feat(server): inference state-graph + supervisor, memory tools, MCP client, schema, routes

- Add state-graph.ts: typed state machine for inference lifecycle
- Add supervisor.ts: agent supervisor pattern for multi-agent coordination
- Add export-formatter.ts: structured export formatting
- Add manage_memory.ts: memory CRUD tool for agent persistence
- Add get_wiki_article.ts: codecontext wiki article retrieval
- Extend memory/index.ts: 3-tier memory (context/daily/core)
- Extend MCP client: mcp-config.ts env-var substitution
- Update schema.sql: agent_sessions, tasks, pending_changes extensions
- Update API types: MessageMetadata, ErrorReason, AgentSessionConfig
- Update routes: chats, messages, sessions — column renames and agent_session_id
- Update inference: error handler, payload builder, stream phase, turn orchestrator

apps/server/CLAUDE.md

@@ -1,7 +1,12 @@
-# apps/server — BooChat backend (deep reference)
+# apps/server — BooChat backend (deep reference) — v2.7.x (last meaningful update: 2026-06)
 
 > Per-app engineering notes for `apps/server/src/`. Cross-cutting commands, database, environment, workflow, and cross-app contracts (WS-frame / provider-type parity, sentinels) live in the **root `CLAUDE.md`**. This file auto-loads when you read/edit files under `apps/server/`.
 
+## These gotchas are load-bearing — do not remove or refactor without understanding why
+- Do NOT remove the abort-signal pinning comment in `stream-phase.ts` — `fullStream` exits cleanly on abort without throwing; the post-iteration `if (signal?.aborted)` check is the only thing that distinguishes cancelled from complete.
+- Do NOT remove `includeUsage: true` from `provider.ts` — the adapter defaults it false; without it, token counts are always NULL.
+- Do NOT add raw `broker.publish()`/`publishUser()` calls — always use `publishFrame`/`publishUserFrame` which Zod-validate against `WsFrameSchema`.
+
 ## Stack
 
 - **Fastify** with `@fastify/websocket` and `@fastify/static` (serves the built frontend).