openfox 2.0.11 → 2.0.13

package/dist/{chunk-M53OBFMI.js → chunk-2VDLXCLO.js}

@@ -1,3 +1,17 @@
+import {
+  COMPACTION_PROMPT,
+  TurnMetrics,
+  buildBasePrompt,
+  buildSubAgentSystemPrompt,
+  buildTopLevelSystemPrompt,
+  consumeStreamGenerator,
+  createChatDoneEvent,
+  createMessageDoneEvent,
+  createMessageStartEvent,
+  createToolCallEvent,
+  createToolResultEvent,
+  streamLLMPure
+} from "./chunk-L7TDUIQY.js";
 import {
   startInspectProxy
 } from "./chunk-DL6ZILAF.js";
@@ -52,11 +66,7 @@ import {
   getGlobalConfigDir
 } from "./chunk-CQGTEGKL.js";
 import {
-  buildModelParams,
-  getBackendCapabilities,
-  getModelProfile,
-  modelSupportsVision,
-  streamWithSegments
+  modelSupportsVision
 } from "./chunk-Z4SWOUWC.js";
 import {
   logger
@@ -1753,655 +1763,6 @@ async function deleteAgent(configDir, agentId) {
   return { success: false };
 }
 
-// src/server/chat/prompts.ts
-function buildBasePrompt(workdir, customInstructions, skills, modelName) {
-  const instructionsSection = customInstructions ? `
-
-## CUSTOM INSTRUCTIONS
-
-${customInstructions}` : "";
-  const modelLine = modelName ? `
-Model: ${modelName}` : "";
-  return `You are OpenFox, an agentic assistant.
-
-Today's date is ${(/* @__PURE__ */ new Date()).toISOString().split("T")[0].replace(/-/g, "/")}
-
-## ENVIRONMENT
-Working directory: ${workdir}
-Platform: ${process.platform} (${process.arch})${modelLine}
-
-## CORE BEHAVIOR
-- Help the user complete any tasks safely and efficiently.
-- Do everything in your capacity to satisfy user requirements.
-- Read code before changing it.
-- Prefer precise, minimal changes.
-- Use available tools when needed.
-- Explain tradeoffs clearly when requirements are ambiguous.
-- Follow repository and project instructions exactly.
-
-## MODE CONTROL
-- OpenFox may append system-generated runtime control messages as USER-role messages wrapped in <system-reminder>...</system-reminder>.
-- These reminders are authoritative framework instructions from OpenFox.
-- Treat them as higher-priority operational constraints than normal user task wording for the current turn.
-- Do not describe them as "the user reminded me"; they are runtime mode/control metadata injected by OpenFox.
-
-## WORKFLOW
-- If the current runtime reminder says planning mode, focus on understanding, exploration, clarification, and criteria quality.
-- If the current runtime reminder says build mode, focus on implementation, verification, and completing approved criteria.
-- Respect tool and permission constraints enforced by the server even if the conversation suggests otherwise.
-
-## IMPORTANT GUARDRAILS
-- NEVER delete/git checkout an already modified file: that would result in a data loss.
-
-
-# Tone and style
-You should be concise, direct, and to the point. When you run a non-trivial bash command, you should explain what the command does and why you are running it, to make sure the user understands what you are doing (this is especially important when you are running a command that will make changes to the user's system).
-Remember that your output will be displayed on a command line interface. Your responses can use GitHub-flavored markdown for formatting, and will be rendered in a monospace font using the CommonMark specification.
-Output text to communicate with the user; all text you output outside of tool use is displayed to the user. Only use tools to complete tasks. Never use tools like Bash or code comments as means to communicate with the user during the session.
-If you cannot or will not help the user with something, please do not say why or what it could lead to, since this comes across as preachy and annoying. Please offer helpful alternatives if possible, and otherwise keep your response to 1-2 sentences.
-Only use emojis if the user explicitly requests it. Avoid using emojis in all communication unless asked.
-IMPORTANT: You should minimize output tokens as much as possible while maintaining helpfulness, quality, and accuracy. Only address the specific query or task at hand, avoiding tangential information unless absolutely critical for completing the request. If you can answer in 1-3 sentences or a short paragraph, please do.
