@rubytech/create-realagent 1.0.826 → 1.0.829

package/payload/platform/plugins/whatsapp-import/PLUGIN.md

@@ -1,46 +0,0 @@
----
-name: whatsapp-import
-description: "Import a WhatsApp `_chat.txt` export (Conversation + Messages with chronological NEXT chain + analysis-derived insights) into the {{productName}} Neo4j graph as a two-phase contract: Phase 1 (`whatsapp-import` skill) lands raw shape via a deterministic Bash entry; Phase 2 (`whatsapp-import-enrich` skill) is operator-driven semantic resolution over the loaded conversation. Skill-only plugin owned by the database-operator specialist. Opt-in per brand — not enabled by default. Distinct from the live `whatsapp` plugin (Baileys QR pairing + in-memory store)."
-tools: []
-always: false
-embed: false
-specialist: database-operator
-metadata: {"platform":{"optional":true,"pluginKey":"whatsapp-import"}}
----
-
-# WhatsApp Import
-
-Ingests a WhatsApp "Export Chat" archive (the `_chat.txt` file plus media attachments) into the {{productName}} Neo4j graph. Skill-only plugin — no MCP server, no admin tools added. The skill runs under the `database-operator` specialist, which owns external-archive ingestion and ad-hoc graph operations.
-
-## Two-phase contract (earlier platform fixes + earlier platform fixes)
-
-The plugin ships two skills that pair as a contract: deterministic load first, then operator-driven semantic enrichment. Splitting them removes the orchestration loop a single bundled phase imposed (the 7h orchestration), and lets re-imports reuse the load phase without redoing the semantic walk.
-
-| Phase | Skill | What it does | Trigger phrase |
-|-------|-------|--------------|----------------|
-| 1 — load | [`whatsapp-import`](skills/whatsapp-import/SKILL.md) | Parses `_chat.txt`, writes `:Conversation:WhatsAppConversation` + `:Message:WhatsAppMessage` with NEXT chain, auto-creates one `:Person {participantStatus:'auto-created'}` per distinct senderName, lands `:Observation {observationStatus:'auto-extracted'}` rows from the chunked Haiku insight pass. Single Bash entry; no MCP envelope between steps. | Operator drops a `_chat.txt` file or its containing export folder into chat. |
-| 2 — enrich | [`whatsapp-import-enrich`](skills/whatsapp-import-enrich/SKILL.md) | Walks `participantStatus='auto-created'` and `observationStatus='auto-extracted'` rows scoped to a chosen conversation, surfaces evidence per row, writes operator-confirmed wiring: `apoc.refactor.mergeNodes` for participant promotion/merge, `:MENTIONS`/`:RELATED_TO` edges with `evidenceSnippet`, `:Task` via `task-create`, `:Preference` via `memory-write`. Idempotent — re-running surfaces only items still in `auto-created`/`auto-extracted`. | Operator asks to "enrich the X chat", "promote auto-created participants from Y", "wire observations from yesterday's import". |
-
-Phase 2 refuses to run against a Conversation whose `c.lastImportedAt` is null. Phase 1 always precedes Phase 2.
-
-## When this applies
-
-The admin agent delegates to `database-operator` when the operator drops a `_chat.txt` (or its containing folder) into chat (→ Phase 1) or names enrichment of an already-loaded conversation (→ Phase 2). For Phase 1 the specialist runs the skill's archive-owner confirmation flow before any line is written, then invokes the deterministic Bash entry (`bin/whatsapp-ingest.sh`) once: parse, archive-write (via `memoryArchiveWrite` in-process), and Haiku insight all run in one Node process — no MCP envelope between steps. For Phase 2 the specialist runs `whatsapp-import-enrich`'s bulk preview, asks the operator to confirm scope, then walks the rows with operator confirmation gates.
-
-## Accepted export shapes
-
-WhatsApp's "Export Chat" emits `[DD/MM/YYYY, HH:MM:SS]` prefixes by default in modern releases and `[DD/MM/YY, HH:MM:SS]` in legacy exports. The parser accepts both year shapes (2-digit mapped `2000+yy`, 4-digit passed through) within a single file and auto-detects DD/MM vs MM/DD ordering from the first matched line — see [skills/whatsapp-import/references/export-parse.md](skills/whatsapp-import/references/export-parse.md) for the full grammar contract.
-
-## What this is not
-
-- **Not** the live `whatsapp` plugin. That plugin (Baileys QR pairing) holds messages in an in-memory store cleared on restart. This plugin imports historical exports into Neo4j as persistent graph nodes. Names are deliberately distinct (`whatsapp-import` vs `whatsapp`) so the boundary is mechanical.
-- **Not** a media-transcription pipeline. Voice notes (`.opus`), photos, PDFs are skipped at parse with a counter logged. Transcription / OCR is a separate plugin decision.
-- **Not** a generic chat-export importer. Each messaging channel (Telegram, iMessage, Signal, SMS dump) ships as its own opt-in plugin under the same shape when requested.
-
-## Relationship to other plugins
-
-- **memory** — Phase 1's underlying write surface, imported in-process by `bin/ingest.mjs` (`memoryArchiveWrite` for bulk Conversation+Messages; direct Cypher `:Observation` writes for the insight pass). Phase 2's enrich skill writes `:Preference` nodes via `mcp__memory__memory-write` and uses `mcp__memory__memory-search` for entity disambiguation. All writes carry `source='whatsapp'` + `createdByAgent='whatsapp-import'` (Phase 1) or `createdByAgent='whatsapp-import-enrich'` (Phase 2) provenance. The legacy `mcp__memory__whatsapp-export-parse` / `whatsapp-export-insight-write` MCP tools and the direct `memory-archive-write` MCP path with `archiveType=whatsapp-export` are blocked at the harness — the Bash entry is the only supported invocation surface for Phase 1.
-- **tasks** — Phase 2's `:Task` writes go through `mcp__tasks__task-create` with `affects=$conversationElementId`. Database-operator's frontmatter `tools:` includes `mcp__tasks__task-create` for this path.
-- **contacts** — Phase 2's mint-new-Person path (an auto-created participant the operator wants to land as a fresh contact) goes through `mcp__contacts__contact-create`.
-- **database-operator specialist** — owns execution for both phases. See [admin/IDENTITY.md](../../../platform/templates/agents/admin/IDENTITY.md) delegation clause and [database-operator.md](../../../platform/templates/specialists/agents/database-operator.md) per-source archive list (which now names both phases under the WhatsApp entry).
-- **linkedin-import** — sister plugin under the same pattern (LinkedIn Basic Data Export). LinkedIn ingestion is single-phase today (no enrich pass) because CSV rows already encode entity types deterministically — no auto-created participants, no auto-extracted observations to walk. Reading [linkedin-import/PLUGIN.md](../linkedin-import/PLUGIN.md) is the fastest way to understand the load-phase shape this plugin's Phase 1 follows.