llmz 0.2.2 → 0.3.0

package/dist/{execute-4IL4QPAR.cjs → execute-2H4F7GFE.cjs}

@@ -2,7 +2,7 @@
 
 
 
-var _chunk2EZSVF5Xcjs = require('./chunk-2EZSVF5X.cjs');
+var _chunk3UZKLXADcjs = require('./chunk-3UZKLXAD.cjs');
 
 
 
@@ -11,13 +11,14 @@ var _chunk2EZSVF5Xcjs = require('./chunk-2EZSVF5X.cjs');
 
 
 
-var _chunk5XPEVOOWcjs = require('./chunk-5XPEVOOW.cjs');
+var _chunkKG5E3UZZcjs = require('./chunk-KG5E3UZZ.cjs');
 
 
 
-var _chunkC2KND2PScjs = require('./chunk-C2KND2PS.cjs');
-require('./chunk-S46T73HR.cjs');
-require('./chunk-MHVORH33.cjs');
+var _chunkOVRDJMZJcjs = require('./chunk-OVRDJMZJ.cjs');
+require('./chunk-RGUG3UMF.cjs');
+require('./chunk-PZE6PIPG.cjs');
+require('./chunk-HMMM32RG.cjs');
 
 
 
@@ -31,19 +32,19 @@ var _chunkPK72FAKDcjs = require('./chunk-PK72FAKD.cjs');
 
 
 var _chunkKIN7Y247cjs = require('./chunk-KIN7Y247.cjs');
-require('./chunk-D6PU5YAS.cjs');
+require('./chunk-DDSJTR6Q.cjs');
 require('./chunk-I56MPOVG.cjs');
 
 
 var _chunk5ZF6NTYIcjs = require('./chunk-5ZF6NTYI.cjs');
 
 
-var _chunkYQT5G77Wcjs = require('./chunk-YQT5G77W.cjs');
+var _chunkHEAQAEFZcjs = require('./chunk-HEAQAEFZ.cjs');
 
 
 
 
-var _chunkFPLSJF4Scjs = require('./chunk-FPLSJF4S.cjs');
+var _chunkGOWPACSKcjs = require('./chunk-GOWPACSK.cjs');
 
 
 
