@psnext/slingcli 2.5.20260722-2 → 2.5.20260722-3

package/node_modules/@earendil-works/pi-coding-agent/dist/core/agent-session.js

@@ -14,6 +14,7 @@
  */
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { basename, dirname } from "node:path";
+import { contentText } from "@earendil-works/pi-ai";
 import { clampThinkingLevel, cleanupSessionResources, getSupportedThinkingLevels, isContextOverflow, isRetryableAssistantError, modelsAreEqual, resetApiProviders, streamSimple, } from "@earendil-works/pi-ai/compat";
 import { getThemeByName, theme } from "../modes/interactive/theme/theme.js";
 import { stripFrontmatter } from "../utils/frontmatter.js";
@@ -35,6 +36,7 @@ import { buildSystemPrompt } from "./system-prompt.js";
 import { createLocalBashOperations } from "./tools/bash.js";
 import { createAllToolDefinitions } from "./tools/index.js";
 import { createToolDefinitionFromAgentTool } from "./tools/tool-definition-wrapper.js";
+import { addUsageToTotals, createUsageTotals } from "./usage-totals.js";
 /**
  * Parse a skill block from message text.
  * Returns null if the text doesn't contain a skill block.
@@ -188,7 +190,7 @@ export class AgentSession {
         throw new Error(formatNoApiKeyFoundMessage(model.provider));
     }
     async _getSummarizationRequestAuth(model) {
-        if (this.agent.streamFn === streamSimple) {
+        if (this.agent.streamFunction === streamSimple) {
             return this._getRequiredRequestAuth(model);
         }
         try {
@@ -243,6 +245,7 @@ export class AgentSession {
                 content: result.content,
                 details: result.details,
                 isError,
+                usage: result.usage,
             });
             if (!hookResult) {
                 return undefined;
@@ -251,6 +254,7 @@ export class AgentSession {
                 content: hookResult.content,
                 details: hookResult.details,
                 isError: hookResult.isError ?? isError,
+                usage: hookResult.usage,
             };
         };
     }
@@ -325,7 +329,7 @@ export class AgentSession {
         // This ensures the UI sees the updated queue state
         if (event.type === "message_start" && event.message.role === "user") {
             this._overflowRecoveryAttempted = false;
-            const messageText = this._getUserMessageText(event.message);
+            const messageText = contentText(event.message.content, "");
             if (messageText) {
                 // Check steering queue first
                 const steeringIndex = this._steeringMessages.indexOf(messageText);
@@ -394,16 +398,6 @@ export class AgentSession {
         }
         return false;
     }
-    /** Extract text content from a message */
-    _getUserMessageText(message) {
-        if (message.role !== "user")
-            return "";
-        const content = message.content;
-        if (typeof content === "string")
-            return content;
-        const textBlocks = content.filter((c) => c.type === "text");
-        return textBlocks.map((c) => c.text).join("");
-    }
     /** Find the last assistant message in agent state (including aborted ones) */
     _findLastAssistantMessage() {
         const messages = this.agent.state.messages;
@@ -570,7 +564,7 @@ export class AgentSession {
         catch {
             // Dispose must succeed even if an abort hook throws.