-IMPORTANT: You should NOT answer with unnecessary preamble or postamble (such as explaining your code or summarizing your action), unless the user asks you to.
-IMPORTANT: Keep your responses short, since they will be displayed on a command line interface. You MUST answer concisely with fewer than 4 lines (not including tool use or code generation), unless user asks for detail. Answer the user's question directly, without elaboration, explanation, or details. One word answers are best. Avoid introductions, conclusions, and explanations. You MUST avoid text before/after your response, such as "The answer is <answer>.", "Here is the content of the file..." or "Based on the information provided, the answer is..." or "Here is what I will do next...". Here are some examples to demonstrate appropriate verbosity:
-<example>
-user: 2 + 2
-assistant: 4
-</example>
-
-<example>
-user: what is 2+2?
-assistant: 4
-</example>
-
-<example>
-user: is 11 a prime number?
-assistant: Yes
-</example>
-
-<example>
-user: what command should I run to list files in the current directory?
-assistant: ls
-</example>
-
-<example>
-user: what command should I run to watch files in the current directory?
-assistant: [use the ls tool to list the files in the current directory, then read docs/commands in the relevant file to find out how to watch files]
-npm run dev
-</example>
-
-<example>
-user: How many golf balls fit inside a jetta?
-assistant: 150000
-</example>
-
-<example>
-user: what files are in the directory src/?
-assistant: [runs ls and sees foo.c, bar.c, baz.c]
-user: which file contains the implementation of foo?
-assistant: src/foo.c
-</example>
-
-<example>
-user: write tests for new feature
-assistant: [uses grep and glob search tools to find where similar tests are defined, uses concurrent read file tool use blocks in one tool call to read relevant files at the same time, uses edit file tool to write new tests]
-</example>
-
-# Proactiveness
-You are allowed to be proactive, but only when the user asks you to do something. You should strive to strike a balance between:
-1. Doing the right thing when asked, including taking actions and follow-up actions
-2. Not surprising the user with actions you take without asking
-For example, if the user asks you how to approach something, you should do your best to answer their question first, and not immediately jump into taking actions.
-3. Do not add additional code explanation summary unless requested by the user. After working on a file, just stop, rather than providing an explanation of what you did.
-
-# Following conventions
-When making changes to files, first understand the file's code conventions. Mimic code style, use existing libraries and utilities, and follow existing patterns.
-- NEVER assume that a given library is available, even if it is well known. Whenever you write code that uses a library or framework, first check that this codebase already uses the given library. For example, you might look at neighboring files, or check the package.json (or cargo.toml, and so on depending on the language).
-- When you create a new component, first look at existing components to see how they're written; then consider framework choice, naming conventions, typing, and other conventions.
-- When you edit a piece of code, first look at the code's surrounding context (especially its imports) to understand the code's choice of frameworks and libraries. Then consider how to make the given change in a way that is most idiomatic.
-- Always follow security best practices. Never introduce code that exposes or logs secrets and keys. Never commit secrets or keys to the repository.
-
-# Code style
-- IMPORTANT: DO NOT ADD ***ANY*** COMMENTS unless asked
-
-# Doing tasks
-The user will primarily request you perform software engineering tasks. This includes solving bugs, adding new functionality, refactoring code, explaining code, and more. For these tasks the following steps are recommended:
-- Use the available search tools to understand the codebase and the user's query. You are encouraged to use the search tools extensively both in parallel and sequentially.
-- Implement the solution using all tools available to you
-- Verify the solution if possible with tests. NEVER assume specific test framework or test script. Check the README or search codebase to determine the testing approach.
-- VERY IMPORTANT: When you have completed a task, you MUST run the lint and typecheck commands (e.g. npm run lint, npm run typecheck, ruff, etc.) with Bash if they were provided to you to ensure your code is correct. If you are unable to find the correct command, ask the user for the command to run and if they supply it, proactively suggest writing it to AGENTS.md so that you will know to run it next time.
-NEVER commit changes unless the user explicitly asks you to. It is VERY IMPORTANT to only commit when explicitly asked, otherwise the user will feel that you are being too proactive.
