@bd7pil/opencode-deep-memory 0.6.0 → 0.7.0

package/dist/index.js

@@ -15284,12 +15284,10 @@ function createCompactingHandler(args) {
 }
 
 // src/compress/pressure.ts
-var FALLBACK_MAX_CONTEXT = 128e3;
+var FALLBACK_MAX_CONTEXT = 1e6;
 var OPENCODE_COMPACTION_RATIO = 0.75;
-var THRESHOLDS = {
-  medium: 0.2,
-  high: 0.4
-};
+var PRESSURE_MEDIUM_TOKENS = 5e4;
+var PRESSURE_HIGH_TOKENS = 15e4;
 var calibratedMaxContext = 0;
 function calibrateFromCompaction(lastInputTokens) {
   if (lastInputTokens <= 0) return;
@@ -15303,67 +15301,33 @@ function maxContextFrom(modelContextWindow) {
   if (calibratedMaxContext > 0) return calibratedMaxContext;
   return FALLBACK_MAX_CONTEXT;
 }
-function estimateTokens2(text) {
-  let cjk = 0;
-  let other = 0;
-  for (const ch of text) {
-    if (/[\u4e00-\u9fff\u3400-\u4dbf\u3000-\u303f\uff00-\uffef\u3040-\u309f\u30a0-\u30ff]/.test(ch)) {
-      cjk++;
-    } else {
-      other++;
-    }
-  }
-  return Math.ceil(cjk * 0.7 + other / 3.8);
-}
-function extractTokensFromMessages(messages) {
-  let total = 0;
-  for (const msg of messages) {
-    for (const part of msg.parts) {
-      if (typeof part !== "object" || part === null) continue;
-      const p = part;
-      if (p["type"] === "text" && typeof p["text"] === "string") {
-        total += estimateTokens2(p["text"]);
-      } else if (p["type"] === "tool") {
-        const state = p["state"];
-        if (state?.["output"] && typeof state["output"] === "string") {
-          total += estimateTokens2(state["output"]);
-        }
-        if (state?.["error"] && typeof state["error"] === "string") {
-          total += estimateTokens2(state["error"]);
-        }
-      } else if (p["type"] === "reasoning" || p["type"] === "thinking") {
-        if (typeof p["text"] === "string") {
-          total += estimateTokens2(p["text"]);
-        }
-      }
-    }
-  }
-  return total;
-}
 function extractInputTokensFromMessages(messages) {
+  let best = 0;
   for (let i = messages.length - 1; i >= 0; i--) {
     const msg = messages[i];
     for (const part of msg.parts) {
       if (typeof part !== "object" || part === null) continue;
       const p = part;
       if (p["type"] === "step-finish") {
-        const tokens = p.tokens;
-        if (tokens?.input && tokens.input > 0) {
-          return tokens.input;
-        }
+        const tokens = p;
+        const input = tokens.tokens?.input ?? 0;
+        const cached2 = tokens.tokens?.cached ?? 0;
+        const total = input + cached2;
+        if (total > best) best = total;
+        if (best > 0) return best;
       }
     }
   }
-  return 0;
+  return best;
 }
 function detectPressure(messages, modelContextWindow) {
   const ctx = maxContextFrom(modelContextWindow || 0);
   const inputTokens = extractInputTokensFromMessages(messages);
-  const estimated = inputTokens > 0 ? inputTokens : extractTokensFromMessages(messages);
+  const estimated = inputTokens > 0 ? inputTokens : 0;
   const ratio = Math.min(estimated / ctx, 1);
   let level;
-  if (ratio >= THRESHOLDS.high) level = "high";
-  else if (ratio >= THRESHOLDS.medium) level = "medium";
+  if (estimated >= PRESSURE_HIGH_TOKENS) level = "high";
+  else if (estimated >= PRESSURE_MEDIUM_TOKENS) level = "medium";
   else level = "low";
   return { level, ratio, estimatedTokens: estimated, maxContext: ctx };
 }
@@ -15627,6 +15591,8 @@ var PROTECTED_TOOLS = /* @__PURE__ */ new Set([
 var NEVER_DEDUP = /* @__PURE__ */ new Set(["read", "bash", "grep", "glob", "find", "search"]);
 var ERROR_PURGE_TURN_THRESHOLD = 4;
 var PROTECTED_HEAD_SINGLE = 2;
+var ASSISTANT_COMPRESS_MIN_LENGTH = 500;
+var ASSISTANT_COMPRESS_SAVINGS_RATIO = 0.6;
 function simpleHash(s) {
   const len = s.length;
   const sampleSize = 500;
@@ -15640,12 +15606,32 @@ function simpleHash(s) {
   }
   return `${len}:${h.toString(36)}`;
 }
+function compressAssistantText(text) {
+  if (text.length < ASSISTANT_COMPRESS_MIN_LENGTH) return text;
+  const lines = text.split("\n");
+  const head = 3;
+  const tail = 3;
+  const kept = [];
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i];
+    if (i < head || i >= lines.length - tail) {
+      kept.push(line);
+      continue;
+    }
+    if (/^#{1,3}\s/.test(line) || /error|fail|warning|critical|important/i.test(line) || /^\s*[-*]\s/.test(line) || /^\s*\d+\.\s/.test(line) || line.trim().startsWith("```") || /^\/[^\s:]+/.test(line)) {
+      kept.push(line);
+    }
+  }
+  const result = kept.join("\n");
+  return result.length < text.length * ASSISTANT_COMPRESS_SAVINGS_RATIO ? result : text;
+}
 function singlePassCompress(messages, state, protectedTail) {
   const stats = {
     toolDedup: 0,
     errorPurge: 0,
     toolOutputCompressed: 0,
     jsonCrushed: 0,
+    assistantCompressed: 0,
     ccrStored: 0
   };
   const totalMessages = messages.length;
@@ -15725,6 +15711,21 @@ function singlePassCompress(messages, state, protectedTail) {
         }
       }
     }
+    if (i < protectedTail && msg.info.role === "assistant") {
+      for (let j = msg.parts.length - 1; j >= 0; j--) {
+        const part = msg.parts[j];
+        if (typeof part !== "object" || part === null) continue;
+        const p = part;
+        if (p["type"] !== "text") continue;
+        const text = p["text"];
+        if (typeof text !== "string") continue;
+        const compressed = compressAssistantText(text);
+        if (compressed !== text) {
+          p["text"] = compressed;
+          stats.assistantCompressed++;
+        }
+      }
+    }
   }
   return stats;
 }
@@ -15766,6 +15767,7 @@ function runCompressionPipeline(ctx) {
     errorPurge: spStats.errorPurge,
     toolOutputCompressed: spStats.toolOutputCompressed,
     jsonCrushed: spStats.jsonCrushed,
+    assistantCompressed: spStats.assistantCompressed,
     ccrStored: spStats.ccrStored,
     nudgeInjected: false,
     pressureLevel: pressure.level,
@@ -15786,7 +15788,7 @@ function runCompressionPipeline(ctx) {
       logger?.debug("compress: memory nudge", { type: memoryNudge.type });
     }
   }
-  const active = stats.toolDedup > 0 || stats.errorPurge > 0 || stats.toolOutputCompressed > 0 || stats.jsonCrushed > 0 || stats.nudgeInjected;
+  const active = stats.toolDedup > 0 || stats.errorPurge > 0 || stats.toolOutputCompressed > 0 || stats.jsonCrushed > 0 || stats.assistantCompressed > 0 || stats.nudgeInjected;
   if (active) {
     logger?.debug("compress: pipeline result", { ...stats });
   } else {
@@ -15976,7 +15978,7 @@ function createMessagesTransformHandler(state, logger) {
       logger
     });
     const ds = pipelineResult.stats;
-    if (ds.toolDedup > 0 || ds.errorPurge > 0 || ds.toolOutputCompressed > 0 || ds.jsonCrushed > 0 || ds.nudgeInjected) {
+    if (ds.toolDedup > 0 || ds.errorPurge > 0 || ds.toolOutputCompressed > 0 || ds.jsonCrushed > 0 || ds.assistantCompressed > 0 || ds.nudgeInjected) {
       logger?.debug("messages.transform: deep compression", { ...ds });
       state.mergeNotify({
         compression: stats,
@@ -16116,6 +16118,29 @@ function createNotifyHandler(client, logger) {
   };
 }
 
