AGENTS.md
+18 −3
https://raw.githubusercontent.com/grunion-ai/weave/HEAD/AGENTS.md
− | | `weave_query` | rows of one table, filtered (where), sorted, paged |− screen*: the eighteen field types with how each renders in the grid and which− | `weave ontology [--concept T] [--since E]` / `weave schema` | `weave space create` / `weave space` / `weave space update` / `weave space delete` / `weave space restore` | `weave create` / `weave get` / `weave query` |+ | | `weave_query` | rows of one table, filtered (where), sorted, 200 a page |+ ### Local file uploads+ For local files, call `weave_call` with `name: "weave_attach_file"` and+ `args: {entity, field, name, mime, prepare: true, size}`. `field` is an optional+ attachments field; `size` is the exact file size in bytes, up to 10 MiB.+ The authenticated HTTP connector returns `{url, method, headers, maxBytes,+ expiresAt}`. A local script sends the file bytes to that URL with the returned+ method and headers. File bytes never need to enter the model context, and no+ separate editor token is needed.+ The upload grant authorizes one file at the prepared destination for ten+ minutes. Every transfer attempt consumes it, even if the transfer fails.+ Prepare a fresh grant after failure or server restart. After an uncertain+ response, read the row's attachments before retrying to avoid a duplicate.+ Keep the returned authorization header private. The server rechecks the+ account's current role before accepting bytes and before attaching them.+ For small inline payloads, `contentBase64` still works, with optional `field`.+ screen*: every field type with how it renders in the grid and which+ | `weave ontology [--concept T] [--since E]` / `weave schema` | `weave space create` / `weave space` / `weave space update` / `weave space members` / `weave space delete` / `weave space restore` | `weave create` / `weave get` / `weave query` |