         }
-        this._extensionRunner.invalidate("This extension ctx is stale after session replacement or reload. Do not use a captured sling or command ctx after ctx.newSession(), ctx.fork(), ctx.switchSession(), or ctx.reload(). For newSession, fork, and switchSession, move post-replacement work into withSession and use the ctx passed to withSession. For reload, do not use the old ctx after await ctx.reload().");
+        this._extensionRunner.invalidate("This extension ctx is stale after session replacement or reload. Do not use a captured pi or command ctx after ctx.newSession(), ctx.fork(), ctx.switchSession(), or ctx.reload(). For newSession, fork, and switchSession, move post-replacement work into withSession and use the ctx passed to withSession. For reload, do not use the old ctx after await ctx.reload().");
         this._disconnectFromAgent();
         this._eventListeners = [];
         cleanupSessionResources(this.sessionId);
@@ -1414,26 +1408,29 @@ export class AgentSession {
             let summary;
             let firstKeptEntryId;
             let tokensBefore;
+            let usage;
             let details;
             if (extensionCompaction) {
                 // Extension provided compaction content
                 summary = extensionCompaction.summary;
                 firstKeptEntryId = extensionCompaction.firstKeptEntryId;
                 tokensBefore = extensionCompaction.tokensBefore;
+                usage = extensionCompaction.usage;
                 details = extensionCompaction.details;
             }
             else {
                 // Generate compaction result
-                const result = await compact(preparation, this.model, apiKey, headers, customInstructions, this._compactionAbortController.signal, this.thinkingLevel, this.agent.streamFn, env);
+                const result = await compact(preparation, this.model, apiKey, headers, customInstructions, this._compactionAbortController.signal, this.thinkingLevel, this.agent.streamFunction, env, this.settingsManager.getRetrySettings(), this._summarizationRetryCallbacks({ source: "compaction", reason: "manual" }));
                 summary = result.summary;
                 firstKeptEntryId = result.firstKeptEntryId;
                 tokensBefore = result.tokensBefore;
+                usage = result.usage;
                 details = result.details;
             }
             if (this._compactionAbortController.signal.aborted) {
                 throw new Error("Compaction cancelled");
             }
-            this.sessionManager.appendCompaction(summary, firstKeptEntryId, tokensBefore, details, fromExtension);
+            this.sessionManager.appendCompaction(summary, firstKeptEntryId, tokensBefore, details, fromExtension, usage);
             const newEntries = this.sessionManager.getEntries();
             const sessionContext = this.sessionManager.buildSessionContext();
             this.agent.state.messages = sessionContext.messages;
@@ -1454,6 +1451,7 @@ export class AgentSession {
                 firstKeptEntryId,
                 tokensBefore,
                 estimatedTokensAfter,
+                usage,
                 details,
             };
             this._emit({
@@ -1600,7 +1598,7 @@ export class AgentSession {
             let apiKey;
             let headers;
             let env;
-            if (this.agent.streamFn === streamSimple) {
+            if (this.agent.streamFunction === streamSimple) {
                 const authResult = await this._modelRuntime.getAuth(this.model);
                 if (!authResult?.auth.apiKey)
                     return false;
@@ -1649,20 +1647,23 @@ export class AgentSession {
             let summary;
             let firstKeptEntryId;
             let tokensBefore;
+            let usage;
             let details;
             if (extensionCompaction) {
                 // Extension provided compaction content
                 summary = extensionCompaction.summary;
                 firstKeptEntryId = extensionCompaction.firstKeptEntryId;
                 tokensBefore = extensionCompaction.tokensBefore;
+                usage = extensionCompaction.usage;
                 details = extensionCompaction.details;
             }
             else {
                 // Generate compaction result
-                const compactResult = await compact(preparation, this.model, apiKey, headers, undefined, this._autoCompactionAbortController.signal, this.thinkingLevel, this.agent.streamFn, env);
+                const compactResult = await compact(preparation, this.model, apiKey, headers, undefined, this._autoCompactionAbortController.signal, this.thinkingLevel, this.agent.streamFunction, env, this.settingsManager.getRetrySettings(), this._summarizationRetryCallbacks({ source: "compaction", reason }));
                 summary = compactResult.summary;
                 firstKeptEntryId = compactResult.firstKeptEntryId;
                 tokensBefore = compactResult.tokensBefore;
+                usage = compactResult.usage;
                 details = compactResult.details;
             }
             if (this._autoCompactionAbortController.signal.aborted) {
@@ -1675,7 +1676,7 @@ export class AgentSession {
                 });
                 return false;
             }
-            this.sessionManager.appendCompaction(summary, firstKeptEntryId, tokensBefore, details, fromExtension);
+            this.sessionManager.appendCompaction(summary, firstKeptEntryId, tokensBefore, details, fromExtension, usage);
             const newEntries = this.sessionManager.getEntries();
             const sessionContext = this.sessionManager.buildSessionContext();
             this.agent.state.messages = sessionContext.messages;
@@ -1696,6 +1697,7 @@ export class AgentSession {
                 firstKeptEntryId,
                 tokensBefore,
                 estimatedTokensAfter,
+                usage,
                 details,
             };
             this._emit({ type: "compaction_end", reason, result, aborted: false, willRetry });
@@ -2086,6 +2088,34 @@ export class AgentSession {
             return false;
         return isRetryableAssistantError(message);
     }
+    /**
+     * Retry policy + callbacks shared by compaction and branch-summary summarization calls.