+// src/shared/model-limits.ts
+var KNOWN_MODEL_LIMITS = {
+  "deepseek-v4-pro": 1e6,
+  "deepseek-v4": 1e6,
+  "deepseek-v3": 64e3,
+  "deepseek-r1": 64e3,
+  "claude-opus-4": 2e5,
+  "claude-sonnet-4": 2e5,
+  "gpt-4o": 128e3,
+  "o1": 2e5,
+  "o3-mini": 2e5,
+  "gemini-2.5-pro": 1e6,
+  "gemini-2.5-flash": 1e6,
+  "qwen-max": 131072
+};
+function lookupModelLimit(modelID) {
+  if (KNOWN_MODEL_LIMITS[modelID]) return KNOWN_MODEL_LIMITS[modelID];
+  for (const [key, limit] of Object.entries(KNOWN_MODEL_LIMITS)) {
+    if (modelID.includes(key)) return limit;
+  }
+  return void 0;
+}
+
 // src/extract/enrich.ts
 import { stat } from "fs/promises";
 
@@ -16447,11 +16472,14 @@ var deepMemoryPlugin = async (input) => {
       const defaultModel = configResult.data?.model;
       if (typeof defaultModel === "string" && defaultModel.includes("/")) {
         const slashIdx = defaultModel.indexOf("/");
-        state.recordFallbackModel({
-          providerID: defaultModel.slice(0, slashIdx),
-          modelID: defaultModel.slice(slashIdx + 1)
-        });
-        logger.debug("resolved fallback model from config", { defaultModel });
+        const providerID = defaultModel.slice(0, slashIdx);
+        const modelID = defaultModel.slice(slashIdx + 1);
+        state.recordFallbackModel({ providerID, modelID });
+        const limit = lookupModelLimit(modelID);
+        if (limit) {
+          state.setModelContextWindow(limit);
+          logger.debug("resolved model context window", { modelID, limit });
+        }
       }
     }).catch((err) => {
       logger.debug("config.get failed, dream/distill will omit model", {