@threaded/ai 1.0.24 → 1.0.26

package/.claude/settings.local.json

@@ -0,0 +1,15 @@
+{
+  "permissions": {
+    "allow": [
+      "WebSearch",
+      "WebFetch(domain:platform.openai.com)",
+      "WebFetch(domain:docs.anthropic.com)",
+      "WebFetch(domain:ai.google.dev)",
+      "WebFetch(domain:docs.x.ai)",
+      "WebFetch(domain:docs.cloud.google.com)",
+      "WebFetch(domain:latenode.com)",
+      "WebFetch(domain:github.com)",
+      "WebFetch(domain:docs.aimlapi.com)"
+    ]
+  }
+}

package/dist/index.cjs

@@ -33,6 +33,7 @@ __export(index_exports, {
   IMAGE_EDIT_MODEL_SCHEMA: () => IMAGE_EDIT_MODEL_SCHEMA,
   IMAGE_MODEL_SCHEMA: () => IMAGE_MODEL_SCHEMA,
   Inherit: () => Inherit,
+  addUsage: () => addUsage,
   appendToLastRequest: () => appendToLastRequest,
   compose: () => compose,
   convertMCPSchemaToToolSchema: () => convertMCPSchemaToToolSchema,
@@ -222,6 +223,11 @@ var maxCalls = (toolConfig, maxCalls2) => ({
   ...toolConfig,
   _maxCalls: maxCalls2
 });
+var addUsage = (existing, promptTokens, completionTokens, totalTokens) => ({
+  promptTokens: (existing?.promptTokens || 0) + promptTokens,
+  completionTokens: (existing?.completionTokens || 0) + completionTokens,
+  totalTokens: (existing?.totalTokens || 0) + totalTokens
+});
 
 // src/embed.ts
 var modelCache = /* @__PURE__ */ new Map();
@@ -533,7 +539,8 @@ var callOpenAI = async (config, ctx) => {
   const body = {
     model: model2,
     messages,
-    stream: !!ctx.stream
+    stream: !!ctx.stream,
+    ...ctx.stream && { stream_options: { include_usage: true } }
   };
   if (schema) {
     body.response_format = {
@@ -578,7 +585,8 @@ var callOpenAI = async (config, ctx) => {
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage: addUsage(ctx.usage, data.usage?.prompt_tokens || 0, data.usage?.completion_tokens || 0, data.usage?.total_tokens || 0)
   };
 };
 var handleOpenAIStream = async (response, ctx) => {
@@ -587,6 +595,7 @@ var handleOpenAIStream = async (response, ctx) => {
   let fullContent = "";
   let toolCalls = [];
   let buffer = "";
+  let streamUsage = null;
   try {
     while (true) {
       if (ctx.abortSignal?.aborted) {
@@ -604,6 +613,9 @@ var handleOpenAIStream = async (response, ctx) => {
           if (!data) continue;
           try {
             const parsed = JSON.parse(data);
+            if (parsed.usage) {
+              streamUsage = parsed.usage;
+            }
             const delta = parsed.choices?.[0]?.delta;
             if (delta?.content) {
               fullContent += delta.content;
@@ -629,10 +641,15 @@ var handleOpenAIStream = async (response, ctx) => {
   if (toolCalls.length > 0) {
     msg.tool_calls = toolCalls;
   }
+  const usage = addUsage(ctx.usage, streamUsage?.prompt_tokens || 0, streamUsage?.completion_tokens || 0, streamUsage?.total_tokens || 0);
+  if (ctx.stream && streamUsage) {
+    ctx.stream({ type: "usage", usage });
+  }
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage
   };
 };
 