@@ -134,12 +135,12 @@ var generateCode = async ({
   onRunStart,
   onRunComplete
 }) => {
-  var _a, _b, _c, _d;
+  var _a, _b;
   const startedAt = Date.now();
   const traces = iteration.traces;
   const modelRef = Array.isArray(iteration.model) ? iteration.model[0] : iteration.model;
   const model = await cognitive.getModelDetails(modelRef).catch((thrown) => {
-    throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`Failed to fetch model details for model "${modelRef}": ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, thrown)}`);
+    throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`Failed to fetch model details for model "${modelRef}": ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, thrown)}`);
   });
   let modelLimit = Math.max(model.input.maxTokens, 8e3);
   if (ctx.maxTokens) {
@@ -151,34 +152,32 @@ var generateCode = async ({
   }
   let messages;
   try {
-    messages = _chunkYQT5G77Wcjs.truncateWrappedContent.call(void 0, {
+    messages = _chunkHEAQAEFZcjs.truncateWrappedContent.call(void 0, {
       messages: iteration.messages,
       tokenLimit: modelLimit - responseLengthBuffer,
       throwOnFailure: true
     }).filter((x) => typeof x.content !== "string" || x.content.trim().length > 0);
   } catch (thrown) {
     const cap = ctx.maxTokens ? ` (context window capped at ${modelLimit} tokens by options.maxTokens \u2014 consider raising or removing it)` : ` (model context window: ${modelLimit} tokens)`;
-    throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`The prompt does not fit in the context window${cap}: ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, thrown)}`);
+    throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`The prompt does not fit in the context window${cap}: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, thrown)}`);
   }
   iteration.messages = messages;
   traces.push({
     type: "llm_call_started",
     started_at: startedAt,
     ended_at: startedAt,
-    model: model.ref
+    model: model.id
   });
   const input = {
-    signal: controller.signal,
-    systemPrompt: (_a = messages.find((x) => x.role === "system")) == null ? void 0 : _a.content,
     model: iteration.model,
     temperature: iteration.temperature,
     responseFormat: "text",
     reasoningEffort: iteration.reasoningEffort,
-    messages: messages.filter((x) => x.role !== "system"),
+    messages,
     stopSequences: ctx.version.getStopTokens(),
     meta: metadata ? { metadata } : void 0
   };
-  let output;
+  let responseMetadata;
   let raw;
   let assistantResponse;
   let timeToFirstToken;
@@ -223,18 +222,20 @@ var generateCode = async ({
       }
     }
   };
-  if (typeof cognitive.generateContentStream === "function") {
-    const parser = new (0, _chunkC2KND2PScjs.StreamingMessageParser)();
+  if (typeof cognitive.generateTextStream === "function") {
+    const parser = new (0, _chunkOVRDJMZJcjs.StreamingMessageParser)();
     const fence = new LeadingFenceFilter();
     const streamController = createJoinedAbortController([controller.signal]);
     const requestedAt = Date.now();
-    const stream = cognitive.generateContentStream({
-      ...input,
-      signal: streamController.signal,
-      // Passed through to the cognitive v2 request: fall back to the next
-      // model/provider when the first token takes too long
-      ...ctx.maxTimeToFirstToken ? { options: { maxTimeToFirstToken: ctx.maxTimeToFirstToken } } : {}
-    });
+    const stream = cognitive.generateTextStream(
+      {
+        ...input,
+        // Passed through to the cognitive request: fall back to the next
+        // model/provider when the first token takes too long
+        ...ctx.maxTimeToFirstToken ? { options: { maxTimeToFirstToken: ctx.maxTimeToFirstToken } } : {}
+      },
+      { signal: streamController.signal }
+    );
     const inactivityTimeout = Math.max(STREAM_INACTIVITY_TIMEOUT, _nullishCoalesce(ctx.maxTimeToFirstToken, () => ( 0)));
     const nextChunk = async () => {
       let timer;
@@ -256,12 +257,14 @@ var generateCode = async ({