-
-- Tool results and user messages may include <system-reminder> tags. <system-reminder> tags contain useful information and reminders. They are NOT part of the user's provided input or the tool result.
-
-# Tool usage policy
-- When doing file search, prefer to use the Task tool in order to reduce context usage.
-- You have the capability to call multiple tools in a single response. When multiple independent pieces of information are requested, batch your tool calls together for optimal performance. When making multiple bash tool calls, you MUST send a single message with multiple tools calls to run the calls in parallel. For example, if you need to run "git status" and "git diff", send a single message with two tool calls to run the calls in parallel.
-
-You MUST answer concisely with fewer than 4 lines of text (not including tool use or code generation), unless user asks for detail.
-
-${instructionsSection}
-${buildSkillsSection(skills)}
-`;
-}
-function buildSkillsSection(skills) {
-  if (!skills || skills.length === 0) return "";
-  const listing = skills.map((s, i) => `${i + 1}. **${s.id}** - ${s.description}`).join("\n");
-  return `
-## AVAILABLE SKILLS
-
-You can load specialized knowledge using the load_skill tool. Only load a skill when you need its instructions for the current task.
-
-${listing}
-
-To load a skill, call load_skill with the skill ID. The skill's detailed instructions will be returned as a tool result.
-`;
-}
-function buildSubAgentsSection(subAgentDefs) {
-  if (subAgentDefs.length === 0) return "";
-  const listing = subAgentDefs.map((agent, i) => {
-    const tools = (agent.metadata.allowedTools || []).join(", ");
-    return `${i + 1}. **${agent.metadata.id}** - ${agent.metadata.description}
-   - Has access to: ${tools}`;
-  }).join("\n\n");
-  return `
-## AVAILABLE SUB-AGENTS
-
-You can call specialized sub-agents for specific tasks using the call_sub_agent tool:
-
-${listing}
-
-To call a sub-agent, use the call_sub_agent tool with:
-- subAgentType: The ID of the sub-agent
-- prompt: Clear description of what you need
-`;
-}
-function buildTopLevelSystemPrompt(workdir, customInstructions, skills, subAgentDefs, modelName) {
-  const base = buildBasePrompt(workdir, customInstructions, skills, modelName);
-  const subAgents = subAgentDefs ? buildSubAgentsSection(subAgentDefs) : "";
-  return base + subAgents;
-}
-function buildSubAgentSystemPrompt(workdir, agentDef, skills, modelName) {
-  const base = buildBasePrompt(workdir, void 0, skills, modelName);
-  return base + "\n\n" + agentDef.prompt;
-}
-function buildToolPermissionsSection(allowedTools, isSubAgent) {
-  if (!allowedTools || allowedTools.length === 0) {
-    return "\n\n## AVAILABLE TOOLS\n\nYou have no tools available.";
-  }
-  const toolPermissions = /* @__PURE__ */ new Map();
-  const baseTools = [];
-  for (const entry of allowedTools) {
-    const colonIdx = entry.indexOf(":");
-    if (colonIdx === -1) {
-      if (entry !== "return_value" || isSubAgent) {
-        baseTools.push(entry);
-      }
-    } else {
-      const toolName = entry.slice(0, colonIdx);
-      const actionsStr = entry.slice(colonIdx + 1);
-      const actions = actionsStr.split(",").filter(Boolean);
-      const existing = toolPermissions.get(toolName) || [];
-      toolPermissions.set(toolName, [...existing, ...actions]);
-    }
-  }
-  if (isSubAgent && !baseTools.includes("return_value") && !toolPermissions.has("return_value")) {
-    baseTools.push("return_value");
-  }
-  const parts = [];
-  for (const [toolName, actions] of toolPermissions) {
-    parts.push(`${toolName}: ${actions.join(", ")}`);
-  }
-  for (const tool of baseTools) {
-    if (!toolPermissions.has(tool)) {
-      parts.push(tool);
-    }
-  }
-  const toolsList = parts.join(", ");
-  return `
-
-## AVAILABLE TOOLS
-
-You have access to these tools: ${toolsList}`;
-}
-function buildAgentReminder(agentDef) {
-  const toolPermissions = buildToolPermissionsSection(agentDef.metadata.allowedTools, agentDef.metadata.subagent);
-  return `<system-reminder>
-${agentDef.prompt}${toolPermissions}
-</system-reminder>`;
-}
-function buildAgentSmallReminder(name) {
-  return `<system-reminder>
-Reminder: you are in '${name}' mode.
