@dreb/coding-agent 2.59.1 → 2.59.2

package/dist/core/agent-session.js

@@ -20,7 +20,7 @@ import { getDocsPath } from "../config.js";
 import { theme } from "../modes/interactive/theme/theme.js";
 import { sleep } from "../utils/sleep.js";
 import { executeBash as executeBashCommand, executeBashWithOperations } from "./bash-executor.js";
-import { calculateContextTokens, collectEntriesForBranchSummary, compact, estimateContextTokens, generateBranchSummary, prepareCompaction, shouldCompact, } from "./compaction/index.js";
+import { calculateContextTokens, collectEntriesForBranchSummary, compact, estimateContextTokens, estimateTokens, generateBranchSummary, prepareCompaction, shouldCompact, } from "./compaction/index.js";
 import { DEFAULT_THINKING_LEVEL } from "./defaults.js";
 import { DispatchArbiter } from "./dispatch-arbiter.js";
 import { exportSessionToHtml } from "./export-html/index.js";
@@ -3346,11 +3346,12 @@ export class AgentSession {
             return undefined;
         // After compaction, the last assistant usage reflects pre-compaction context size.
         // We can only trust usage from an assistant that responded after the latest compaction.
-        // If no such assistant exists, context token count is unknown until the next LLM response.
+        // Until then, estimate every rebuilt message independently so the stale kept
+        // assistant usage cannot leak into the current context value.
         const branchEntries = this.sessionManager.getBranch();
         const latestCompaction = getLatestCompactionEntry(branchEntries);
         if (latestCompaction) {
-            // Check if there's a valid assistant usage after the compaction boundary
+            // Check if there's a valid assistant usage after the compaction boundary.
             const compactionIndex = branchEntries.lastIndexOf(latestCompaction);
             let hasPostCompactionUsage = false;
             for (let i = branchEntries.length - 1; i > compactionIndex; i--) {
@@ -3359,23 +3360,22 @@ export class AgentSession {
                     const assistant = entry.message;
                     if (assistant.stopReason !== "aborted" && assistant.stopReason !== "error") {
                         const contextTokens = calculateContextTokens(assistant.usage);
-                        if (contextTokens > 0) {
+                        if (contextTokens > 0)
                             hasPostCompactionUsage = true;
-                        }
                         break;
                     }
                 }
             }
             if (!hasPostCompactionUsage) {
-                return { tokens: null, contextWindow, percent: null };
+                const tokens = this.messages.reduce((total, message) => total + estimateTokens(message), 0);
+                return { tokens, contextWindow, percent: (tokens / contextWindow) * 100 };
             }
         }
         const estimate = estimateContextTokens(this.messages);
-        const percent = (estimate.tokens / contextWindow) * 100;
         return {
             tokens: estimate.tokens,
             contextWindow,
-            percent,
+            percent: (estimate.tokens / contextWindow) * 100,
         };
     }
     /**