       try {
         chunk = await nextChunk();
       } catch (thrown) {
-        throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`LLM generation failed: ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, thrown)}`);
+        throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`LLM generation failed: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, thrown)}`);
       }
       if (chunk.done) {
-        output = chunk.value;
         break;
       }
+      if ((_a = chunk.value) == null ? void 0 : _a.metadata) {
+        responseMetadata = chunk.value.metadata;
+      }
       const delta = (_b = chunk.value) == null ? void 0 : _b.output;
       if (!delta) {
         continue;
@@ -271,22 +274,25 @@ var generateCode = async ({
       raw += delta;
       await dispatchSends(parser.push(fence.push(delta)));
     }
+    if (!responseMetadata) {
+      throw new (0, _chunkPK72FAKDcjs.CognitiveError)("LLM streaming completed without metadata");
+    }
     const remaining = fence.flush();
     if (remaining) {
       await dispatchSends(parser.push(remaining));
     }
     await dispatchSends(parser.finish());
-    assistantResponse = _chunkC2KND2PScjs.toParsedAssistantResponse.call(void 0, parser.items, raw);
+    assistantResponse = _chunkOVRDJMZJcjs.toParsedAssistantResponse.call(void 0, parser.items, raw);
   } else {
-    output = await cognitive.generateContent(input).catch((thrown) => {
-      throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`LLM generation failed: ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, thrown)}`);
+    const response = await cognitive.generateText(input, { signal: controller.signal }).catch((thrown) => {
+      throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`LLM generation failed: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, thrown)}`);
     });
-    const out = typeof ((_d = (_c = output.output.choices) == null ? void 0 : _c[0]) == null ? void 0 : _d.content) === "string" ? output.output.choices[0].content : null;
-    if (!out) {
+    if (!response.output) {
       throw new (0, _chunkPK72FAKDcjs.CognitiveError)("LLM did not return any text output");
     }
-    raw = out;
-    assistantResponse = ctx.version.parseAssistantResponse(out);
+    responseMetadata = response.metadata;
+    raw = response.output;
+    assistantResponse = ctx.version.parseAssistantResponse(raw);
     for (const send of assistantResponse.sends) {
       await (onSend == null ? void 0 : onSend(send));
     }
@@ -294,29 +300,35 @@ var generateCode = async ({
   iteration.code = assistantResponse.code;
   iteration.sends = assistantResponse.sends;
   iteration.next = assistantResponse.next;
+  const usage = responseMetadata.usage;
   iteration.llm = {
-    cached: output.meta.cached || false,
+    cached: responseMetadata.cached || false,
     ended_at: Date.now(),
     started_at: startedAt,
     status: "success",
-    tokens: output.meta.tokens.input + output.meta.tokens.output,
-    spend: output.meta.cost.input + output.meta.cost.output,
+    tokens: usage.inputTokens + usage.outputTokens,
+    spend: _nullishCoalesce(responseMetadata.cost, () => ( usage.inputCost + usage.outputCost)),
     output: assistantResponse.raw,
-    model: `${output.meta.model.integration}:${output.meta.model.model}`,
+    model: `${responseMetadata.provider}:${responseMetadata.model}`,
     time_to_first_token: timeToFirstToken,
     time_to_last_token: timeToLastToken,
-    usage: output.output.usage
+    usage: {
+      inputTokens: usage.inputTokens,
+      inputCost: usage.inputCost,
+      outputTokens: usage.outputTokens,
+      outputCost: usage.outputCost
+    }
   };
   if (iteration.tokens) {
-    iteration.tokens.input = output.meta.tokens.input;
-    iteration.tokens.output = output.meta.tokens.output;
-    iteration.tokens.total = output.meta.tokens.input + output.meta.tokens.output;
+    iteration.tokens.input = usage.inputTokens;
+    iteration.tokens.output = usage.outputTokens;