+     * Uses the same `settings.retry` budget/backoff as agent-turn retries so a single transient
+     * stream drop no longer fails the whole operation. `source` carries the context
+     * the TUI needs to render the retry and recreate the underlying indicator.
+     */
+    _summarizationRetryCallbacks(source) {
+        return {
+            onRetryScheduled: (attempt, maxAttempts, delayMs, errorMessage) => {
+                this._emit({
+                    type: "summarization_retry_scheduled",
+                    attempt,
+                    maxAttempts,
+                    delayMs,
+                    errorMessage,
+                });
+            },
+            onRetryAttemptStart: () => {
+                this._emit({
+                    type: "summarization_retry_attempt_start",
+                    ...source,
+                });
+            },
+            onRetryFinished: () => {
+                this._emit({ type: "summarization_retry_finished" });
+            },
+        };
+    }
     /**
      * Prepare a retryable error for continuation with exponential backoff.
      * @returns true if the caller should continue the agent, false otherwise
@@ -2331,6 +2361,7 @@ export class AgentSession {
             // Run default summarizer if needed
             let summaryText;
             let summaryDetails;
+            let summaryUsage;
             if (options.summarize && entriesToSummarize.length > 0 && !extensionSummary) {
                 const model = this.model;
                 const { apiKey, headers, env } = await this._getSummarizationRequestAuth(model);
@@ -2344,7 +2375,9 @@ export class AgentSession {
                     customInstructions,
                     replaceInstructions,
                     reserveTokens: branchSummarySettings.reserveTokens,
-                    streamFn: this.agent.streamFn,
+                    streamFn: this.agent.streamFunction,
+                    retry: this.settingsManager.getRetrySettings(),
+                    callbacks: this._summarizationRetryCallbacks({ source: "branchSummary" }),
                 });
                 if (result.aborted) {
                     return { cancelled: true, aborted: true };
@@ -2353,6 +2386,7 @@ export class AgentSession {
                     throw new Error(result.error);
                 }
                 summaryText = result.summary;
+                summaryUsage = result.usage;
                 summaryDetails = {
                     readFiles: result.readFiles || [],
                     modifiedFiles: result.modifiedFiles || [],
@@ -2361,6 +2395,7 @@ export class AgentSession {
             else if (extensionSummary) {
                 summaryText = extensionSummary.summary;
                 summaryDetails = extensionSummary.details;
+                summaryUsage = extensionSummary.usage;
             }
             // Determine the new leaf position based on target type
             let newLeafId;
@@ -2368,18 +2403,12 @@ export class AgentSession {
             if (targetEntry.type === "message" && targetEntry.message.role === "user") {
                 // User message: leaf = parent (null if root), text goes to editor
                 newLeafId = targetEntry.parentId;
-                editorText = this._extractUserMessageText(targetEntry.message.content);
+                editorText = contentText(targetEntry.message.content, "");
             }
             else if (targetEntry.type === "custom_message") {
                 // Custom message: leaf = parent (null if root), text goes to editor
                 newLeafId = targetEntry.parentId;
-                editorText =
-                    typeof targetEntry.content === "string"
-                        ? targetEntry.content
-                        : targetEntry.content
-                            .filter((c) => c.type === "text")
-                            .map((c) => c.text)
-                            .join("");
+                editorText = contentText(targetEntry.content, "");
             }
             else {
                 // Non-user message: leaf = selected node
@@ -2390,7 +2419,7 @@ export class AgentSession {
             let summaryEntry;
             if (summaryText) {