-</system-reminder>`;
-}
-var WORKFLOW_KICKOFF_PROMPT = (criteriaCount) => `Implement the task and make sure you fulfil the ${criteriaCount} criteria.`;
-var COMPACTION_PROMPT = `You are a helpful AI assistant tasked with summarizing conversations for continuation.
-
-Summarize the conversation history concisely, preserving:
-1. What was done and what is currently being worked on
-2. All file modifications made (file paths and what changed)
-3. All errors encountered and how they were resolved
-4. Current progress on each task
-5. Important technical decisions and WHY they were made
-6. Requirements that should persist
-7. Next steps or pending actions that should be continued after compaction
-8. The user's current question, prompt, or active request
-
-Do not respond to any questions in the conversation, only output the summary.
-Be thorough but concise. Output as a structured summary.`;
-
-// src/server/chat/auto-patterns.ts
-function matchRetryPatterns(content, thinking, patterns) {
-  const matches = [];
-  for (const config of patterns) {
-    if (!config.active) continue;
-    let regex;
-    try {
-      regex = new RegExp(config.pattern);
-    } catch {
-      continue;
-    }
-    const testContent = config.field === "thinking" ? false : regex.test(content);
-    const testThinking = config.field === "content" ? false : thinking !== void 0 && regex.test(thinking);
-    if (testContent) {
-      matches.push({ pattern: config.pattern, field: config.field, matchedContent: content });
-    }
-    if (testThinking) {
-      matches.push({ pattern: config.pattern, field: config.field, matchedContent: thinking });
-    }
-  }
-  return matches;
-}
-
-// src/server/chat/stream-utils.ts
-function buildStreamRequestObject(params) {
-  const { messages, tools, toolChoice, reasoningEffort, signal, modelSettings } = params;
-  return {
-    messages,
-    ...tools && { tools },
-    ...toolChoice && { toolChoice },
-    ...reasoningEffort && { reasoningEffort },
-    ...signal && { signal },
-    ...modelSettings && { modelSettings }
-  };
-}
-function buildStreamRequest(client, options) {
-  return streamWithSegments(client, buildStreamRequestObject(options));
-}
-
-// src/server/chat/stats.ts
-var roundTo1 = (n) => Math.round(n * 10) / 10;
-function computeAggregatedStats(input) {
-  const {
-    identity,
-    mode,
-    totalPrefillTokens,
-    totalPrefillIncrement,
-    totalGenTokens,
-    totalPrefillTime,
-    totalGenTime,
-    totalToolTime,
-    totalTime,
-    llmCalls
-  } = input;
-  const prefillSource = totalPrefillIncrement ?? totalPrefillTokens;
-  return {
-    ...identity,
-    mode,
-    totalTime,
-    toolTime: totalToolTime,
-    prefillTokens: totalPrefillTokens,
-    ...totalPrefillIncrement !== void 0 && { prefTokenIncrement: totalPrefillIncrement },
-    prefillSpeed: totalPrefillTime > 0 ? roundTo1(prefillSource / totalPrefillTime) : 0,
-    generationTokens: totalGenTokens,
-    generationSpeed: totalGenTime > 0 ? roundTo1(totalGenTokens / totalGenTime) : 0,
-    ...llmCalls ? { llmCalls } : {}
-  };
-}
-
-// src/server/chat/stream-pure.ts
-function createEmptyStreamResult(aborted, modelParams, patternMatch) {
-  return {
-    content: "",
-    toolCalls: [],
-    segments: [],
-    usage: { promptTokens: 0, completionTokens: 0 },
-    timing: { ttft: 0, completionTime: 0, tps: 0, prefillTps: 0 },
-    aborted,
-    modelParams,
-    finishReason: "stop",
-    ...patternMatch ? { patternMatch } : {}
-  };
-}
-async function* streamLLMPure(options) {
-  const { messageId, systemPrompt, llmClient, messages, tools, toolChoice, signal, reasoningEffort, retryPatterns } = options;
-  const llmMessages = [{ role: "system", content: systemPrompt }, ...messages];