+    iteration.tokens.total = usage.inputTokens + usage.outputTokens;
   }
   traces.push({
     type: "llm_call_success",
     started_at: startedAt,
     ended_at: iteration.llm.ended_at,
-    model: model.ref,
+    model: model.id,
     code: _nullishCoalesce(iteration.code, () => ( ""))
   });
 };
@@ -395,7 +407,7 @@ var interpretVMResult = async ({
     return;
   }
   const returnValue = result.return_value;
-  const codeRequestsInspection = _chunk2EZSVF5Xcjs.hasTopLevelReturn.call(void 0, _nullishCoalesce(iteration.code, () => ( "")));
+  const codeRequestsInspection = _chunk3UZKLXADcjs.hasTopLevelReturn.call(void 0, _nullishCoalesce(iteration.code, () => ( "")));
   if (iteration.next && !codeRequestsInspection) {
     await applyNextExit({ iteration, controller, onExit });
     return;
@@ -420,7 +432,7 @@ var applyNextExit = async ({
   if (!next) {
     throw new Error("applyNextExit called without a \u25A0next block on the iteration");
   }
-  const attempt = (props) => _chunk5XPEVOOWcjs.parseExit.call(void 0, Object.keys(props).length ? { action: next.name, value: props } : { action: next.name }, iteration.exits);
+  const attempt = (props) => _chunkKG5E3UZZcjs.parseExit.call(void 0, Object.keys(props).length ? { action: next.name, value: props } : { action: next.name }, iteration.exits);
   let parsedExit = attempt(next.props);
   const keys = Object.keys(next.props);
   if (!parsedExit.success && keys.length === 1 && ["props", "value", "data"].includes(keys[0])) {
@@ -458,7 +470,7 @@ var applyNextExit = async ({
       type: "exit_error",
       exit_error: {
         exit: returnExit.name,
-        message: `Error executing exit ${returnExit.name}: ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, err)}`,
+        message: `Error executing exit ${returnExit.name}: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, err)}`,
         return_value: returnValue
       }
     });
@@ -639,7 +651,7 @@ var buildVMContext = ({
   onAfterTool
 }) => {
   const traces = iteration.traces;
-  const vmContext = { ..._chunkFPLSJF4Scjs.stripInvalidIdentifiers.call(void 0, iteration.variables) };
+  const vmContext = { ..._chunkGOWPACSKcjs.stripInvalidIdentifiers.call(void 0, iteration.variables) };
   for (const obj of iteration.objects) {
     const internalValues = {};
     const instance = {};
@@ -663,7 +675,7 @@ var buildVMContext = ({
           const parsed = schema.safeParse(value2);
           if (!parsed.success) {
             throw new (0, _chunkPK72FAKDcjs.AssignmentError)(
-              `Invalid value for Object property ${obj.name}.${name}: ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, parsed.error)}`
+              `Invalid value for Object property ${obj.name}.${name}: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, parsed.error)}`
             );
           }
           internalValues[name] = parsed.data;
@@ -714,7 +726,7 @@ var buildVMContext = ({
 // src/runtime/execute.ts
 var executeContext = async (props) => {
   var _a, _b;
-  await _chunkFPLSJF4Scjs.init.call(void 0, );
+  await _chunkGOWPACSKcjs.init.call(void 0, );
   const result = await executeContextInternal(props);
   try {
     (_b = (_a = result.context.chat) == null ? void 0 : _a.onExecutionDone) == null ? void 0 : _b.call(_a, result);
@@ -727,8 +739,8 @@ var executeContextInternal = async (props) => {
   const controller = createJoinedAbortController([props.signal]);
   const { onIterationStart, onIterationEnd, onTrace, onExit, onBeforeExecution, onAfterTool, onBeforeTool } = props;
   const client = _nullishCoalesce(props.client, () => ( new (0, _client.Client)()));
-  const cognitive = _cognitive.Cognitive.isCognitiveClient(client) || _chunk5XPEVOOWcjs._CustomModelClient.isCustomClient(client) ? client : _cognitive.CognitiveBeta.isBetaClient(client) ? _cognitive.cognitiveFromBeta.call(void 0, client) : new (0, _cognitive.Cognitive)({ client, __experimental_beta: true });
-  const ctx = new (0, _chunk5XPEVOOWcjs.Context)({