                 // Create summary at target position (can be null for root)
-                const summaryId = this.sessionManager.branchWithSummary(newLeafId, summaryText, summaryDetails, fromExtension);
+                const summaryId = this.sessionManager.branchWithSummary(newLeafId, summaryText, summaryDetails, fromExtension, summaryUsage);
                 summaryEntry = this.sessionManager.getEntry(summaryId);
                 // Attach label to the summary entry
                 if (label) {
@@ -2438,24 +2467,13 @@ export class AgentSession {
                 continue;
             if (entry.message.role !== "user")
                 continue;
-            const text = this._extractUserMessageText(entry.message.content);
+            const text = contentText(entry.message.content, "");
             if (text) {
                 result.push({ entryId: entry.id, text });
             }
         }
         return result;
     }
-    _extractUserMessageText(content) {
-        if (typeof content === "string")
-            return content;
-        if (Array.isArray(content)) {
-            return content
-                .filter((c) => c.type === "text")
-                .map((c) => c.text)
-                .join("");
-        }
-        return "";
-    }
     /**
      * Get session statistics. Aggregates over ALL session entries (including
      * history that was compacted away), so token/cost totals reflect what was
@@ -2467,12 +2485,11 @@ export class AgentSession {
         let toolResults = 0;
         let totalMessages = 0;
         let toolCalls = 0;
-        let totalInput = 0;
-        let totalOutput = 0;
-        let totalCacheRead = 0;
-        let totalCacheWrite = 0;
-        let totalCost = 0;
+        const usageTotals = createUsageTotals();
         for (const entry of this.sessionManager.getEntries()) {
+            if ((entry.type === "branch_summary" || entry.type === "compaction") && entry.usage) {
+                addUsageToTotals(usageTotals, entry.usage);
+            }
             if (entry.type !== "message")
                 continue;
             totalMessages++;
@@ -2482,6 +2499,9 @@ export class AgentSession {
             }
             else if (message.role === "toolResult") {
                 toolResults++;
+                if (message.usage) {
+                    addUsageToTotals(usageTotals, message.usage);
+                }
             }
             else if (message.role === "assistant") {
                 assistantMessages++;
@@ -2489,12 +2509,7 @@ export class AgentSession {
                 if (Array.isArray(assistantMsg.content)) {
                     toolCalls += assistantMsg.content.filter((c) => c.type === "toolCall").length;
                 }
-                const usage = assistantMsg.usage;
-                totalInput += usage.input;
-                totalOutput += usage.output;
-                totalCacheRead += usage.cacheRead;
-                totalCacheWrite += usage.cacheWrite;
-                totalCost += usage.cost.total;
+                addUsageToTotals(usageTotals, assistantMsg.usage);
             }
         }
         return {
@@ -2506,13 +2521,13 @@ export class AgentSession {
             toolResults,
             totalMessages,
             tokens: {
-                input: totalInput,
-                output: totalOutput,
-                cacheRead: totalCacheRead,
-                cacheWrite: totalCacheWrite,
-                total: totalInput + totalOutput + totalCacheRead + totalCacheWrite,
+                input: usageTotals.input,
+                output: usageTotals.output,
+                cacheRead: usageTotals.cacheRead,
+                cacheWrite: usageTotals.cacheWrite,
+                total: usageTotals.input + usageTotals.output + usageTotals.cacheRead + usageTotals.cacheWrite,
             },
-            cost: totalCost,
+            cost: usageTotals.cost,
             contextUsage: this.getContextUsage(),
         };
     }

package/node_modules/@earendil-works/pi-coding-agent/dist/core/compaction/branch-summarization.js

@@ -4,9 +4,9 @@
  * When navigating to a different point in the session tree, this generates
  * a summary of the branch being left so context isn't lost.