@@ -768,10 +785,13 @@ Return only the JSON object, no other text or formatting.`;
       }
     ];
   }
+  const inputTokens = data.usage?.input_tokens || 0;
+  const outputTokens = data.usage?.output_tokens || 0;
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage: addUsage(ctx.usage, inputTokens, outputTokens, inputTokens + outputTokens)
   };
 };
 var handleAnthropicStream = async (response, ctx) => {
@@ -780,6 +800,8 @@ var handleAnthropicStream = async (response, ctx) => {
   let fullContent = "";
   const toolCalls = [];
   let buffer = "";
+  let inputTokens = 0;
+  let outputTokens = 0;
   try {
     while (true) {
       if (ctx.abortSignal?.aborted) {
@@ -796,6 +818,12 @@ var handleAnthropicStream = async (response, ctx) => {
           if (!data) continue;
           try {
             const parsed = JSON.parse(data);
+            if (parsed.type === "message_start" && parsed.message?.usage) {
+              inputTokens = parsed.message.usage.input_tokens || 0;
+            }
+            if (parsed.type === "message_delta" && parsed.usage) {
+              outputTokens = parsed.usage.output_tokens || 0;
+            }
             if (parsed.type === "content_block_delta" && parsed.delta?.text) {
               fullContent += parsed.delta.text;
               if (ctx.stream) {
@@ -835,10 +863,15 @@ var handleAnthropicStream = async (response, ctx) => {
   if (toolCalls.length > 0) {
     msg.tool_calls = toolCalls.map(({ index, ...tc }) => tc);
   }
+  const usage = addUsage(ctx.usage, inputTokens, outputTokens, inputTokens + outputTokens);
+  if (ctx.stream && (inputTokens || outputTokens)) {
+    ctx.stream({ type: "usage", usage });
+  }
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage
   };
 };
 
@@ -991,10 +1024,12 @@ var callGoogle = async (config, ctx) => {
   if (toolCalls.length > 0) {
     msg.tool_calls = toolCalls;
   }
+  const um = data.usageMetadata;
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage: addUsage(ctx.usage, um?.promptTokenCount || 0, um?.candidatesTokenCount || 0, um?.totalTokenCount || 0)
   };
 };
 var handleGoogleStream = async (response, ctx) => {
@@ -1003,6 +1038,7 @@ var handleGoogleStream = async (response, ctx) => {
   let fullContent = "";
   const toolCalls = [];
   let buffer = "";
+  let usageMetadata = null;
   try {
     while (true) {
       if (ctx.abortSignal?.aborted) {
@@ -1019,6 +1055,9 @@ var handleGoogleStream = async (response, ctx) => {
           if (!data) continue;
           try {
             const parsed = JSON.parse(data);
+            if (parsed.usageMetadata) {
+              usageMetadata = parsed.usageMetadata;
+            }
             const candidate = parsed.candidates?.[0];
             const parts = candidate?.content?.parts || [];
             for (const part of parts) {
@@ -1058,10 +1097,16 @@ var handleGoogleStream = async (response, ctx) => {
   if (toolCalls.length > 0) {
     msg.tool_calls = toolCalls;
   }
+  const um = usageMetadata;
+  const usage = addUsage(ctx.usage, um?.promptTokenCount || 0, um?.candidatesTokenCount || 0, um?.totalTokenCount || 0);
+  if (ctx.stream && um) {
+    ctx.stream({ type: "usage", usage });
+  }
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage
   };
 };
 
@@ -1110,7 +1155,8 @@ var callXAI = async (config, ctx) => {
   const body = {
     model: model2,
     messages,
-    stream: !!ctx.stream
+    stream: !!ctx.stream,
+    ...ctx.stream && { stream_options: { include_usage: true } }
   };
   if (schema) {
     body.response_format = {
@@ -1155,7 +1201,8 @@ var callXAI = async (config, ctx) => {
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage: addUsage(ctx.usage, data.usage?.prompt_tokens || 0, data.usage?.completion_tokens || 0, data.usage?.total_tokens || 0)
   };
 };
 var handleXAIStream = async (response, ctx) => {
@@ -1164,6 +1211,7 @@ var handleXAIStream = async (response, ctx) => {
   let fullContent = "";
   let toolCalls = [];
   let buffer = "";
+  let streamUsage = null;
   try {
     while (true) {
       if (ctx.abortSignal?.aborted) {
@@ -1181,6 +1229,9 @@ var handleXAIStream = async (response, ctx) => {
           if (!data) continue;
           try {
             const parsed = JSON.parse(data);
+            if (parsed.usage) {
+              streamUsage = parsed.usage;
+            }
             const delta = parsed.choices?.[0]?.delta;
             if (delta?.content) {
               fullContent += delta.content;
@@ -1206,10 +1257,15 @@ var handleXAIStream = async (response, ctx) => {
   if (toolCalls.length > 0) {
     msg.tool_calls = toolCalls;
   }
+  const usage = addUsage(ctx.usage, streamUsage?.prompt_tokens || 0, streamUsage?.completion_tokens || 0, streamUsage?.total_tokens || 0);
+  if (ctx.stream && streamUsage) {
+    ctx.stream({ type: "usage", usage });
+  }
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage
   };
 };
 