+  const cognitive = _cognitive.Cognitive.isCognitiveClient(client) || _chunkKG5E3UZZcjs._CustomModelClient.isCustomClient(client) ? client : new (0, _cognitive.Cognitive)({ client });
+  const ctx = new (0, _chunkKG5E3UZZcjs.Context)({
     chat: props.chat,
     instructions: props.instructions,
     objects: props.objects,
@@ -746,7 +758,7 @@ var executeContextInternal = async (props) => {
   try {
     while (true) {
       if (ctx.iterations.length >= ctx.loop) {
-        return new (0, _chunk5XPEVOOWcjs.ErrorExecutionResult)(ctx, new (0, _chunkPK72FAKDcjs.LoopExceededError)());
+        return new (0, _chunkKG5E3UZZcjs.ErrorExecutionResult)(ctx, new (0, _chunkPK72FAKDcjs.LoopExceededError)());
       }
       const iteration = await ctx.nextIteration();
       const unsubscribeTrace = iteration.traces.onPush((traces) => {
@@ -777,7 +789,7 @@ var executeContextInternal = async (props) => {
               }
             }
           });
-          return new (0, _chunk5XPEVOOWcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(controller.signal.reason, () => ( "The operation was aborted")));
+          return new (0, _chunkKG5E3UZZcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(controller.signal.reason, () => ( "The operation was aborted")));
         }
         const iterationResult = await executeIterationWithErrorHandling({
           iteration,
@@ -807,7 +819,7 @@ var executeContextInternal = async (props) => {
       }
     }
   } catch (error) {
-    return new (0, _chunk5XPEVOOWcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(error, () => ( "Unknown error")));
+    return new (0, _chunkKG5E3UZZcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(error, () => ( "Unknown error")));
   }
 };
 var executeIterationStartHook = async ({
@@ -837,7 +849,7 @@ var executeIterationStartHook = async ({
       } : {
         type: "execution_error",
         execution_error: {
-          message: `Error in onIterationStart hook: ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, err)}`,
+          message: `Error in onIterationStart hook: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, err)}`,
           stack: _chunkKIN7Y247cjs.cleanStackTrace.call(void 0, 
             err instanceof Error ? _nullishCoalesce(err.stack, () => ( "No stack trace available")) : "No stack trace available"
           )
@@ -882,7 +894,7 @@ var executeIterationWithErrorHandling = async ({
         status: iteration.status.type === "pending" ? {
           type: "execution_error",
           execution_error: {
-            message: "An unexpected error occurred: " + _chunkFPLSJF4Scjs.getErrorMessage.call(void 0, err),
+            message: "An unexpected error occurred: " + _chunkGOWPACSKcjs.getErrorMessage.call(void 0, err),
             stack: _chunkKIN7Y247cjs.cleanStackTrace.call(void 0, 
               err instanceof Error ? _nullishCoalesce(err.stack, () => ( "No stack trace available")) : "No stack trace available"
             )
@@ -894,7 +906,7 @@ var executeIterationWithErrorHandling = async ({
     if (!controller.signal.aborted) {
       return {
         type: "return",
-        result: new (0, _chunk5XPEVOOWcjs.ErrorExecutionResult)(ctx, err)
+        result: new (0, _chunkKG5E3UZZcjs.ErrorExecutionResult)(ctx, err)
       };
     }
     await finalizeIteration({
@@ -910,7 +922,7 @@ var executeIterationWithErrorHandling = async ({
     });
     return {
       type: "return",
-      result: new (0, _chunk5XPEVOOWcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(controller.signal.reason, () => ( "The operation was aborted")))
+      result: new (0, _chunkKG5E3UZZcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(controller.signal.reason, () => ( "The operation was aborted")))
     };
   }
 };
@@ -922,7 +934,7 @@ var transformIterationResult = ({
     const exitName = iteration.status.exit_success.exit_name;
     return {
       type: "return",
-      result: new (0, _chunk5XPEVOOWcjs.SuccessExecutionResult)(ctx, {
+      result: new (0, _chunkKG5E3UZZcjs.SuccessExecutionResult)(ctx, {
         exit: iteration.exits.find((x) => x.name === exitName),
