llmz 0.2.3 → 0.3.1

package/dist/context.d.ts

@@ -456,7 +456,7 @@ export declare class Iteration implements Serializable<Iteration.JSON> {
     end(status: IterationStatus): void;
     toJSON(): {
         id: string;
-        messages: LLMzPrompts.Message[];
+        messages: import("@botpress/cognitive").CognitiveMessage[];
         code: string | undefined;
         model: Models | Models[];
         temperature: number;
@@ -600,7 +600,7 @@ export declare class Context implements Serializable<Context.JSON> {
         id: string;
         iterations: {
             id: string;
-            messages: LLMzPrompts.Message[];
+            messages: import("@botpress/cognitive").CognitiveMessage[];
             code: string | undefined;
             model: Models | Models[];
             temperature: number;

package/dist/custom-client.d.ts

@@ -1,7 +1,8 @@
 import { Model } from '@botpress/cognitive';
 import { RuntimeCognitive } from './runtime/types.js';
-export type RuntimeGenerateContentInput = Parameters<RuntimeCognitive['generateContent']>[0];
-export type RuntimeGenerateContentOutput = Awaited<ReturnType<RuntimeCognitive['generateContent']>>;
+export type RuntimeGenerateContentInput = Parameters<RuntimeCognitive['generateText']>[0];
+export type RuntimeGenerateContentOptions = Parameters<RuntimeCognitive['generateText']>[1];
+export type RuntimeGenerateContentOutput = Awaited<ReturnType<RuntimeCognitive['generateText']>>;
 /**
  * Internal escape hatch for supplying a custom model client that bypasses Cognitive.
  *
@@ -14,5 +15,5 @@ export declare abstract class _CustomModelClient implements RuntimeCognitive {
     private readonly ['$$COGNITIVE'];
     static isCustomClient(obj: any): obj is _CustomModelClient;
     abstract getModelDetails(model: string): Promise<Model>;
-    abstract generateContent(input: RuntimeGenerateContentInput): Promise<RuntimeGenerateContentOutput>;
+    abstract generateText(input: RuntimeGenerateContentInput, options?: RuntimeGenerateContentOptions): Promise<RuntimeGenerateContentOutput>;
 }

package/dist/{execute-MRZZ4AYY.cjs → execute-2H4F7GFE.cjs}

@@ -135,7 +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;
@@ -166,20 +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;
@@ -224,18 +222,20 @@ var generateCode = async ({
       }
     }
   };
-  if (typeof cognitive.generateContentStream === "function") {
+  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;
@@ -260,9 +260,11 @@ var generateCode = async ({
         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;
@@ -272,6 +274,9 @@ 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));
@@ -279,15 +284,15 @@ var generateCode = async ({
     await dispatchSends(parser.finish());
     assistantResponse = _chunkOVRDJMZJcjs.toParsedAssistantResponse.call(void 0, parser.items, raw);
   } else {
-    output = await cognitive.generateContent(input).catch((thrown) => {
+    const response = await cognitive.generateText(input, { signal: controller.signal }).catch((thrown) => {
       throw new (0, _chunkPK72FAKDcjs.CognitiveError)(`LLM generation failed: ${_chunkGOWPACSKcjs.getErrorMessage.call(void 0, thrown)}`);
     });
-    const out = typeof ((_d = (_c = output.output.choices) == null ? void 0 : _c[0]) == null ? void 0 : _d.content) === "string" ? output.output.choices[0].content : null;
-    if (!out) {
+    if (!response.output) {
       throw new (0, _chunkPK72FAKDcjs.CognitiveError)("LLM did not return any text output");
     }
-    raw = out;
-    assistantResponse = ctx.version.parseAssistantResponse(out);
+    responseMetadata = response.metadata;
+    raw = response.output;
+    assistantResponse = ctx.version.parseAssistantResponse(raw);
     for (const send of assistantResponse.sends) {
       await (onSend == null ? void 0 : onSend(send));
     }
@@ -295,29 +300,35 @@ var generateCode = async ({
   iteration.code = assistantResponse.code;
   iteration.sends = assistantResponse.sends;
   iteration.next = assistantResponse.next;
+  const usage = responseMetadata.usage;
   iteration.llm = {
-    cached: output.meta.cached || false,
+    cached: responseMetadata.cached || false,
     ended_at: Date.now(),
     started_at: startedAt,
     status: "success",
-    tokens: output.meta.tokens.input + output.meta.tokens.output,
-    spend: output.meta.cost.input + output.meta.cost.output,
+    tokens: usage.inputTokens + usage.outputTokens,
+    spend: _nullishCoalesce(responseMetadata.cost, () => ( usage.inputCost + usage.outputCost)),
     output: assistantResponse.raw,
-    model: `${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, () => ( ""))
   });
 };
@@ -728,7 +739,7 @@ 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) || _chunkKG5E3UZZcjs._CustomModelClient.isCustomClient(client) ? client : _cognitive.CognitiveBeta.isBetaClient(client) ? _cognitive.cognitiveFromBeta.call(void 0, client) : new (0, _cognitive.Cognitive)({ client, __experimental_beta: true });
+  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,

