snow-ai 0.3.0 → 0.3.2

package/dist/utils/contextCompressor.js

@@ -1,20 +1,45 @@
-import OpenAI from 'openai';
-import { GoogleGenAI } from '@google/genai';
-import Anthropic from '@anthropic-ai/sdk';
 import { getOpenAiConfig, getCustomHeaders, getCustomSystemPrompt } from './apiConfig.js';
 import { SYSTEM_PROMPT } from '../api/systemPrompt.js';
 /**
  * Compression request prompt - asks AI to summarize conversation with focus on task continuity
  */
 const COMPRESSION_PROMPT = 'Please provide a concise summary of our conversation so far. Focus on: 1) The current task or goal we are working on, 2) Key decisions and approaches we have agreed upon, 3) Important context needed to continue, 4) Any pending or unfinished work. Keep it brief but ensure I can seamlessly continue assisting with the task.';
+/**
+ * Parse Server-Sent Events (SSE) stream
+ */
+async function* parseSSEStream(reader) {
+    const decoder = new TextDecoder();
+    let buffer = '';
+    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) {
+            const trimmed = line.trim();
+            if (!trimmed || trimmed.startsWith(':'))
+                continue;
+            if (trimmed === 'data: [DONE]') {
+                return;
+            }
+            if (trimmed.startsWith('data: ')) {
+                const data = trimmed.slice(6);
+                try {
+                    yield JSON.parse(data);
+                }
+                catch (e) {
+                    console.error('Failed to parse SSE data:', data);
+                }
+            }
+        }
+    }
+}
 /**
  * Compress context using OpenAI Chat Completions API
  */
 async function compressWithChatCompletions(baseUrl, apiKey, modelName, conversationMessages, systemPrompt) {
-    const client = new OpenAI({
-        apiKey,
-        baseURL: baseUrl,
-    });
     const customHeaders = getCustomHeaders();
     // Build messages with system prompt support
     const messages = [];
@@ -48,17 +73,29 @@ async function compressWithChatCompletions(baseUrl, apiKey, modelName, conversat
         stream: true,
         stream_options: { include_usage: true },
     };
-    // Use streaming to avoid timeout
-    const stream = (await client.chat.completions.create(requestPayload, {
-        headers: customHeaders,
-    }));
+    const response = await fetch(`${baseUrl}/chat/completions`, {
+        method: 'POST',
+        headers: {
+            'Content-Type': 'application/json',
+            'Authorization': `Bearer ${apiKey}`,
+            ...customHeaders
+        },
+        body: JSON.stringify(requestPayload)
+    });
+    if (!response.ok) {
+        const errorText = await response.text();
+        throw new Error(`OpenAI API error: ${response.status} ${response.statusText} - ${errorText}`);
+    }
+    if (!response.body) {
+        throw new Error('No response body from OpenAI API');
+    }
     let summary = '';
     let usage = {
         prompt_tokens: 0,
         completion_tokens: 0,
         total_tokens: 0,
     };
-    for await (const chunk of stream) {
+    for await (const chunk of parseSSEStream(response.body.getReader())) {
         const delta = chunk.choices[0]?.delta;
         if (delta?.content) {
             summary += delta.content;
@@ -84,10 +121,6 @@ async function compressWithChatCompletions(baseUrl, apiKey, modelName, conversat
  * Compress context using OpenAI Responses API
  */
 async function compressWithResponses(baseUrl, apiKey, modelName, conversationMessages, systemPrompt) {
-    const client = new OpenAI({
-        apiKey,
-        baseURL: baseUrl,
-    });
     const customHeaders = getCustomHeaders();
     // Build instructions
     const instructions = systemPrompt || SYSTEM_PROMPT;
@@ -130,17 +163,29 @@ async function compressWithResponses(baseUrl, apiKey, modelName, conversationMes
         input,
         stream: true,
     };
-    // Use streaming to avoid timeout
-    const stream = await client.responses.create(requestPayload, {
-        headers: customHeaders,
+    const response = await fetch(`${baseUrl}/responses`, {
+        method: 'POST',
+        headers: {
+            'Content-Type': 'application/json',
+            'Authorization': `Bearer ${apiKey}`,
+            ...customHeaders
+        },
+        body: JSON.stringify(requestPayload)
     });
+    if (!response.ok) {
+        const errorText = await response.text();
+        throw new Error(`OpenAI Responses API error: ${response.status} ${response.statusText} - ${errorText}`);
+    }
+    if (!response.body) {
+        throw new Error('No response body from OpenAI Responses API');
+    }
     let summary = '';
     let usage = {
         prompt_tokens: 0,
         completion_tokens: 0,
         total_tokens: 0,
     };
-    for await (const chunk of stream) {
+    for await (const chunk of parseSSEStream(response.body.getReader())) {
         const eventType = chunk.type;
         // Handle text content delta
         if (eventType === 'response.output_text.delta') {