  */
-import { completeSimple } from "@earendil-works/pi-ai/compat";
+import { contentText } from "@earendil-works/pi-ai";
 import { convertToLlm, createBranchSummaryMessage, createCompactionSummaryMessage, createCustomMessage, } from "../messages.js";
-import { estimateTokens } from "./compaction.js";
+import { completeSummarization, estimateTokens } from "./compaction.js";
 import { computeFileLists, createFileOps, extractFileOpsFromMessage, formatFileOperations, SUMMARIZATION_SYSTEM_PROMPT, serializeConversation, } from "./utils.js";
 // ============================================================================
 // Entry Collection
@@ -185,7 +185,7 @@ Keep each section concise. Preserve exact file paths, function names, and error
  * @param options - Generation options
  */
 export async function generateBranchSummary(entries, options) {
-    const { model, apiKey, headers, env, signal, customInstructions, replaceInstructions, reserveTokens = 16384, streamFn, } = options;
+    const { model, apiKey, headers, env, signal, customInstructions, replaceInstructions, reserveTokens = 16384, streamFn, retry, callbacks, } = options;
     // Token budget = context window minus reserved space for prompt + response
     const contextWindow = model.contextWindow || 128000;
     const tokenBudget = contextWindow - reserveTokens;
@@ -218,12 +218,11 @@ export async function generateBranchSummary(entries, options) {
     ];
     // Call LLM for summarization. Prefer the session stream function so SDK
     // request behavior (timeouts, retries, attribution headers) stays consistent
-    // without running through agent state/events.
+    // without running through agent state/events. Retried via completeSummarization
+    // so transient stream drops reuse the configured retry policy.
     const context = { systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages };
     const requestOptions = { apiKey, headers, env, signal, maxTokens: 2048 };
-    const response = streamFn
-        ? await (await streamFn(model, context, requestOptions)).result()
-        : await completeSimple(model, context, requestOptions);
+    const response = await completeSummarization(model, context, requestOptions, streamFn, retry, callbacks);
     // Check if aborted or errored
     if (response.stopReason === "aborted") {
         return { aborted: true };
@@ -231,10 +230,7 @@ export async function generateBranchSummary(entries, options) {
     if (response.stopReason === "error") {
         return { error: response.errorMessage || "Summarization failed" };
     }
-    let summary = response.content
-        .filter((c) => c.type === "text")
-        .map((c) => c.text)
-        .join("\n");
+    let summary = contentText(response.content);
     // Prepend preamble to provide context about the branch summary
     summary = BRANCH_SUMMARY_PREAMBLE + summary;
     // Compute file lists and append to summary
@@ -242,6 +238,7 @@ export async function generateBranchSummary(entries, options) {
     summary += formatFileOperations(readFiles, modifiedFiles);
     return {
         summary: summary || "No summary generated",
+        usage: response.usage,
         readFiles,
         modifiedFiles,
     };

package/node_modules/@earendil-works/pi-coding-agent/dist/core/compaction/compaction.js

@@ -4,6 +4,7 @@
  * Pure functions for compaction logic. The session manager handles I/O,
  * and after compaction the session is reloaded.