-  const profile = getModelProfile(llmClient.getModel());
-  const backend = getBackendCapabilities(llmClient.getBackend());
-  const userTemp = options.modelSettings?.temperature;
-  const userTopP = options.modelSettings?.topP;
-  const userTopK = options.modelSettings?.topK;
-  const userMaxTokens = options.modelSettings?.maxTokens;
-  const temperature = userTemp ?? profile.temperature;
-  const maxTokens = userMaxTokens ?? profile.defaultMaxTokens;
-  const topP = userTopP ?? profile.topP;
-  const topK = userTopK ?? (backend.supportsTopK ? profile.topK : void 0);
-  const modelParams = buildModelParams({ temperature, topP, topK, maxTokens });
-  logger.debug("LLM request settings", {
-    model: llmClient.getModel(),
-    profile: profile.name,
-    temperature,
-    maxTokens,
-    topP,
-    topK,
-    userConfigured: {
-      temperature: userTemp !== void 0 ? `user:${userTemp}` : "default",
-      topP: userTopP !== void 0 ? `user:${userTopP}` : "default",
-      topK: userTopK !== void 0 ? `user:${userTopK}` : "default",
-      maxTokens: userMaxTokens !== void 0 ? `user:${userMaxTokens}` : "default"
-    }
-  });
-  const patternAbortController = new AbortController();
-  const combinedSignal = signal ? AbortSignal.any([signal, patternAbortController.signal]) : patternAbortController.signal;
-  const stream = buildStreamRequest(llmClient, {
-    messages: llmMessages,
-    tools,
-    toolChoice,
-    reasoningEffort,
-    signal: combinedSignal,
-    modelSettings: options.modelSettings
-  });
-  const seenToolIndices = /* @__PURE__ */ new Set();
-  const toolNames = /* @__PURE__ */ new Map();
-  const toolIds = /* @__PURE__ */ new Map();
-  const returnValueArgs = /* @__PURE__ */ new Map();
-  const toolArgs = /* @__PURE__ */ new Map();
-  let result = null;
-  let aborted = false;
-  let accumulatedContent = "";
-  let accumulatedThinking = "";
-  let patternMatch;
-  const activePatterns = retryPatterns?.filter((p) => p.active) ?? [];
-  try {
-    while (true) {
-      if (signal?.aborted) {
-        aborted = true;
-        break;
-      }
-      const { value, done } = await stream.next();
-      if (done) {
-        result = value;
-        break;
-      }
-      switch (value.type) {
-        case "text_delta":
-          accumulatedContent += value.content;
-          yield {
-            type: "message.delta",
-            data: { messageId, content: value.content }
-          };
-          break;
-        case "thinking_delta":
-          accumulatedThinking += value.content;
-          yield {
-            type: "message.thinking",
-            data: { messageId, content: value.content }
-          };
-          break;
-        case "tool_call_delta": {
-          if (value.name) {
-            const existingName = toolNames.get(value.index) ?? "";
-            toolNames.set(value.index, existingName + value.name);
-          }
-          if (value.id) {
-            toolIds.set(value.index, value.id);
-          }
-          if (value.arguments) {
-            const existingArgs = toolArgs.get(value.index) ?? "";
-            toolArgs.set(value.index, existingArgs + value.arguments);
-          }
-          const fullName = toolNames.get(value.index);
-          if (!seenToolIndices.has(value.index) && fullName) {
-            seenToolIndices.add(value.index);
-            const accumulatedArgs = toolArgs.get(value.index);
-            yield {
-              type: "tool.preparing",
-              data: {
-                messageId,
-                index: value.index,
-                name: fullName,
-                ...accumulatedArgs ? { arguments: accumulatedArgs } : {}
-              }
-            };
