"Preserve all opaque identifiers exactly as written (no shortening or reconstruction), " +

"including UUIDs, hashes, IDs, hostnames, IPs, ports, URLs, and file names.";

const HANDOFF_INSTRUCTIONS = [

"Generate a concise recovery briefing for a new LLM taking over this session.",

"The previous model hit a quota limit and you are providing the context for a smooth handoff.",

"",

"LEADER HIERARCHY REINFORCEMENT:",

"- Explicitly state that the new model is the LEADER (Orchestrator).",

"- Identify any active autonomous units (like AutoClaw) as SUBORDINATES.",

"- Instruct the new model to NOT perform the subordinate's task, but to supervise and provide strategic commands.",

"",

"MUST CAPTURE:",

"- Current high-level goal and project path.",

"- Status of the latest tool executions (especially AutoClaw/Subagents).",

"- Critical files currently being modified.",

"- Pending items and next intended steps.",

].join("\n");

export type CompactionSummarizationInstructions = {

identifierPolicy?: AgentCompactionIdentifierPolicy;

identifierInstructions?: string;

maxContextTokens: number;

maxHistoryShare?: number;

parts?: number;

mode?: "share" | "handoff";

}): {

messages: AgentMessage[];

droppedMessagesList: AgentMessage[];

keptTokens: number;

budgetTokens: number;

} {

const maxHistoryShare = params.maxHistoryShare ?? 0.5;

const isHandoff = params.mode === "handoff";

const defaultShare = isHandoff ? 0.2 : 0.5; // Stricter budget for handoff snapshots

const maxHistoryShare = params.maxHistoryShare ?? defaultShare;

const budgetTokens = Math.max(1, Math.floor(params.maxContextTokens * maxHistoryShare));

let keptMessages = params.messages;

const allDroppedMessages: AgentMessage[] = [];

};

}

/**

* Generates a concise handoff summary for model transitions, enforcing a 4000 token limit.

*/

export async function summarizeForHandoff(params: {

messages: AgentMessage[];

model: NonNullable<ExtensionContext["model"]>;

apiKey: string;

headers?: Record<string, string>;

signal: AbortSignal;

maxChunkTokens: number;

contextWindow: number;

customInstructions?: string;

summarizationInstructions?: CompactionSummarizationInstructions;

}): Promise<string> {

const custom = params.customInstructions?.trim();

const handoffInstructions = custom

? `${HANDOFF_INSTRUCTIONS}\n\n${custom}`

: HANDOFF_INSTRUCTIONS;

// Use a hard cap of 4000 tokens for the handoff summary as per plan

const handoffMaxTokens = 4000;

return summarizeWithFallback({

...params,

reserveTokens: SUMMARIZATION_OVERHEAD_TOKENS,

maxChunkTokens: Math.min(params.maxChunkTokens, handoffMaxTokens),

customInstructions: handoffInstructions,

});

}

export function resolveContextWindowTokens(model?: ExtensionContext["model"]): number {

const effective =

(model as { contextTokens?: number } | undefined)?.contextTokens ?? model?.contextWindow;