@gendive/chatllm 0.17.19 → 0.17.20

package/dist/react/index.js

@@ -483,7 +483,7 @@ var parseExtractionResult = (text) => {
 function useInfoExtraction(options) {
   const [isExtracting, setIsExtracting] = (0, import_react2.useState)(false);
   const [lastExtraction, setLastExtraction] = (0, import_react2.useState)(null);
-  const { apiEndpoint, model, minConfidence = 0.8, globalMemory } = options;
+  const { apiEndpoint, model, minConfidence = 0.8, globalMemory, onCallLLM } = options;
   const extractInfo = (0, import_react2.useCallback)(
     async (messages) => {
       if (messages.length === 0) {
@@ -492,41 +492,44 @@ function useInfoExtraction(options) {
       setIsExtracting(true);
       try {
         const prompt = buildExtractionPrompt(messages);
-        const response = await fetch(apiEndpoint, {
-          method: "POST",
-          headers: { "Content-Type": "application/json" },
-          body: JSON.stringify({
-            messages: [{ role: "user", content: prompt }],
-            model: model || "default"
-          })
-        });
-        if (!response.ok) {
-          throw new Error(`API error: ${response.status}`);
-        }
-        const reader = response.body?.getReader();
-        if (!reader) {
-          return [];
-        }
-        const decoder = new TextDecoder();
-        let buffer = "";
-        let fullResponse = "";
-        while (true) {
-          const { done, value } = await reader.read();
-          if (done) break;
-          buffer += decoder.decode(value, { stream: true });
-          const lines = buffer.split("\n");
-          buffer = lines.pop() || "";
-          for (const line of lines) {
-            if (line.startsWith("data: ")) {
-              const data = line.slice(6);
-              if (data === "[DONE]") continue;
-              try {
-                const parsed = JSON.parse(data);
-                {
+        let fullResponse;
+        if (onCallLLM) {
+          fullResponse = await onCallLLM(prompt, model || "default");
+        } else {
+          const response = await fetch(apiEndpoint, {
+            method: "POST",
+            headers: { "Content-Type": "application/json" },
+            body: JSON.stringify({
+              messages: [{ role: "user", content: prompt }],
+              model: model || "default"
+            })
+          });
+          if (!response.ok) {
+            throw new Error(`API error: ${response.status}`);
+          }
+          const reader = response.body?.getReader();
+          if (!reader) {
+            return [];
+          }
+          const decoder = new TextDecoder();
+          let buffer = "";
+          fullResponse = "";
+          while (true) {
+            const { done, value } = await reader.read();
+            if (done) break;
+            buffer += decoder.decode(value, { stream: true });
+            const lines = buffer.split("\n");
+            buffer = lines.pop() || "";
+            for (const line of lines) {
+              if (line.startsWith("data: ")) {
+                const data = line.slice(6);
+                if (data === "[DONE]") continue;
+                try {
+                  const parsed = JSON.parse(data);
                   const chunk = parsed.content ?? parsed.text ?? "";
                   if (chunk) fullResponse += chunk;
+                } catch {
                 }
-              } catch {
               }
             }
           }
@@ -553,7 +556,7 @@ function useInfoExtraction(options) {
         setIsExtracting(false);
       }
     },
-    [apiEndpoint, model, minConfidence, globalMemory]
+    [apiEndpoint, model, minConfidence, globalMemory, onCallLLM]
   );
   return {
     extractInfo,
@@ -1609,6 +1612,33 @@ var removeSessionCache = (storageKey, sessionId) => {
 };
 