-          } else if (seenToolIndices.has(value.index) && value.arguments) {
-            const name = toolNames.get(value.index);
-            if (name === "run_command" || name === "return_value") {
-              const accumulatedArgs = toolArgs.get(value.index);
-              if (accumulatedArgs) {
-                yield {
-                  type: "tool.preparing",
-                  data: { messageId, index: value.index, name, arguments: accumulatedArgs }
-                };
-              }
-            }
-          }
-          if (fullName === "return_value" && value.arguments) {
-            const toolCallId = toolIds.get(value.index);
-            if (toolCallId) {
-              const prevRaw = returnValueArgs.get(value.index) ?? "";
-              const newRaw = prevRaw + value.arguments;
-              returnValueArgs.set(value.index, newRaw);
-              const contentStart = newRaw.indexOf('"content"');
-              if (contentStart >= 0) {
-                const colonPos = newRaw.indexOf(":", contentStart + 9);
-                if (colonPos >= 0) {
-                  const valueStart = newRaw.indexOf('"', colonPos + 1);
-                  if (valueStart >= 0) {
-                    const prevContent = prevRaw.length > valueStart + 1 ? prevRaw.slice(valueStart + 1).replace(/"\s*\}\s*$/, "") : "";
-                    const currentContent = newRaw.slice(valueStart + 1).replace(/"\s*\}\s*$/, "");
-                    const delta = currentContent.slice(prevContent.length);
-                    if (delta) {
-                      const unescaped = delta.replace(/\\n/g, "\n").replace(/\\t/g, "	").replace(/\\"/g, '"').replace(/\\\\/g, "\\");
-                      yield {
-                        type: "tool.output",
-                        data: { messageId, toolCallId, stream: "stdout", content: unescaped }
-                      };
-                    }
-                  }
-                }
-              }
-            }
-          }
-          break;
-        }
-        case "error":
-          yield {
-            type: "chat.error",
-            data: { error: value.error, recoverable: true }
-          };
-          break;
-      }
-      if (activePatterns.length > 0 && (accumulatedContent || accumulatedThinking)) {
-        const matches = matchRetryPatterns(accumulatedContent, accumulatedThinking || void 0, activePatterns);
-        if (matches.length > 0) {
-          patternMatch = matches[0];
-          patternAbortController.abort();
-          break;
-        }
-      }
-    }
-  } catch (error) {
-    if (error instanceof Error && error.message === "Aborted") {
-      aborted = true;
-    } else {
-      throw error;
-    }
-  }
-  if (patternMatch) {
-    return createEmptyStreamResult(false, modelParams, patternMatch);
-  }
-  if (!result) {
-    return createEmptyStreamResult(aborted, modelParams);
-  }
-  const baseResult = {
-    content: result.content,
-    toolCalls: result.toolCalls,
-    segments: result.segments,
-    usage: {
-      promptTokens: result.response.usage.promptTokens,
-      completionTokens: result.response.usage.completionTokens
-    },
-    timing: result.timing,
-    aborted,
-    modelParams,
-    finishReason: result.response.finishReason
-  };
-  if (result.thinkingContent) {
-    baseResult.thinkingContent = result.thinkingContent;
-  }
-  return baseResult;
-}
-var TurnMetrics = class {
-  startTime;
-  totalPrefillTokens = 0;
-  totalPrefillIncrement = 0;
-  totalPrefillTime = 0;
-  // seconds
-  totalGenTokens = 0;
-  totalGenTime = 0;
-  // seconds
-  totalToolTime = 0;
-  // seconds
-  llmCalls = [];
-  modelParams = {};
-  constructor() {
-    this.startTime = performance.now();
-  }
-  /** Add metrics from an LLM call.