         result: iteration.status.exit_success.return_value
       })
@@ -931,17 +943,17 @@ var transformIterationResult = ({
   if (iteration.status.type === "callback_requested") {
     return {
       type: "return",
-      result: new (0, _chunk5XPEVOOWcjs.PartialExecutionResult)(
+      result: new (0, _chunkKG5E3UZZcjs.PartialExecutionResult)(
         ctx,
         iteration.status.callback_requested.signal,
-        _chunk5XPEVOOWcjs.Snapshot.fromSignal(iteration.status.callback_requested.signal)
+        _chunkKG5E3UZZcjs.Snapshot.fromSignal(iteration.status.callback_requested.signal)
       )
     };
   }
   if (iteration.status.type === "aborted") {
     return {
       type: "return",
-      result: new (0, _chunk5XPEVOOWcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(iteration.error, () => ( "The operation was aborted")))
+      result: new (0, _chunkKG5E3UZZcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(iteration.error, () => ( "The operation was aborted")))
     };
   }
   if (iteration.status.type === "thinking_requested" || iteration.status.type === "exit_error" || iteration.status.type === "execution_error" || iteration.status.type === "invalid_code_error") {
@@ -949,7 +961,7 @@ var transformIterationResult = ({
   }
   return {
     type: "return",
-    result: new (0, _chunk5XPEVOOWcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(iteration.error, () => ( `Unknown error. Status: ${iteration.status.type}`)))
+    result: new (0, _chunkKG5E3UZZcjs.ErrorExecutionResult)(ctx, _nullishCoalesce(iteration.error, () => ( `Unknown error. Status: ${iteration.status.type}`)))
   };
 };
 var executeIteration = async ({
@@ -966,10 +978,10 @@ var executeIteration = async ({
   var _a, _b;
   const runCode = (code) => {
     const vmContext = buildVMContext({ ctx, iteration, controller, onBeforeTool, onAfterTool });
-    return _chunk2EZSVF5Xcjs.runAsyncFunction.call(void 0, vmContext, code, iteration.traces, controller.signal, ctx.timeout).catch((err) => {
+    return _chunk3UZKLXADcjs.runAsyncFunction.call(void 0, vmContext, code, iteration.traces, controller.signal, ctx.timeout).catch((err) => {
       return {
         success: false,
-        error: err instanceof Error ? err : new Error(_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, err)),
+        error: err instanceof Error ? err : new Error(_chunkGOWPACSKcjs.getErrorMessage.call(void 0, err)),
         lines_executed: [],
         traces: [],
         variables: {}
@@ -985,7 +997,7 @@ var executeIteration = async ({
     controller,
     metadata,
     // Pre-warm the VM while the model is still writing the ■run block
-    onRunStart: () => _chunk2EZSVF5Xcjs.warmupVM.call(void 0, ),
+    onRunStart: () => _chunk3UZKLXADcjs.warmupVM.call(void 0, ),
     onRunComplete: canExecuteEarly ? (code) => {
       if (!code.length || controller.signal.aborted || earlyExecution) {
         return;
@@ -1006,7 +1018,7 @@ var executeIteration = async ({
       try {
         await ctx.chat.handler(component);
       } catch (err) {
-        throw new Error(`Error while sending message (\u25A0send=${send.name}): ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, err)}`);
+        throw new Error(`Error while sending message (\u25A0send=${send.name}): ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, err)}`);
       }
       iteration.traces.push({ type: "yield", value: component, started_at: sendStartedAt, ended_at: Date.now() });
     },
@@ -1035,7 +1047,7 @@ var executeIteration = async ({
       iteration.end({
         type: "execution_error",
         execution_error: {
-          message: `Error in onBeforeExecution hook: ${_chunkFPLSJF4Scjs.getErrorMessage.call(void 0, err)}`,
+          message: `Error in onBeforeExecution hook: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, err)}`,
           stack: _chunkKIN7Y247cjs.cleanStackTrace.call(void 0, 
             err instanceof Error ? _nullishCoalesce(err.stack, () => ( "No stack trace available")) : "No stack trace available"
           )

