optimize chat token allocation for maximum input/output space

- replace aggressive 1.5x buffer with smart model-based allocation
- 128k+ models: only 2k buffer (was wasting 10k+ tokens)
- 32k models: 1.5k buffer max
- 8k models: conservative buffer
- follow-up responses now have maximum room for both history and output
- especially important for long conversations on large context models

src/chat.js

@@ -13,7 +13,21 @@ function estimateTokens(text) {
 function calculateAvailableTokens(chatHistory, maxTokens) {
 	const inputText = chatHistory.map(msg => msg.content).join('');
 	const inputTokens = estimateTokens(inputText);
-	const buffer = Math.max(2048, Math.floor(inputTokens * 1.5)); // Use 1.5x input tokens as buffer, minimum 2048
+	
+	// For large context models (>32k), use a more reasonable buffer
+	// For smaller models, be more conservative
+	let buffer;
+	if (maxTokens > 32000) {
+		// Large context models: just reserve 10% or 2k tokens for safety
+		buffer = Math.min(2000, Math.floor(maxTokens * 0.1));
+	} else if (maxTokens > 8000) {
+		// Medium models: reserve 20% or 1.5k tokens
+		buffer = Math.min(1500, Math.floor(maxTokens * 0.2));
+	} else {
+		// Small models: keep conservative approach
+		buffer = Math.max(1000, Math.floor(inputTokens * 0.5));
+	}
+	
 	const availableTokens = Math.max(100, maxTokens - inputTokens - buffer);
 	console.log(`Token calculation: max=${maxTokens}, input≈${inputTokens}, buffer=${buffer}, available≈${availableTokens}`);
 	return availableTokens;