package/dist/{execute-DSGJ2UNL.js → execute-DPCZ6PKH.js}

@@ -52,7 +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) {
@@ -135,7 +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;
@@ -166,20 +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;
@@ -224,18 +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;
@@ -260,9 +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;
@@ -272,6 +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));
@@ -279,15 +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: ${getErrorMessage(thrown)}`);
     });
-    const out = typeof ((_d = (_c = output.output.choices) == null ? void 0 : _c[0]) == null ? void 0 : _d.content) === "string" ? output.output.choices[0].content : null;
-    if (!out) {
+    if (!response.output) {
       throw new CognitiveError("LLM did not return any text output");
     }
-    raw = out;
-    assistantResponse = ctx.version.parseAssistantResponse(out);
+    responseMetadata = response.metadata;
+    raw = response.output;
+    assistantResponse = ctx.version.parseAssistantResponse(raw);
     for (const send of assistantResponse.sends) {
       await (onSend == null ? void 0 : onSend(send));
     }
@@ -295,29 +300,35 @@ var generateCode = async ({
   iteration.code = assistantResponse.code;
   iteration.sends = assistantResponse.sends;
   iteration.next = assistantResponse.next;
+  const usage = responseMetadata.usage;
   iteration.llm = {
-    cached: output.meta.cached || false,
+    cached: responseMetadata.cached || false,
     ended_at: Date.now(),
     started_at: startedAt,
     status: "success",
-    tokens: output.meta.tokens.input + output.meta.tokens.output,
-    spend: output.meta.cost.input + output.meta.cost.output,
+    tokens: usage.inputTokens + usage.outputTokens,
+    spend: responseMetadata.cost ?? usage.inputCost + usage.outputCost,
     output: assistantResponse.raw,
-    model: `${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 ?? ""
   });
 };
@@ -728,7 +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/index.cjs

@@ -69,7 +69,7 @@ var _chunkGOWPACSKcjs = require('./chunk-GOWPACSK.cjs');
 var _chunkHC2QZTVTcjs = require('./chunk-HC2QZTVT.cjs');
 
 // package.json
-var version = "0.2.3";
+var version = "0.3.1";
 
 // src/objects.ts
 var _zui = require('@bpinternal/zui');
@@ -1527,11 +1527,11 @@ var utils = {
   stripTruncationTags: _chunkHEAQAEFZcjs.stripTruncationTags
 };
 var execute = async (props) => {
-  const { executeContext } = await Promise.resolve().then(() => _interopRequireWildcard(require("./execute-MRZZ4AYY.cjs")));
+  const { executeContext } = await Promise.resolve().then(() => _interopRequireWildcard(require("./execute-2H4F7GFE.cjs")));
   return executeContext(props);
 };
 var init = async () => {
-  await Promise.resolve().then(() => _interopRequireWildcard(require("./execute-MRZZ4AYY.cjs")));
+  await Promise.resolve().then(() => _interopRequireWildcard(require("./execute-2H4F7GFE.cjs")));
   await Promise.resolve().then(() => _interopRequireWildcard(require("./component-2NYFY6CW.cjs")));
   await Promise.resolve().then(() => _interopRequireWildcard(require("./tool-VE2YHIY2.cjs")));
   await Promise.resolve().then(() => _interopRequireWildcard(require("./exit-MX56VN2C.cjs")));

package/dist/index.js

@@ -69,7 +69,7 @@ import {
 } from "./chunk-34GJUS6U.js";
 
 // package.json
-var version = "0.2.3";
+var version = "0.3.1";
 
 // src/objects.ts
 import { z } from "@bpinternal/zui";
@@ -1527,11 +1527,11 @@ var utils = {
   stripTruncationTags
 };
 var execute = async (props) => {
-  const { executeContext } = await import("./execute-DSGJ2UNL.js");
+  const { executeContext } = await import("./execute-DPCZ6PKH.js");
   return executeContext(props);
 };
 var init = async () => {
-  await import("./execute-DSGJ2UNL.js");
+  await import("./execute-DPCZ6PKH.js");
   await import("./component-ASNZHQUR.js");
   await import("./tool-KD7FL45V.js");
   await import("./exit-UQ3EUX5D.js");

