@pi-vault/pi-dcp 0.1.0

package/src/utils/message-ids.ts

@@ -0,0 +1,45 @@
+export function formatMessageRef(index: number): string {
+  return `m${String(index).padStart(4, "0")}`;
+}
+
+export function formatBlockRef(blockId: number): string {
+  return `b${blockId}`;
+}
+
+export function parseMessageRef(ref: string): number | undefined {
+  const match = /^m(\d{4})$/.exec(ref);
+  if (!match) return undefined;
+  return parseInt(match[1], 10);
+}
+
+export function parseBlockRef(ref: string): number | undefined {
+  const match = /^b(\d+)$/.exec(ref);
+  if (!match) return undefined;
+  const n = parseInt(match[1], 10);
+  if (n <= 0) return undefined;
+  return n;
+}
+
+export type ParsedBoundaryId =
+  | { type: "message"; index: number }
+  | { type: "block"; blockId: number };
+
+export function parseBoundaryId(id: string): ParsedBoundaryId | undefined {
+  const msgIndex = parseMessageRef(id);
+  if (msgIndex !== undefined) return { type: "message", index: msgIndex };
+
+  const blockId = parseBlockRef(id);
+  if (blockId !== undefined) return { type: "block", blockId };
+
+  return undefined;
+}
+
+export function formatMessageIdTag(
+  ref: string,
+  attrs?: { priority?: number },
+): string {
+  if (attrs?.priority !== undefined) {
+    return `<dcp-message-id priority="${attrs.priority}">${ref}</dcp-message-id>`;
+  }
+  return `<dcp-message-id>${ref}</dcp-message-id>`;
+}

package/src/utils/tokens.ts

@@ -0,0 +1,54 @@
+/**
+ * Token counting using character-based estimation.
+ *
+ * Uses length/4 as a rough approximation. Pi's built-in ctx.getContextUsage()
+ * provides accurate context-level token counts for threshold decisions. These
+ * per-message estimates are for relative comparisons (compression savings,
+ * priority ranking).
+ */
+export function countTokens(text: string): number {
+  if (text.length === 0) return 0;
+  return Math.max(1, Math.round(text.length / 4));
+}
+
+export function countTokensBatch(texts: string[]): number {
+  if (texts.length === 0) return 0;
+  return countTokens(texts.join(" "));
+}
+
+/**
+ * Extract all text content from a message-shaped object for token counting.
+ * Handles UserMessage (string | TextContent[]), AssistantMessage (TextContent +
+ * ToolCallContent), and ToolResultMessage.
+ */
+export function extractMessageText(message: {
+  role: string;
+  content?: unknown;
+}): string {
+  const content = message.content;
+  if (!content) return "";
+  if (typeof content === "string") return content;
+  if (!Array.isArray(content)) return "";
+
+  const parts: string[] = [];
+  for (const part of content) {
+    if (typeof part !== "object" || part === null) continue;
+    const p = part as Record<string, unknown>;
+    if (p.type === "text" && typeof p.text === "string") {
+      parts.push(p.text);
+    } else if (p.type === "toolCall") {
+      if (typeof p.name === "string") parts.push(p.name);
+      if (p.arguments && typeof p.arguments === "object") {
+        parts.push(JSON.stringify(p.arguments));
+      }
+    }
+  }
+  return parts.join(" ");
+}
+
+export function countMessageTokens(message: {
+  role: string;
+  content?: unknown;
+}): number {
+  return countTokens(extractMessageText(message));
+}