package/dist/{execute-6AYBMTRM.js → execute-DPCZ6PKH.js}

@@ -2,7 +2,7 @@ import {
   hasTopLevelReturn,
   runAsyncFunction,
   warmupVM
-} from "./chunk-EJIDFDQP.js";
+} from "./chunk-YRS7KRKP.js";
 import {
   Context,
   ErrorExecutionResult,
@@ -11,13 +11,14 @@ import {
   SuccessExecutionResult,
   _CustomModelClient,
   parseExit
-} from "./chunk-2MSLOTFO.js";
+} from "./chunk-XMFQVARN.js";
 import {
   StreamingMessageParser,
   toParsedAssistantResponse
-} from "./chunk-5ISVDYBA.js";
-import "./chunk-3OSQ4XL4.js";
-import "./chunk-GDWRFHM5.js";
+} from "./chunk-OXEKAOVI.js";
+import "./chunk-UGHFNCPD.js";
+import "./chunk-TDALRNGL.js";
+import "./chunk-GDJTKI7J.js";
 import {
   AssignmentError,
   CodeExecutionError,
@@ -31,19 +32,19 @@ import {
 import {
   cleanStackTrace
 } from "./chunk-YEAWWJSJ.js";
-import "./chunk-XVVYNOTX.js";
+import "./chunk-HB2G32AK.js";
 import "./chunk-KWQTYXAT.js";
 import {
   createJsxComponent
 } from "./chunk-XTEHT2TK.js";
 import {
   truncateWrappedContent
-} from "./chunk-HM5H556L.js";
+} from "./chunk-72ZWBNZV.js";
 import {
   getErrorMessage,
   init,
   stripInvalidIdentifiers
-} from "./chunk-YX27XM67.js";
+} from "./chunk-GLXADNTR.js";
 import {
   clamp_default,
   isEqual_default
@@ -51,7 +52,7 @@ import {
 
 // src/runtime/execute.ts
 import { Client } from "@botpress/client";
-import { Cognitive, CognitiveBeta, cognitiveFromBeta } from "@botpress/cognitive";
+import { Cognitive } from "@botpress/cognitive";
 