 // src/react/hooks/useChatUI.ts
+var parseSSEResponse = async (response) => {
+  const reader = response.body?.getReader();
+  if (!reader) return "";
+  const decoder = new TextDecoder();
+  let buffer = "";
+  let result = "";
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    buffer += decoder.decode(value, { stream: true });
+    const lines = buffer.split("\n");
+    buffer = lines.pop() || "";
+    for (const line of lines) {
+      if (line.startsWith("data: ")) {
+        const data = line.slice(6);
+        if (data === "[DONE]") continue;
+        try {
+          const parsed = JSON.parse(data);
+          const chunk = parsed.content ?? parsed.text ?? "";
+          if (chunk) result += chunk;
+        } catch {
+        }
+      }
+    }
+  }
+  return result;
+};
 var DEFAULT_STORAGE_KEY = "chatllm_sessions";
 var DEFAULT_COMPRESSION_THRESHOLD = 20;
 var DEFAULT_KEEP_RECENT = 6;
@@ -1814,11 +1844,37 @@ var useChatUI = (options) => {
   );
   const projectMemoryRaw = useGlobalMemory(projectMemoryOptions);
   const projectMemory = enableProjects ? projectMemoryRaw : null;
+  const callInternalLLM = (0, import_react5.useCallback)(async (prompt, model) => {
+    if (onSendMessageRef.current) {
+      const modelConfig = models.find((m) => m.id === model);
+      const provider = modelConfig?.provider || "ollama";
+      const result = await onSendMessageRef.current({
+        messages: [{ role: "user", content: prompt }],
+        model,
+        provider,
+        apiKey
+      });
+      if (typeof result === "string") return result;
+      if (typeof result === "object" && "content" in result) return result.content;
+      return parseSSEResponse(new Response(result));
+    }
+    const response = await fetch(apiEndpoint, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({
+        messages: [{ role: "user", content: prompt }],
+        model
+      })
+    });
+    if (!response.ok) return "";
+    return parseSSEResponse(response);
+  }, [apiEndpoint, apiKey, models]);
   const infoExtraction = useInfoExtraction({
     apiEndpoint,
     model: selectedModel,
     minConfidence: 0.8,
-    globalMemory
+    globalMemory,
+    onCallLLM: callInternalLLM
   });
   const currentSession = sessions.find((s) => s.id === currentSessionId) || null;
   const messages = currentSession?.messages.filter((m) => !m.hidden) || [];
@@ -2016,46 +2072,11 @@ ${conversationText}
 
 \uC694\uC57D:`;
     try {
-      const response = await fetch(apiEndpoint, {
-        method: "POST",
-        headers: { "Content-Type": "application/json" },
-        body: JSON.stringify({
-          messages: [{ role: "user", content: summaryPrompt }],
-          model
-        })
-      });
-      if (!response.ok) return "";
-      const reader = response.body?.getReader();
-      if (!reader) return "";
-      const decoder = new TextDecoder();
-      let buffer = "";
-      let summary = "";
-      while (true) {
-        const { done, value } = await reader.read();
-        if (done) break;
-        buffer += decoder.decode(value, { stream: true });
-        const lines = buffer.split("\n");
-        buffer = lines.pop() || "";
-        for (const line of lines) {
-          if (line.startsWith("data: ")) {
-            const data = line.slice(6);
-            if (data === "[DONE]") continue;
-            try {
-              const parsed = JSON.parse(data);
-              {
-                const chunk = parsed.content ?? parsed.text ?? "";
-                if (chunk) summary += chunk;
-              }
-            } catch {
-            }
-          }
-        }
-      }
-      return summary;
+      return await callInternalLLM(summaryPrompt, model);
     } catch {
       return "";
     }
-  }, [apiEndpoint]);
+  }, [callInternalLLM]);
   const incrementalCompressContext = (0, import_react5.useCallback)(
     async (existingSummary, newMessages, model) => {
       const newConversation = newMessages.map((m) => `${m.role === "user" ? "\uC0AC\uC6A9\uC790" : "AI"}: ${m.content}`).join("\n\n");
@@ -2075,47 +2096,13 @@ ${newConversation}
 
 \uD1B5\uD569 \uC694\uC57D:`;
       try {
-        const response = await fetch(apiEndpoint, {
-          method: "POST",
-          headers: { "Content-Type": "application/json" },
-          body: JSON.stringify({
-            messages: [{ role: "user", content: mergePrompt }],
-            model
-          })
-        });
-        if (!response.ok) return existingSummary;
-        const reader = response.body?.getReader();
-        if (!reader) return existingSummary;
-        const decoder = new TextDecoder();
-        let buffer = "";
-        let summary = "";
-        while (true) {
-          const { done, value } = await reader.read();
-          if (done) break;
-          buffer += decoder.decode(value, { stream: true });
-          const lines = buffer.split("\n");
-          buffer = lines.pop() || "";
-          for (const line of lines) {
-            if (line.startsWith("data: ")) {
-              const data = line.slice(6);
-              if (data === "[DONE]") continue;
-              try {
-                const parsed = JSON.parse(data);
-                {
-                  const chunk = parsed.content ?? parsed.text ?? "";
-                  if (chunk) summary += chunk;
-                }
-              } catch {
-              }
-            }
-          }
-        }
+        const summary = await callInternalLLM(mergePrompt, model);
         return summary || existingSummary;
       } catch {
         return existingSummary;
       }
     },
-    [apiEndpoint]
+    [callInternalLLM]
   );
   const estimateTokens = (0, import_react5.useCallback)((messages2) => {
     return messages2.reduce((sum, m) => sum + Math.ceil(m.content.length / 4), 0);