  */
+import { contentText, retryAssistantCall } from "@earendil-works/pi-ai";
 import { completeSimple } from "@earendil-works/pi-ai/compat";
 import { convertToLlm } from "../messages.js";
 import { buildSessionContext, sessionEntryToContextMessages, } from "../session-manager.js";
@@ -48,6 +49,28 @@ function getMessageFromEntryForCompaction(entry) {
     }
     return sessionEntryToContextMessages(entry)[0];
 }
+function combineUsage(first, second) {
+    return {
+        input: first.input + second.input,
+        output: first.output + second.output,
+        cacheRead: first.cacheRead + second.cacheRead,
+        cacheWrite: first.cacheWrite + second.cacheWrite,
+        ...(first.cacheWrite1h !== undefined || second.cacheWrite1h !== undefined
+            ? { cacheWrite1h: (first.cacheWrite1h ?? 0) + (second.cacheWrite1h ?? 0) }
+            : {}),
+        ...(first.reasoning !== undefined || second.reasoning !== undefined
+            ? { reasoning: (first.reasoning ?? 0) + (second.reasoning ?? 0) }
+            : {}),
+        totalTokens: first.totalTokens + second.totalTokens,
+        cost: {
+            input: first.cost.input + second.cost.input,
+            output: first.cost.output + second.cost.output,
+            cacheRead: first.cost.cacheRead + second.cost.cacheRead,
+            cacheWrite: first.cost.cacheWrite + second.cost.cacheWrite,
+            total: first.cost.total + second.cost.total,
+        },
+    };
+}
 export const DEFAULT_COMPACTION_SETTINGS = {
     enabled: true,
     reserveTokens: 16384,
@@ -407,18 +430,26 @@ function createSummarizationOptions(model, maxTokens, apiKey, headers, env, sign
     }
     return options;
 }
-async function completeSummarization(model, context, options, streamFn) {
-    if (!streamFn) {
-        return completeSimple(model, context, options);
-    }
-    const stream = await streamFn(model, context, options);
-    return stream.result();
+/**
+ * Shared choke point for every compaction/branch-summary summarization call. Wraps the
+ * single LLM call in {@link retryAssistantCall} so transient stream drops (e.g.
+ * `terminated`, socket close) honor the configured retry policy instead of failing
+ * the whole compaction on the first attempt. Deterministic errors and aborts return
+ * immediately (see {@link retryAssistantCall}).
+ */
+export async function completeSummarization(model, context, options, streamFn, retry, callbacks) {
+    const produce = async () => streamFn ? (await streamFn(model, context, options)).result() : completeSimple(model, context, options);
+    return retryAssistantCall(produce, retry, options.signal, callbacks);
 }
 /**
  * Generate a summary of the conversation using the LLM.
  * If previousSummary is provided, uses the update prompt to merge.
  */
-export async function generateSummary(currentMessages, model, reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env) {
+export async function generateSummary(currentMessages, model, reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env, retry, callbacks) {
+    return (await generateSummaryWithUsage(currentMessages, model, reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env, retry, callbacks)).text;
+}
+/** Generate or update a conversation summary and return its provider usage. */
+export async function generateSummaryWithUsage(currentMessages, model, reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env, retry, callbacks) {
     const maxTokens = Math.min(Math.floor(0.8 * reserveTokens), model.maxTokens > 0 ? model.maxTokens : Number.POSITIVE_INFINITY);
     // Use update prompt if we have a previous summary, otherwise initial prompt
     let basePrompt = previousSummary ? UPDATE_SUMMARIZATION_PROMPT : SUMMARIZATION_PROMPT;
@@ -443,15 +474,12 @@ export async function generateSummary(currentMessages, model, reserveTokens, api
         },
     ];
     const completionOptions = createSummarizationOptions(model, maxTokens, apiKey, headers, env, signal, thinkingLevel);
-    const response = await completeSummarization(model, { systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages }, completionOptions, streamFn);
+    const response = await completeSummarization(model, { systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages }, completionOptions, streamFn, retry, callbacks);
     if (response.stopReason === "error") {
         throw new Error(`Summarization failed: ${response.errorMessage || "Unknown error"}`);
     }
-    const textContent = response.content
-        .filter((c) => c.type === "text")
-        .map((c) => c.text)
-        .join("\n");
-    return textContent;
+    const textContent = contentText(response.content);
+    return { text: textContent, usage: response.usage };
 }
 export function prepareCompaction(pathEntries, settings) {
     if (pathEntries.length > 0 && pathEntries[pathEntries.length - 1].type === "compaction") {