package/dist/index.workerd.js

@@ -14768,20 +14768,18 @@ ${rest}`;
         type: "llm_call_started",
         started_at: startedAt,
         ended_at: startedAt,
-        model: model.ref
+        model: model.id
       });
       const input = {
-        signal: controller.signal,
-        systemPrompt: messages.find((x) => x.role === "system")?.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;
@@ -14826,18 +14824,20 @@ ${rest}`;
           }
         }
       };
-      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;
@@ -14862,9 +14862,11 @@ ${rest}`;
             throw new CognitiveError(`LLM generation failed: ${getErrorMessage(thrown)}`);
           }
           if (chunk.done) {
-            output = chunk.value;
             break;
           }
+          if (chunk.value?.metadata) {
+            responseMetadata = chunk.value.metadata;
+          }
           const delta = chunk.value?.output;
           if (!delta) {
             continue;
@@ -14874,6 +14876,9 @@ ${rest}`;
           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));
@@ -14881,15 +14886,15 @@ ${rest}`;
         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 output.output.choices?.[0]?.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?.(send);
         }
@@ -14897,29 +14902,35 @@ ${rest}`;
       iteration.code = assistantResponse.code;
       iteration.sends = assistantResponse.sends;
       iteration.next = assistantResponse.next;
+      const usage = responseMetadata.usage;
       iteration.llm = {
-        cached: output.meta.cached || false,
+        cached: responseMetadata.cached || false,
         ended_at: Date.now(),
         started_at: startedAt,
         status: "success",
-        tokens: output.meta.tokens.input + output.meta.tokens.output,
-        spend: output.meta.cost.input + output.meta.cost.output,
+        tokens: usage.inputTokens + usage.outputTokens,
+        spend: responseMetadata.cost ?? usage.inputCost + usage.outputCost,
         output: assistantResponse.raw,
-        model: `${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 ?? ""
       });
     };
@@ -15353,7 +15364,7 @@ __export(execute_exports, {
   executeContext: () => executeContext
 });
 import { Client } from "@botpress/client";
-import { Cognitive, CognitiveBeta, cognitiveFromBeta } from "@botpress/cognitive";
+import { Cognitive } from "@botpress/cognitive";
 var executeContext, executeContextInternal, executeIterationStartHook, executeIterationWithErrorHandling, transformIterationResult, executeIteration;
 var init_execute = __esm({
   "src/runtime/execute.ts"() {
@@ -15385,7 +15396,7 @@ var init_execute = __esm({
       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,
@@ -15748,7 +15759,7 @@ var init_execute = __esm({
 });
 
 // package.json
-var version = "0.2.3";
+var version = "0.3.1";
 
 // src/index.ts
 init_tool();

package/dist/prompts/prompt.d.ts

@@ -1,4 +1,4 @@
-import { type GenerateContentInput } from '@botpress/cognitive';
+import { type CognitiveMessage } from '@botpress/cognitive';
 import { Component } from '../component.js';
 import { Exit } from '../exit.js';
 import type { ParsedItem } from '../message-stream/types.js';
@@ -29,10 +29,7 @@ export type ParsedAssistantResponse = {
     next?: ParsedNext;
 };
 export declare namespace LLMzPrompts {
-    type Message = GenerateContentInput['messages'][number] | {
-        role: 'system';
-        content: string;
-    };
+    type Message = CognitiveMessage;
     /**
      * The raw (pre-truncation, unwrapped) content of each named section of the
      * system prompt. Used to measure the context size of each part of the prompt.
@@ -51,7 +48,7 @@ export declare namespace LLMzPrompts {
         message: Message;
         parts: SystemPromptParts;
     };
-    type MessageContent = Extract<GenerateContentInput['messages'][number]['content'], any[]>[number];
+    type MessageContent = Extract<CognitiveMessage['content'], any[]>[number];
     type InitialStateProps = {
         instructions?: string;
         transcript: TranscriptArray;

package/dist/result.d.ts

@@ -408,7 +408,7 @@ export declare class SuccessExecutionResult<TOutput = unknown> extends Execution
             id: string;
             iterations: {
                 id: string;
-                messages: import("./index.js").LLMzPrompts.Message[];
+                messages: import("@botpress/cognitive").CognitiveMessage[];
                 code: string | undefined;
                 model: import("@botpress/cognitive").Models | import("@botpress/cognitive").Models[];
                 temperature: number;
@@ -573,7 +573,7 @@ export declare class ErrorExecutionResult extends ExecutionResult implements Ser
             id: string;
             iterations: {
                 id: string;
-                messages: import("./index.js").LLMzPrompts.Message[];
+                messages: import("@botpress/cognitive").CognitiveMessage[];
                 code: string | undefined;
                 model: import("@botpress/cognitive").Models | import("@botpress/cognitive").Models[];
                 temperature: number;
@@ -758,7 +758,7 @@ export declare class PartialExecutionResult extends ExecutionResult implements S
             id: string;
             iterations: {
                 id: string;
-                messages: import("./index.js").LLMzPrompts.Message[];
+                messages: import("@botpress/cognitive").CognitiveMessage[];
                 code: string | undefined;
                 model: import("@botpress/cognitive").Models | import("@botpress/cognitive").Models[];
                 temperature: number;