@vladchatware/dsh-fs-docker

Docker implementation of the filesystem seam (ctx.fs). Paths, contents, and atomic staging files stay inside the shared container owned by @vladchatware/dsh-docker. Load the owner first, then this service in place of dsh-fs-local; dsh-tool-fs and the policy layer then operate on the container execution world unchanged.

Implementation notes and known limits:

• Canonical targets are absolute guest paths transported as base64 with strict NUL framing; resolve/processPath/fileUrl/contains project container paths into the seam.
• Version tokens derive from guest stat facts (path, kind, size, mode, mtime) — the Docker metadata API exposes no per-file revision, so two writes within the mtime tick with identical facts collide and the stale guard degrades to a no-op.
• Atomic writes stage into a private sibling temp name and commit with a container-side mv (a guarded ln -T no-clobber publication for createIfAbsent), mirroring the E2B adapter.
• lstat reports symlinks via an in-guest helper because dockerode exposes no no-follow primitive.
• Text/binary handling mirrors the seam: UTF-8 validation and an 8192-byte NUL sample reject binary and invalid content as FS_NOT_TEXT.
• The logged image must contain the GNU utilities this adapter invokes (realpath, base64, stat, ln, chmod — debian:bookworm-slim does).

Model Experience

Indirectly, through dsh-tool-fs, which renders remote UTF-8 content, directory results, mutation acknowledgements, and provider errors while container identity and transport remain internal.

KV Cache effect

No direct invalidation; the named consumer owns any request-prefix changes.

Known Limitations and Deferred Work

• No host synchronization — the container cwd starts empty; local files are neither uploaded nor reflected back.
• Mutation coordination is host-process-local — a concurrent container process can still race replacement; version guards detect only metadata changes Docker exposes.
• Whole-file mutation costs remain — overwrite diffs and literal edits read complete files through the archive transport.
• File transfer cost — every read and write crosses the Docker archive API tar in/out of the container.