@@ -544,21 +572,29 @@ Be concise. Focus on what's needed to understand the kept suffix.`;
  * @param preparation - Pre-calculated preparation from prepareCompaction()
  * @param customInstructions - Optional custom focus for the summary
  */
-export async function compact(preparation, model, apiKey, headers, customInstructions, signal, thinkingLevel, streamFn, env) {
+export async function compact(preparation, model, apiKey, headers, customInstructions, signal, thinkingLevel, streamFn, env, retry, callbacks) {
     const { firstKeptEntryId, messagesToSummarize, turnPrefixMessages, isSplitTurn, tokensBefore, previousSummary, fileOps, settings, } = preparation;
     // Generate summaries and merge into one
     let summary;
+    let summaryUsage;
     if (isSplitTurn && turnPrefixMessages.length > 0) {
-        const historyResult = messagesToSummarize.length > 0
-            ? await generateSummary(messagesToSummarize, model, settings.reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env)
-            : "No prior history.";
-        const turnPrefixResult = await generateTurnPrefixSummary(turnPrefixMessages, model, settings.reserveTokens, apiKey, headers, env, signal, thinkingLevel, streamFn);
+        let historyText = "No prior history.";
+        let historyUsage;
+        if (messagesToSummarize.length > 0) {
+            const historyResult = await generateSummaryWithUsage(messagesToSummarize, model, settings.reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env, retry, callbacks);
+            historyText = historyResult.text;
+            historyUsage = historyResult.usage;
+        }
+        const turnPrefixResult = await generateTurnPrefixSummary(turnPrefixMessages, model, settings.reserveTokens, apiKey, headers, env, signal, thinkingLevel, streamFn, retry, callbacks);
         // Merge into single summary
-        summary = `${historyResult}\n\n---\n\n**Turn Context (split turn):**\n\n${turnPrefixResult}`;
+        summary = `${historyText}\n\n---\n\n**Turn Context (split turn):**\n\n${turnPrefixResult.text}`;
+        summaryUsage = historyUsage ? combineUsage(historyUsage, turnPrefixResult.usage) : turnPrefixResult.usage;
     }
     else {
         // Just generate history summary
-        summary = await generateSummary(messagesToSummarize, model, settings.reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env);
+        const result = await generateSummaryWithUsage(messagesToSummarize, model, settings.reserveTokens, apiKey, headers, signal, customInstructions, previousSummary, thinkingLevel, streamFn, env, retry, callbacks);
+        summary = result.text;
+        summaryUsage = result.usage;
     }
     // Compute file lists and append to summary
     const { readFiles, modifiedFiles } = computeFileLists(fileOps);
@@ -570,13 +606,14 @@ export async function compact(preparation, model, apiKey, headers, customInstruc
         summary,
         firstKeptEntryId,
         tokensBefore,
+        usage: summaryUsage,
         details: { readFiles, modifiedFiles },
     };
 }
 /**
  * Generate a summary for a turn prefix (when splitting a turn).
  */
-async function generateTurnPrefixSummary(messages, model, reserveTokens, apiKey, headers, env, signal, thinkingLevel, streamFn) {
+async function generateTurnPrefixSummary(messages, model, reserveTokens, apiKey, headers, env, signal, thinkingLevel, streamFn, retry, callbacks) {
     const maxTokens = Math.min(Math.floor(0.5 * reserveTokens), model.maxTokens > 0 ? model.maxTokens : Number.POSITIVE_INFINITY); // Smaller budget for turn prefix
     const llmMessages = convertToLlm(messages);
     const conversationText = serializeConversation(llmMessages);
@@ -588,13 +625,13 @@ async function generateTurnPrefixSummary(messages, model, reserveTokens, apiKey,
             timestamp: Date.now(),
         },
     ];
-    const response = await completeSummarization(model, { systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages }, createSummarizationOptions(model, maxTokens, apiKey, headers, env, signal, thinkingLevel), streamFn);
+    const response = await completeSummarization(model, { systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages }, createSummarizationOptions(model, maxTokens, apiKey, headers, env, signal, thinkingLevel), streamFn, retry, callbacks);
     if (response.stopReason === "error") {
         throw new Error(`Turn prefix summarization failed: ${response.errorMessage || "Unknown error"}`);
     }
-    return response.content
-        .filter((c) => c.type === "text")
-        .map((c) => c.text)
-        .join("\n");
+    return {
+        text: contentText(response.content),
+        usage: response.usage,
+    };
 }
 //# sourceMappingURL=compaction.js.map

package/node_modules/@earendil-works/pi-coding-agent/dist/core/compaction/utils.js

@@ -1,6 +1,7 @@
 /**
  * Shared utilities for compaction and branch summarization.
