        attributes: {
+          "fallom.sdk_version": "2",
+          "fallom.method": "agent.generate",
+          "fallom.agent_name": agent.name || "unknown"
+        }
+      }).catch(() => {
       });
       throw error;
     }
@@ -2023,6 +1886,9 @@ var FallomSession = class {
   /**
    * Wrap a Vercel AI SDK model to trace all calls (PostHog style).
    * Returns the same model type with tracing injected.
+   * 
+   * Note: This only captures tokens/timing, not prompt/completion content.
+   * Use wrapAISDK for full content tracing.
    */
   traceModel(model) {
     const ctx = this.ctx;
@@ -2048,17 +1914,18 @@ var FallomSession = class {
             trace_id: traceId,
             span_id: spanId,
             parent_span_id: traceCtx?.parentSpanId,
-            name: "generateText",
+            name: "doGenerate",
             kind: "llm",
             model: modelId,
             start_time: new Date(startTime).toISOString(),
             end_time: new Date(endTime).toISOString(),
             duration_ms: endTime - startTime,
             status: "OK",
-            prompt_tokens: usage?.promptTokens,
-            completion_tokens: usage?.completionTokens,
-            total_tokens: usage?.totalTokens,
-            attributes: shouldCaptureContent() && usage ? { "fallom.raw.usage": JSON.stringify(usage) } : void 0
+            attributes: {
+              "fallom.sdk_version": "2",
+              "fallom.method": "traceModel.doGenerate",
+              ...usage ? { "fallom.raw.usage": JSON.stringify(usage) } : {}
+            }
           }).catch(() => {
           });
           return result;
@@ -2071,14 +1938,15 @@ var FallomSession = class {
             trace_id: traceId,
             span_id: spanId,
             parent_span_id: traceCtx?.parentSpanId,
-            name: "generateText",
+            name: "doGenerate",
             kind: "llm",
             model: model.modelId || "unknown",
             start_time: new Date(startTime).toISOString(),
             end_time: new Date(endTime).toISOString(),
             duration_ms: endTime - startTime,
             status: "ERROR",
-            error_message: error instanceof Error ? error.message : String(error)
+            error_message: error instanceof Error ? error.message : String(error),
+            attributes: { "fallom.sdk_version": "2", "fallom.method": "traceModel.doGenerate" }
           }).catch(() => {
           });
           throw error;
@@ -2103,14 +1971,19 @@ var FallomSession = class {
             trace_id: traceId,
             span_id: spanId,
             parent_span_id: traceCtx?.parentSpanId,
-            name: "streamText",
+            name: "doStream",
             kind: "llm",
             model: modelId,
             start_time: new Date(startTime).toISOString(),
             end_time: new Date(Date.now()).toISOString(),
             duration_ms: Date.now() - startTime,
             status: "OK",
-            is_streaming: true
+            is_streaming: true,
+            attributes: {
+              "fallom.sdk_version": "2",
+              "fallom.method": "traceModel.doStream",
+              "fallom.is_streaming": true
+            }
           }).catch(() => {
           });
           return result;
@@ -2122,7 +1995,7 @@ var FallomSession = class {
             trace_id: traceId,
             span_id: spanId,
             parent_span_id: traceCtx?.parentSpanId,
-            name: "streamText",
+            name: "doStream",
             kind: "llm",
             model: modelId,
             start_time: new Date(startTime).toISOString(),
@@ -2130,7 +2003,12 @@ var FallomSession = class {
             duration_ms: Date.now() - startTime,
             status: "ERROR",
             error_message: error instanceof Error ? error.message : String(error),
-            is_streaming: true
+            is_streaming: true,
+            attributes: {
+              "fallom.sdk_version": "2",
+              "fallom.method": "traceModel.doStream",
+              "fallom.is_streaming": true
+            }
           }).catch(() => {
           });
           throw error;
@@ -2186,7 +2064,7 @@ var promptCache = /* @__PURE__ */ new Map();
 var promptABCache = /* @__PURE__ */ new Map();
 var promptContext = null;
 var SYNC_TIMEOUT2 = 2e3;
-function log5(msg) {
+function log4(msg) {
   if (debugMode3) {
     console.log(`[Fallom Prompts] ${msg}`);
   }
@@ -2289,10 +2167,10 @@ async function get2(promptKey, options = {}) {
   const { variables, version, debug = false } = options;
   debugMode3 = debug;
   ensureInit2();
-  log5(`get() called: promptKey=${promptKey}`);
+  log4(`get() called: promptKey=${promptKey}`);
   let promptData = promptCache.get(promptKey);
   if (!promptData) {
-    log5("Not in cache, fetching...");
+    log4("Not in cache, fetching...");
     await fetchPrompts(SYNC_TIMEOUT2);
     promptData = promptCache.get(promptKey);
   }
@@ -2314,7 +2192,7 @@ async function get2(promptKey, options = {}) {
     promptKey,
     promptVersion: targetVersion
   });
-  log5(`\u2705 Got prompt: ${promptKey} v${targetVersion}`);
+  log4(`\u2705 Got prompt: ${promptKey} v${targetVersion}`);
   return {
     key: promptKey,
     version: targetVersion,
@@ -2326,10 +2204,10 @@ async function getAB(abTestKey, sessionId, options = {}) {
   const { variables, debug = false } = options;
   debugMode3 = debug;
   ensureInit2();
-  log5(`getAB() called: abTestKey=${abTestKey}, sessionId=${sessionId}`);
+  log4(`getAB() called: abTestKey=${abTestKey}, sessionId=${sessionId}`);
   let abData = promptABCache.get(abTestKey);
   if (!abData) {
-    log5("Not in cache, fetching...");
+    log4("Not in cache, fetching...");
     await fetchPromptABTests(SYNC_TIMEOUT2);
     abData = promptABCache.get(abTestKey);
   }
@@ -2344,8 +2222,8 @@ async function getAB(abTestKey, sessionId, options = {}) {
     throw new Error(`Prompt A/B test '${abTestKey}' has no current version.`);
   }
   const { variants } = versionData;
-  log5(`A/B test '${abTestKey}' has ${variants?.length ?? 0} variants`);
-  log5(`Version data: ${JSON.stringify(versionData, null, 2)}`);
+  log4(`A/B test '${abTestKey}' has ${variants?.length ?? 0} variants`);
+  log4(`Version data: ${JSON.stringify(versionData, null, 2)}`);
   if (!variants || variants.length === 0) {
     throw new Error(
       `Prompt A/B test '${abTestKey}' has no variants configured.`
@@ -2391,7 +2269,7 @@ async function getAB(abTestKey, sessionId, options = {}) {
     abTestKey,
     variantIndex: selectedIndex
   });
-  log5(
+  log4(
     `\u2705 Got prompt from A/B: ${promptKey} v${targetVersion} (variant ${selectedIndex})`
   );
   return {