@@ -150,7 +195,7 @@ async function compressWithResponses(baseUrl, apiKey, modelName, conversationMes
             }
         }
         // Handle usage info
-        if (eventType === 'response.done') {
+        if (eventType === 'response.completed') {
             const response = chunk.response;
             if (response?.usage) {
                 usage = {
@@ -173,26 +218,7 @@ async function compressWithResponses(baseUrl, apiKey, modelName, conversationMes
  * Compress context using Gemini API
  */
 async function compressWithGemini(baseUrl, apiKey, modelName, conversationMessages, systemPrompt) {
-    const clientConfig = {
-        apiKey,
-    };
     const customHeaders = getCustomHeaders();
-    // Support custom baseUrl and headers for proxy servers
-    if (baseUrl && baseUrl !== 'https://api.openai.com/v1') {
-        clientConfig.httpOptions = {
-            baseUrl,
-            headers: {
-                'x-goog-api-key': apiKey,
-                ...customHeaders,
-            },
-        };
-    }
-    else if (Object.keys(customHeaders).length > 0) {
-        clientConfig.httpOptions = {
-            headers: customHeaders,
-        };
-    }
-    const client = new GoogleGenAI(clientConfig);
     // Build system instruction
     const systemInstruction = systemPrompt || SYSTEM_PROMPT;
     // Build contents array with conversation history
@@ -220,30 +246,81 @@ async function compressWithGemini(baseUrl, apiKey, modelName, conversationMessag
                 text: COMPRESSION_PROMPT,
             }],
     });
-    const requestConfig = {
-        model: modelName,
-        systemInstruction,
+    const requestBody = {
         contents,
+        systemInstruction: systemInstruction ? { parts: [{ text: systemInstruction }] } : undefined,
     };
-    // Use streaming to avoid timeout
-    const stream = await client.models.generateContentStream(requestConfig);
+    // Extract model name
+    const effectiveBaseUrl = baseUrl && baseUrl !== 'https://api.openai.com/v1'
+        ? baseUrl
+        : 'https://generativelanguage.googleapis.com/v1beta';
+    const model = modelName.startsWith('models/') ? modelName : `models/${modelName}`;
+    const url = `${effectiveBaseUrl}/${model}:streamGenerateContent?key=${apiKey}&alt=sse`;
+    const response = await fetch(url, {
+        method: 'POST',
+        headers: {
+            'Content-Type': 'application/json',
+            'Authorization': `Bearer ${apiKey}`,
+            ...customHeaders
+        },
+        body: JSON.stringify(requestBody)
+    });
+    if (!response.ok) {
+        const errorText = await response.text();
+        throw new Error(`Gemini API error: ${response.status} ${response.statusText} - ${errorText}`);
+    }
+    if (!response.body) {
+        throw new Error('No response body from Gemini API');
+    }
     let summary = '';
     let usage = {
         prompt_tokens: 0,
         completion_tokens: 0,
         total_tokens: 0,
     };
-    for await (const chunk of stream) {
-        if (chunk.text) {
-            summary += chunk.text;
-        }
-        // Collect usage info
-        if (chunk.usageMetadata) {
-            usage = {
-                prompt_tokens: chunk.usageMetadata.promptTokenCount || 0,
-                completion_tokens: chunk.usageMetadata.candidatesTokenCount || 0,
-                total_tokens: chunk.usageMetadata.totalTokenCount || 0,
-            };
+    // Parse SSE stream
+    const reader = response.body.getReader();
+    const decoder = new TextDecoder();
+    let buffer = '';
+    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) {
+            const trimmed = line.trim();
+            if (!trimmed || trimmed.startsWith(':'))
+                continue;
+            if (trimmed.startsWith('data: ')) {
+                const data = trimmed.slice(6);
+                try {
+                    const chunk = JSON.parse(data);
+                    // Process candidates
+                    if (chunk.candidates && chunk.candidates.length > 0) {
+                        const candidate = chunk.candidates[0];
+                        if (candidate.content && candidate.content.parts) {
+                            for (const part of candidate.content.parts) {
+                                if (part.text) {
+                                    summary += part.text;
+                                }
+                            }
+                        }
+                    }
+                    // Collect usage info
+                    if (chunk.usageMetadata) {
+                        usage = {
+                            prompt_tokens: chunk.usageMetadata.promptTokenCount || 0,
+                            completion_tokens: chunk.usageMetadata.candidatesTokenCount || 0,
+                            total_tokens: chunk.usageMetadata.totalTokenCount || 0,
+                        };
+                    }
+                }
+                catch (e) {