-   * @param previousContextTokens - context size BEFORE this LLM call (for computing the non-cached increment)
-   */
-  addLLMCall(timing, promptTokens, completionTokens, previousContextTokens, modelParams) {
-    const callIndex = this.llmCalls.length + 1;
-    this.totalPrefillTokens += promptTokens;
-    this.totalPrefillTime += timing.ttft;
-    this.totalGenTokens += completionTokens;
-    this.totalGenTime += timing.completionTime;
-    if (modelParams) {
-      this.modelParams = modelParams;
-    }
-    const prefTokenIncrement = previousContextTokens !== void 0 ? Math.max(0, promptTokens - previousContextTokens) : void 0;
-    if (prefTokenIncrement !== void 0) {
-      this.totalPrefillIncrement += prefTokenIncrement;
-    }
-    const prefillSource = prefTokenIncrement ?? promptTokens;
-    this.llmCalls = [
-      ...this.llmCalls,
-      {
-        callIndex,
-        promptTokens,
-        completionTokens,
-        ...prefTokenIncrement !== void 0 && { prefTokenIncrement },
-        ttft: timing.ttft,
-        completionTime: timing.completionTime,
-        prefillSpeed: timing.ttft > 0 ? Math.round(prefillSource / timing.ttft * 10) / 10 : 0,
-        generationSpeed: timing.completionTime > 0 ? Math.round(completionTokens / timing.completionTime * 10) / 10 : 0,
-        totalTime: Math.round((timing.ttft + timing.completionTime) * 10) / 10,
-        timestamp: (/* @__PURE__ */ new Date()).toISOString(),
-        ...this.modelParams.temperature !== void 0 && { temperature: this.modelParams.temperature },
-        ...this.modelParams.topP !== void 0 && { topP: this.modelParams.topP },
-        ...this.modelParams.topK !== void 0 && { topK: this.modelParams.topK },
-        ...this.modelParams.maxTokens !== void 0 && { maxTokens: this.modelParams.maxTokens }
-      }
-    ];
-  }
-  /** Set model parameters for tracking */
-  setModelParams(params) {
-    this.modelParams = params;
-  }
-  /** Add tool execution time (in milliseconds) */
-  addToolTime(durationMs) {
-    this.totalToolTime += durationMs / 1e3;
-  }
-  /** Build final stats object */
-  buildStats(identity, mode) {
-    return computeAggregatedStats({
-      identity,
-      mode,
-      totalPrefillTokens: this.totalPrefillTokens,
-      ...this.totalPrefillIncrement > 0 && { totalPrefillIncrement: this.totalPrefillIncrement },
-      totalGenTokens: this.totalGenTokens,
-      totalPrefillTime: this.totalPrefillTime,
-      totalGenTime: this.totalGenTime,
-      totalToolTime: this.totalToolTime,
-      totalTime: (performance.now() - this.startTime) / 1e3,
-      llmCalls: this.llmCalls.map((call) => ({
-        ...identity,
-        ...call
-      }))
-    });
-  }
-};
-function createMessageStartEvent(messageId, role, content, options) {
-  return {
-    type: "message.start",
-    data: {
-      messageId,
-      role,
-      ...content !== void 0 && { content },
-      ...options?.contextWindowId && { contextWindowId: options.contextWindowId },
-      ...options?.subAgentId && { subAgentId: options.subAgentId },
-      ...options?.subAgentType && { subAgentType: options.subAgentType },
-      ...options?.isSystemGenerated && { isSystemGenerated: options.isSystemGenerated },
-      ...options?.messageKind && { messageKind: options.messageKind },
-      ...options?.metadata && { metadata: options.metadata }
-    }
-  };
-}
-function createMessageDoneEvent(messageId, options) {
-  return {
-    type: "message.done",
-    data: {
-      messageId,
-      ...options?.stats && { stats: options.stats },
-      ...options?.segments && { segments: options.segments },
-      ...options?.partial && { partial: options.partial }
-    }
-  };
-}
-function createToolCallEvent(messageId, toolCall) {
-  return {
-    type: "tool.call",
-    data: { messageId, toolCall }
-  };
-}
-function createToolResultEvent(messageId, toolCallId, result) {
-  return {