  */
+import { contentText } from "@earendil-works/pi-ai";
 export function createFileOps() {
     return {
         read: new Set(),
@@ -94,24 +95,15 @@ export function serializeConversation(messages) {
     const parts = [];
     for (const msg of messages) {
         if (msg.role === "user") {
-            const content = typeof msg.content === "string"
-                ? msg.content
-                : msg.content
-                    .filter((c) => c.type === "text")
-                    .map((c) => c.text)
-                    .join("");
+            const content = contentText(msg.content, "");
             if (content)
                 parts.push(`[User]: ${content}`);
         }
         else if (msg.role === "assistant") {
-            const textParts = [];
             const thinkingParts = [];
             const toolCalls = [];
             for (const block of msg.content) {
-                if (block.type === "text") {
-                    textParts.push(block.text);
-                }
-                else if (block.type === "thinking") {
+                if (block.type === "thinking") {
                     thinkingParts.push(block.thinking);
                 }
                 else if (block.type === "toolCall") {
@@ -125,18 +117,15 @@ export function serializeConversation(messages) {
             if (thinkingParts.length > 0) {
                 parts.push(`[Assistant thinking]: ${thinkingParts.join("\n")}`);
             }
-            if (textParts.length > 0) {
-                parts.push(`[Assistant]: ${textParts.join("\n")}`);
+            if (msg.content.some((block) => block.type === "text")) {
+                parts.push(`[Assistant]: ${contentText(msg.content)}`);
             }
             if (toolCalls.length > 0) {
                 parts.push(`[Assistant tool calls]: ${toolCalls.join("; ")}`);
             }
         }
         else if (msg.role === "toolResult") {
-            const content = msg.content
-                .filter((c) => c.type === "text")
-                .map((c) => c.text)
-                .join("");
+            const content = contentText(msg.content, "");
             if (content) {
                 parts.push(`[Tool result]: ${truncateForSummary(content, TOOL_RESULT_MAX_CHARS)}`);
             }

package/node_modules/@earendil-works/pi-coding-agent/dist/core/extensions/runner.js

@@ -657,6 +657,10 @@ export class ExtensionRunner {
                         currentEvent.isError = handlerResult.isError;
                         modified = true;
                     }
+                    if (handlerResult.usage !== undefined) {
+                        currentEvent.usage = handlerResult.usage;
+                        modified = true;
+                    }
                 }
                 catch (err) {
                     const message = err instanceof Error ? err.message : String(err);
@@ -677,6 +681,7 @@ export class ExtensionRunner {
             content: currentEvent.content,
             details: currentEvent.details,
             isError: currentEvent.isError,
+            usage: currentEvent.usage,
         };
     }
     async emitToolCall(event) {

package/node_modules/@earendil-works/pi-coding-agent/dist/core/model-resolver.js

@@ -40,6 +40,8 @@ export const defaultModelPerProvider = {
     "kimi-coding": "kimi-for-coding",
     "cloudflare-workers-ai": "@cf/moonshotai/kimi-k2.6",
     "cloudflare-ai-gateway": "workers-ai/@cf/moonshotai/kimi-k2.6",
+    "qwen-token-plan": "qwen3.7-max",
+    "qwen-token-plan-cn": "qwen3.7-max",
     xiaomi: "mimo-v2.5-pro",
     "xiaomi-token-plan-cn": "mimo-v2.5-pro",
     "xiaomi-token-plan-ams": "mimo-v2.5-pro",