chore(docs,scripts): drop references left by the delete-first sweep (CL-8406)

README.md

@@ -19,8 +19,10 @@ bun run dev
 
 `bun run dev` starts the hub and the web dev server, applies migrations,
 and signs up a local admin (`alice@example.com` / `password123` unless
-`HUB_ADMIN_EMAIL` / `HUB_ADMIN_PASSWORD` are set). Connect an inference
-provider from the UI; keys are never read from the environment.
+`HUB_ADMIN_EMAIL` / `HUB_ADMIN_PASSWORD` are set). The first sign-in asks
+for one inference provider; keys are never read from the environment. A
+local [Ollama](https://ollama.com) works with no key, see
+[docs/local-dev.md](docs/local-dev.md).
 
 `bun run reset` wipes local state. It refuses anything but a local database.
 
@@ -31,8 +33,7 @@ provider from the UI; keys are never read from the environment.
 | `apps/`       | `hub` (API), `web` (React client), `sidecar` (execution host)    |
 | `packages/`   | Corbits libraries                                                |
 | `tools/`      | `@corbits/*` agent tool packages                                 |
-| `agents/`     | `@corbits/assistant` (Myra)                                      |
-| `skills/`     | `@corbits/*` skill packages                                      |
+| `agents/`     | `@corbits/myra`, the one shipped agent                           |
 | `vendor/intx` | Hand-copied Interchange packages, see [VENDORED.md](VENDORED.md) |
 
 ## Develop
@@ -47,5 +48,5 @@ Conventions live in [AGENTS.md](AGENTS.md).
 ## License
 
 The application is GPLv2 with the [AI Exception](GPLv2-AI-Exception.md);
-see [LICENSE.md](LICENSE.md). Libraries under `packages/`, `tools/`,
-`agents/`, and `skills/` are LGPL-2.1.
+see [LICENSE.md](LICENSE.md). Libraries under `packages/`, `tools/`, and
+`agents/` are LGPL-2.1.

docs/local-dev.md

@@ -1,16 +1,17 @@
 # Running fully local (Ollama)
 
 Workbench can run with no cloud LLM key at all, using a local
-[Ollama](https://ollama.com) instance as the inference provider. Connect
-the Ollama provider from the UI — its connect card asks for a base URL
-instead of a token and needs no key. Set `OLLAMA_BASE_URL` in `.env` too
-(e.g. `http://localhost:11434`, or a tailscale-tunneled origin): `bun run
+[Ollama](https://ollama.com) instance as the inference provider. On the
+first sign-in, pick "Ollama (local)": it asks for the base URL (default
+`http://localhost:11434/v1`) and a model name you have pulled, and needs
+no key. It rides Interchange's stock `openai-compatible` plugin. Set
+`OLLAMA_BASE_URL` in `.env` too (e.g. `http://localhost:11434`): `bun run
 dev` reads it to also mount the memory plane against the same origin when
 `EMBED_BASE_URL` is unset.
 
-Tool-heavy turns (anything that calls `mcp_list_tools`, dispatches a task,
-or chains several tool calls) take noticeably longer on small local
-models. That is a model-capability limit, not a platform bug.
+Tool-heavy turns (several tool calls in one turn) take noticeably longer
+on small local models. That is a model-capability limit, not a platform
+bug.
 
 ## Migrations after pulling
 

scripts/db-setup.ts

@@ -217,7 +217,7 @@ export async function setupDatabase(
 // would leave those tables behind. Always dropped alongside the target
 // schema so a reset is a true clean slate for every installed
 // package's tables, not only the platform's.
-const PACKAGE_SCHEMAS = ["mailbox", "cron", "webhook_triggers", "memory"] as const;
+const PACKAGE_SCHEMAS = ["mailbox", "cron", "memory", "artifacts"] as const;
 
 /**
  * Drop the target schema and everything in it (platform tables and