-    type: "tool.result",
-    data: { messageId, toolCallId, result }
-  };
-}
-function createChatDoneEvent(messageId, reason, stats, agentType) {
-  return {
-    type: "chat.done",
-    data: {
-      messageId,
-      reason,
-      ...stats && { stats },
-      ...agentType && { agentType }
-    }
-  };
-}
-async function consumeStreamGenerator(gen, onEvent) {
-  let result;
-  while (true) {
-    result = await gen.next();
-    if (result.done) {
-      return result.value;
-    }
-    onEvent(result.value);
-  }
-}
-
 // src/server/context/instructions.ts
 import { readFile as readFile6, access as access3 } from "fs/promises";
 import { join as join5, dirname as dirname3 } from "path";
@@ -2843,7 +2204,7 @@ async function runTopLevelAgentLoop(config, turnMetrics) {
   let returnValueResult;
   let currentMaxTokensOverride;
   let lastPatternMatch;
-  let compacting = false;
+  let compacting = config.initialCompacting ?? false;
   let returnValueNudgeCount = 0;
   for (; ; ) {
     if (signal?.aborted) throw new Error("Aborted");
@@ -2966,18 +2327,9 @@ ${CONTINUE_PROMPT}` : CONTINUE_PROMPT;
     sessionManager.setCurrentContextSize(sessionId, result.usage.promptTokens);
     if (!compacting) {
       const contextState = sessionManager.getContextState(sessionId);
-      const { shouldCompact } = await import("./compactor-5VQNEYQX.js");
+      const { shouldCompact, appendCompactionPrompt } = await import("./compactor-R7GRFBOU.js");
       if (shouldCompact(contextState.currentTokens, contextState.maxTokens, runtimeConfig.context.compactionThreshold)) {
-        const compactPromptMsgId = crypto.randomUUID();
-        append(
-          createMessageStartEvent(compactPromptMsgId, "user", COMPACTION_PROMPT, {
-            ...currentWindowMessageOptions ?? {},
-            isSystemGenerated: true,
-            messageKind: "auto-prompt",
-            metadata: { type: "compaction", name: "Compaction", color: "#64748b" }
-          })
-        );
-        append({ type: "message.done", data: { messageId: compactPromptMsgId } });
+        appendCompactionPrompt(sessionId, append);
         compacting = true;
         continue;
       }
@@ -3126,6 +2478,7 @@ ${COMPACTION_PROMPT}`,
         });
         logger.warn("Compaction produced empty summary, continuing", { sessionId });
         compacting = false;
+        if (config.initialCompacting) break;
         continue;
       }
       const closedWindowId = getCurrentContextWindowId(sessionId) ?? "";
@@ -3149,6 +2502,7 @@ ${COMPACTION_PROMPT}`,
       append(createChatDoneEvent(assistantMsgId, "complete"));
       config.injectAgentReminder?.();
       compacting = false;
+      if (config.initialCompacting) break;
       continue;
     }
     if (config.requireReturnValue && !returnValueContent) {
@@ -3777,7 +3131,7 @@ var callSubAgentTool = {
       };
     }
     try {
-      const { getToolRegistryForAgent: getToolRegistryForAgent2 } = await import("./tools-UT5WFO7R.js");
+      const { getToolRegistryForAgent: getToolRegistryForAgent2 } = await import("./tools-THXBQJ7A.js");
       const toolRegistry = getToolRegistryForAgent2(agentDef);
       const turnMetrics = new TurnMetrics();
       const result = await executeSubAgent({
@@ -4801,20 +4155,6 @@ export {
   agentExists,
   saveAgent,
   deleteAgent,
-  buildBasePrompt,
-  buildTopLevelSystemPrompt,
-  buildAgentReminder,
-  buildAgentSmallReminder,
-  WORKFLOW_KICKOFF_PROMPT,
-  COMPACTION_PROMPT,
-  streamLLMPure,
-  TurnMetrics,
-  createMessageStartEvent,
-  createMessageDoneEvent,
-  createToolCallEvent,
-  createToolResultEvent,
-  createChatDoneEvent,
-  consumeStreamGenerator,
   getAllInstructions,
   loadDefaultSkills,
   loadUserSkills,
@@ -4846,4 +4186,4 @@ export {
   getToolRegistryForAgent,
   createToolRegistry
 };
-//# sourceMappingURL=chunk-M53OBFMI.js.map
+//# sourceMappingURL=chunk-2VDLXCLO.js.map