+                    console.error('Failed to parse Gemini SSE data:', data);
+                }
+            }
         }
     }
     if (!summary) {
@@ -258,18 +335,7 @@ async function compressWithGemini(baseUrl, apiKey, modelName, conversationMessag
  * Compress context using Anthropic API
  */
 async function compressWithAnthropic(baseUrl, apiKey, modelName, conversationMessages, systemPrompt) {
-    const clientConfig = {
-        apiKey,
-    };
-    if (baseUrl && baseUrl !== 'https://api.openai.com/v1') {
-        clientConfig.baseURL = baseUrl;
-    }
     const customHeaders = getCustomHeaders();
-    clientConfig.defaultHeaders = {
-        'Authorization': `Bearer ${apiKey}`,
-        ...customHeaders,
-    };
-    const client = new Anthropic(clientConfig);
     // Build messages array with conversation history
     const messages = [];
     // If custom system prompt exists, add default as first user message
@@ -298,27 +364,73 @@ async function compressWithAnthropic(baseUrl, apiKey, modelName, conversationMes
         max_tokens: 4096,
         system: systemParam,
         messages,
+        stream: true
     };
-    // Use streaming to avoid timeout
-    const stream = await client.messages.stream(requestPayload);
+    const effectiveBaseUrl = baseUrl && baseUrl !== 'https://api.openai.com/v1'
+        ? baseUrl
+        : 'https://api.anthropic.com/v1';
+    const response = await fetch(`${effectiveBaseUrl}/messages`, {
+        method: 'POST',
+        headers: {
+            'Content-Type': 'application/json',
+            'x-api-key': apiKey,
+            'authorization': `Bearer ${apiKey}`,
+            ...customHeaders
+        },
+        body: JSON.stringify(requestPayload)
+    });
+    if (!response.ok) {
+        const errorText = await response.text();
+        throw new Error(`Anthropic API error: ${response.status} ${response.statusText} - ${errorText}`);
+    }
+    if (!response.body) {
+        throw new Error('No response body from Anthropic API');
+    }
     let summary = '';
     let usage = {
         prompt_tokens: 0,
         completion_tokens: 0,
         total_tokens: 0,
     };
-    for await (const event of stream) {
-        if (event.type === 'content_block_delta' && event.delta.type === 'text_delta') {
-            summary += event.delta.text;
-        }
-        // Collect usage info from message_stop event
-        if (event.type === 'message_stop') {
-            const finalMessage = await stream.finalMessage();
-            usage = {
-                prompt_tokens: finalMessage.usage.input_tokens,
-                completion_tokens: finalMessage.usage.output_tokens,
-                total_tokens: finalMessage.usage.input_tokens + finalMessage.usage.output_tokens,
-            };
+    // Parse Anthropic SSE stream
+    const reader = response.body.getReader();
+    const decoder = new TextDecoder();
+    let buffer = '';
+    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) {
+            const trimmed = line.trim();
+            if (!trimmed || trimmed.startsWith(':'))
+                continue;
+            if (trimmed.startsWith('event: ')) {
+                continue;
+            }
+            if (trimmed.startsWith('data: ')) {
+                const data = trimmed.slice(6);
+                try {
+                    const event = JSON.parse(data);
+                    if (event.type === 'content_block_delta' && event.delta?.type === 'text_delta') {
+                        summary += event.delta.text;
+                    }
+                    // Collect usage info from message_start event
+                    if (event.type === 'message_start' && event.message?.usage) {
+                        usage.prompt_tokens = event.message.usage.input_tokens || 0;
+                    }
+                    // Collect usage info from message_delta event
+                    if (event.type === 'message_delta' && event.usage) {
+                        usage.completion_tokens = event.usage.output_tokens || 0;
+                        usage.total_tokens = usage.prompt_tokens + usage.completion_tokens;
+                    }
+                }
+                catch (e) {
+                    console.error('Failed to parse Anthropic SSE data:', data);
+                }
+            }
         }
     }
     if (!summary) {

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "snow-ai",
-	"version": "0.3.0",
+	"version": "0.3.2",
 	"description": "Intelligent Command Line Assistant powered by AI",
 	"license": "MIT",
 	"bin": {
@@ -39,8 +39,6 @@
 		"dist"
 	],
 	"dependencies": {
-		"@anthropic-ai/sdk": "^0.65.0",
-		"@google/genai": "^1.23.0",
 		"@inkjs/ui": "^2.0.0",
 		"@modelcontextprotocol/sdk": "^1.17.3",
 		"chalk-template": "^1.1.2",
@@ -57,7 +55,6 @@
 		"ink-text-input": "^6.0.0",
 		"ink-tree-select": "^2.3.1",
 		"meow": "^11.0.0",
-		"openai": "^6.1.0",
 		"puppeteer-core": "^24.25.0",
 		"react": "^18.2.0",
 		"string-width": "^7.2.0",