 // src/abort-signal.ts
 function createJoinedAbortController(signals) {
@@ -134,7 +135,7 @@ var generateCode = async ({
   onRunStart,
   onRunComplete
 }) => {
-  var _a, _b, _c, _d;
+  var _a, _b;
   const startedAt = Date.now();
   const traces = iteration.traces;
   const modelRef = Array.isArray(iteration.model) ? iteration.model[0] : iteration.model;
@@ -165,20 +166,18 @@ var generateCode = async ({
     type: "llm_call_started",
     started_at: startedAt,
     ended_at: startedAt,
-    model: model.ref
+    model: model.id
   });
   const input = {
-    signal: controller.signal,
-    systemPrompt: (_a = messages.find((x) => x.role === "system")) == null ? void 0 : _a.content,
     model: iteration.model,
     temperature: iteration.temperature,
     responseFormat: "text",
     reasoningEffort: iteration.reasoningEffort,
-    messages: messages.filter((x) => x.role !== "system"),
+    messages,
     stopSequences: ctx.version.getStopTokens(),
     meta: metadata ? { metadata } : void 0
   };
-  let output;
+  let responseMetadata;
   let raw;
   let assistantResponse;
   let timeToFirstToken;
@@ -223,18 +222,20 @@ var generateCode = async ({
       }
     }
   };
-  if (typeof cognitive.generateContentStream === "function") {
+  if (typeof cognitive.generateTextStream === "function") {
     const parser = new StreamingMessageParser();
     const fence = new LeadingFenceFilter();
     const streamController = createJoinedAbortController([controller.signal]);
     const requestedAt = Date.now();
-    const stream = cognitive.generateContentStream({
-      ...input,
-      signal: streamController.signal,
-      // Passed through to the cognitive v2 request: fall back to the next
-      // model/provider when the first token takes too long
-      ...ctx.maxTimeToFirstToken ? { options: { maxTimeToFirstToken: ctx.maxTimeToFirstToken } } : {}
-    });
+    const stream = cognitive.generateTextStream(
+      {
+        ...input,
+        // Passed through to the cognitive request: fall back to the next
+        // model/provider when the first token takes too long
+        ...ctx.maxTimeToFirstToken ? { options: { maxTimeToFirstToken: ctx.maxTimeToFirstToken } } : {}
+      },
+      { signal: streamController.signal }
+    );
     const inactivityTimeout = Math.max(STREAM_INACTIVITY_TIMEOUT, ctx.maxTimeToFirstToken ?? 0);
     const nextChunk = async () => {
       let timer;
@@ -259,9 +260,11 @@ var generateCode = async ({
         throw new CognitiveError(`LLM generation failed: ${getErrorMessage(thrown)}`);
       }
       if (chunk.done) {
-        output = chunk.value;
         break;
       }
+      if ((_a = chunk.value) == null ? void 0 : _a.metadata) {
+        responseMetadata = chunk.value.metadata;
+      }
       const delta = (_b = chunk.value) == null ? void 0 : _b.output;
       if (!delta) {
         continue;
@@ -271,6 +274,9 @@ var generateCode = async ({
       raw += delta;
       await dispatchSends(parser.push(fence.push(delta)));
     }
+    if (!responseMetadata) {
+      throw new CognitiveError("LLM streaming completed without metadata");
+    }
     const remaining = fence.flush();
     if (remaining) {
       await dispatchSends(parser.push(remaining));
@@ -278,15 +284,15 @@ var generateCode = async ({
     await dispatchSends(parser.finish());
     assistantResponse = toParsedAssistantResponse(parser.items, raw);
   } else {
-    output = await cognitive.generateContent(input).catch((thrown) => {
+    const response = await cognitive.generateText(input, { signal: controller.signal }).catch((thrown) => {
       throw new CognitiveError(`LLM generation failed: ${getErrorMessage(thrown)}`);
     });
-    const out = typeof ((_d = (_c = output.output.choices) == null ? void 0 : _c[0]) == null ? void 0 : _d.content) === "string" ? output.output.choices[0].content : null;
-    if (!out) {
+    if (!response.output) {
       throw new CognitiveError("LLM did not return any text output");
     }
-    raw = out;
-    assistantResponse = ctx.version.parseAssistantResponse(out);
+    responseMetadata = response.metadata;
+    raw = response.output;
+    assistantResponse = ctx.version.parseAssistantResponse(raw);
     for (const send of assistantResponse.sends) {
       await (onSend == null ? void 0 : onSend(send));
     }
@@ -294,29 +300,35 @@ var generateCode = async ({
   iteration.code = assistantResponse.code;
   iteration.sends = assistantResponse.sends;
   iteration.next = assistantResponse.next;
+  const usage = responseMetadata.usage;
   iteration.llm = {
-    cached: output.meta.cached || false,
+    cached: responseMetadata.cached || false,
     ended_at: Date.now(),
     started_at: startedAt,
     status: "success",
-    tokens: output.meta.tokens.input + output.meta.tokens.output,
-    spend: output.meta.cost.input + output.meta.cost.output,
+    tokens: usage.inputTokens + usage.outputTokens,
+    spend: responseMetadata.cost ?? usage.inputCost + usage.outputCost,
     output: assistantResponse.raw,
-    model: `${output.meta.model.integration}:${output.meta.model.model}`,
+    model: `${responseMetadata.provider}:${responseMetadata.model}`,
     time_to_first_token: timeToFirstToken,
     time_to_last_token: timeToLastToken,
-    usage: output.output.usage
+    usage: {
+      inputTokens: usage.inputTokens,
+      inputCost: usage.inputCost,
+      outputTokens: usage.outputTokens,
+      outputCost: usage.outputCost
+    }
   };
   if (iteration.tokens) {
-    iteration.tokens.input = output.meta.tokens.input;
-    iteration.tokens.output = output.meta.tokens.output;
-    iteration.tokens.total = output.meta.tokens.input + output.meta.tokens.output;
+    iteration.tokens.input = usage.inputTokens;
+    iteration.tokens.output = usage.outputTokens;
+    iteration.tokens.total = usage.inputTokens + usage.outputTokens;
   }
   traces.push({
     type: "llm_call_success",
     started_at: startedAt,
     ended_at: iteration.llm.ended_at,
-    model: model.ref,
+    model: model.id,
     code: iteration.code ?? ""
   });
 };
@@ -727,7 +739,7 @@ var executeContextInternal = async (props) => {
   const controller = createJoinedAbortController([props.signal]);
   const { onIterationStart, onIterationEnd, onTrace, onExit, onBeforeExecution, onAfterTool, onBeforeTool } = props;
   const client = props.client ?? new Client();
-  const cognitive = Cognitive.isCognitiveClient(client) || _CustomModelClient.isCustomClient(client) ? client : CognitiveBeta.isBetaClient(client) ? cognitiveFromBeta(client) : new Cognitive({ client, __experimental_beta: true });
+  const cognitive = Cognitive.isCognitiveClient(client) || _CustomModelClient.isCustomClient(client) ? client : new Cognitive({ client });
   const ctx = new Context({
     chat: props.chat,
     instructions: props.instructions,

package/dist/exit-MX56VN2C.cjs

@@ -0,0 +1,8 @@
+"use strict";Object.defineProperty(exports, "__esModule", {value: true});
+
+var _chunkDDSJTR6Qcjs = require('./chunk-DDSJTR6Q.cjs');
+require('./chunk-GOWPACSK.cjs');
+require('./chunk-HC2QZTVT.cjs');
+
+
+exports.Exit = _chunkDDSJTR6Qcjs.Exit;

package/dist/{exit-VJATJTQI.js → exit-UQ3EUX5D.js}

@@ -1,7 +1,7 @@
 import {
   Exit
-} from "./chunk-XVVYNOTX.js";
-import "./chunk-YX27XM67.js";
+} from "./chunk-HB2G32AK.js";
+import "./chunk-GLXADNTR.js";
 import "./chunk-34GJUS6U.js";
 export {
   Exit