feat: add conversation import — 5-format normalizer pipeline

[AliceLJY]

Summary

• Multi-format conversation normalization for importing chat history from 5 platforms
• Auto-detects format from content structure — zero manual config needed
• Framework-agnostic ingest via callback pattern — not coupled to any specific pipeline

Supported formats

Format	Detection	Notes
Claude Code JSONL	type: "human"/"assistant"	.claude/projects/ directory
Claude.ai JSON	chat_conversations wrapper	Official export format
ChatGPT JSON	mapping tree structure	conversations.json from data export
Slack JSON	ts + text fields	Channel export format
Plaintext	User:/Human:/Assistant:/AI: prefixes	Universal fallback

Changes

File	Change
packages/core/src/conversation-importer.ts	New — 5 normalizers + format detection + batch ingest
packages/core/src/index.ts	Export new functions and types
packages/core/test/conversation-importer.test.mjs	26 tests

Integration example

import { detectFormat, normalizeConversation, ingestNormalizedMessages } from "@anthropic/memory-core";

const content = readFileSync("conversations.json", "utf-8");
const format = detectFormat(content);
const messages = normalizeConversation(content, format);

await ingestNormalizedMessages(messages, "project:myapp", async (text, scope, role) => {
  await pipeline.ingest({ text, scope, category: "events", importance: 0.5, source: "conversation_import" });
  return "stored";
});

Test plan

• 26 unit tests pass (node --test packages/core/test/conversation-importer.test.mjs)
• All 5 formats + auto-detection + edge cases covered
• No changes to existing files except index.ts exports

🤖 Generated with Claude Code