@@ -1242,7 +1298,8 @@ var callLocal = async (config, ctx) => {
   const body = {
     model: model2,
     messages,
-    stream: !!ctx.stream
+    stream: !!ctx.stream,
+    ...ctx.stream && { stream_options: { include_usage: true } }
   };
   if (schema) {
     body.response_format = {
@@ -1290,7 +1347,8 @@ var callLocal = async (config, ctx) => {
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage: addUsage(ctx.usage, data.usage?.prompt_tokens || 0, data.usage?.completion_tokens || 0, data.usage?.total_tokens || 0)
   };
 };
 var handleLocalStream = async (response, ctx) => {
@@ -1299,6 +1357,7 @@ var handleLocalStream = async (response, ctx) => {
   let fullContent = "";
   let toolCalls = [];
   let buffer = "";
+  let streamUsage = null;
   try {
     while (true) {
       if (ctx.abortSignal?.aborted) {
@@ -1316,6 +1375,9 @@ var handleLocalStream = async (response, ctx) => {
           if (!data) continue;
           try {
             const parsed = JSON.parse(data);
+            if (parsed.usage) {
+              streamUsage = parsed.usage;
+            }
             const delta = parsed.choices?.[0]?.delta;
             if (delta?.content) {
               fullContent += delta.content;
@@ -1341,10 +1403,15 @@ var handleLocalStream = async (response, ctx) => {
   if (toolCalls.length > 0) {
     msg.tool_calls = toolCalls;
   }
+  const usage = addUsage(ctx.usage, streamUsage?.prompt_tokens || 0, streamUsage?.completion_tokens || 0, streamUsage?.total_tokens || 0);
+  if (ctx.stream && streamUsage) {
+    ctx.stream({ type: "usage", usage });
+  }
   return {
     ...ctx,
     lastResponse: msg,
-    history: [...ctx.history, msg]
+    history: [...ctx.history, msg],
+    usage
   };
 };
 
@@ -1642,7 +1709,12 @@ var createThread = (id, store) => {
       };
       const finalContext = await (workflow || model())(initialContext);
       if (options?.abortSignal?.aborted) {
-        return finalContext;
+        const abortedHistory = [
+          ...initialContext.history,
+          { role: "assistant", content: "[Response interrupted]" }
+        ];
+        await store.set(id, abortedHistory);
+        return { ...finalContext, history: abortedHistory };
       }
       await store.set(id, finalContext.history);
       return finalContext;
@@ -1856,6 +1928,7 @@ var scopeContext = (config, ctx) => {
   }
   scopedCtx.stream = ctx.stream;
   scopedCtx.abortSignal = ctx.abortSignal;
+  scopedCtx.usage = ctx.usage;
   if (config.tools) {
     const toolDefinitions = config.tools.map(toolConfigToToolDefinition);
     const toolExecutors = config.tools.reduce(
@@ -1909,7 +1982,8 @@ var scope = (config, ...steps) => {
       history: config.silent ? ctx.history : scopedCtx.history,
       lastResponse: config.silent ? ctx.lastResponse : scopedCtx.lastResponse,
       lastRequest: config.silent ? ctx.lastRequest : scopedCtx.lastRequest,
-      stopReason: config.silent ? ctx.stopReason : scopedCtx.stopReason
+      stopReason: config.silent ? ctx.stopReason : scopedCtx.stopReason,
+      usage: scopedCtx.usage
     };
   };
 };
@@ -1971,6 +2045,7 @@ var rateLimited = (config) => (fn) => {
   IMAGE_EDIT_MODEL_SCHEMA,
   IMAGE_MODEL_SCHEMA,
   Inherit,
+  addUsage,
   appendToLastRequest,
   compose,
   convertMCPSchemaToToolSchema,