Files
hyprland-headless-stream-pl…/nix/mcp.nix
T
HermenixandClaude Opus 5 0a63c9b7fc feat(mcp)!: state as resources, acts as tools, and --toolsets for the rest
BREAKING: `status`, `session_list`, `session_status`, `session_logs` and
`pointer_state` are no longer tools. They are `hhstream://…` resources. 26
tools became 21, and eight resources appeared beside them.

The reason is the split the protocol itself draws, which the first cut ignored.
A resource is **application**-controlled context: the host decides what to put
in front of the model, and "what is on this machine right now" is the host's
question. A tool is **model**-controlled: an act the model chooses to perform.
Nothing in those five verbs acts. So the line between the halves is *who
decides*, not what a call costs.

    hhstream://status                     text, the units and the frame socket
    hhstream://sessions          {?all}   the broker's own list object
    hhstream://pointer                    the lease holder, buttons and keys
    hhstream://containers                 every container, plus the orphans
    hhstream://screen                     a frame, as an image blob
    hhstream://sessions/{id}     {?events}
    hhstream://sessions/{id}/logs {?lines}
    hhstream://outputs/{name}             one output's frame, as a blob

What did *not* move is the interesting part. `shot` and `record` stay tools
though neither changes the desktop: capturing is an act at a moment, and an
agent in a loop has to be able to decide to look *now* -- in many clients a
resource is something a person attaches rather than something a model can
fetch. The same frames are readable as `hhstream://screen` and
`hhstream://outputs/{name}` as well, for a host that wants one in context
itself. `session_container` stays a tool for a harder reason: with an id the
CLI runs `container.reconcile`, which writes the record back, so it is not a
read at all. Only the no-id listing is a resource. **A resource read must never
write**, and `resources.py` says so where it would be tempting.

Three consequences worth naming:

* A refusal cannot be content. `ReadResourceResult` has no `isError`, so a CLI
  refusal is raised as `-32602` with the same remedy text in `message` and the
  same `{ok, exit_code, error, reason, hint, remedy}` in `data`. Returning the
  broker's failure object as the *body* would make "no such session" read as a
  successful read of a session.
* Arguments become query parameters, and only declared ones are accepted -- for
  the reason every tool schema sets `additionalProperties: false`. A parameter
  silently ignored is a read that answered a different question.
* A v1 tool name is told where it went: `session_list` comes back with "It is a
  RESOURCE now ... read hhstream://sessions instead". "No such tool" alone
  reads as a broken server rather than a moved surface.

`--toolsets frames,sessions,pointer,keyboard` lands with it, because 21 flat
tools is the same problem one size smaller. There is **no grouping primitive in
MCP** -- `Tool` carries name, title, description, inputSchema, outputSchema,
icons, annotations, _meta and execution, and nowhere to put a family -- so the
only honest lever is which tools the *server* lists, chosen at launch. An
unknown name is argparse's exit 2, never a silently empty server, and a tool
outside the selection is unreachable rather than merely unlisted. Resources are
deliberately not filtered by it: they are the host's context, not the model's
budget.

`--list-resources` joins `--list-tools`; both are in the package's own
entry-point check, with `--toolsets` beside them.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011ZeH65zd6pFTgBSu52W9PE
2026-10-11 11:49:46 +00:00

115 lines
4.2 KiB
Nix

# `hhstream-mcp` -- the MCP server, as its own derivation.
#
# Separate from `nix/package.nix` on purpose, and the separation is the point:
# the frame socket, the CLI and the broker are stdlib-only by policy
# (`pyproject.toml`, `docs/CONTRACT-sessions.md` §6 rule 3), the MCP SDK is a
# third-party dependency, and `packages.hhstream` runs as the compositor's own
# units. So the dependency lives here, in a package nothing else imports, and
# `packages.hhstream` keeps the closure it had.
#
# `hhstream` is a *library* dependency here and not a binary one: the tools call
# `hhstream.cli`'s handlers in process (see that module's docstring), so what
# this package needs is the import path. `toPythonModule` is what makes an
# application's site-packages visible to a dependent, which a bare
# `buildPythonApplication` output is not.
#
# `grim` and `ffmpeg` do go on PATH, because the CLI itself spawns them --
# `shot --output` and `record` are two of the twenty-one tools, and the server
# is launched by an MCP client whose PATH is not ours to predict. `--prefix`
# and not `--set`, so a deployment that wants its own ahead of these puts them
# on PATH and wins.
{
lib,
python3,
hhstream,
grim,
ffmpeg,
makeWrapper,
}:
python3.pkgs.buildPythonApplication {
pname = "hhstream-mcp";
version = "0.1.0";
pyproject = true;
# The package directory is the project root: its own pyproject.toml lives
# there and maps the package name onto `.`. See the comment in that file for
# why it is not at the repository root.
src = ../src/hhstream_mcp;
build-system = [ python3.pkgs.setuptools ];
dependencies = [
python3.pkgs.mcp
python3.pkgs.anyio
# An application is not importable by a dependent until this converts it,
# and `hhstream_mcp.invoke` imports `hhstream.cli` directly.
(python3.pkgs.toPythonModule hhstream)
];
nativeBuildInputs = [ makeWrapper ];
makeWrapperArgs = [
"--prefix"
"PATH"
":"
"${lib.makeBinPath [
grim
ffmpeg
]}"
];
# The suite lives in `tests/` at the repository root and runs in the flake's
# `checks.unit`, which is also where `mcp` reaches it. Same choice as
# nix/package.nix, for the same reason.
doCheck = false;
# Cheap proof that both tables build and the SDK wiring imports: `tools` and
# `resources` construct every entry at import time, and `server` is the only
# module that touches `mcp`. A missing SDK, a schema that is not a dict, or a
# tool named twice all fail here rather than at an agent's first call.
pythonImportsCheck = [
"hhstream_mcp.tools"
"hhstream_mcp.resources"
"hhstream_mcp.server"
];
# `pythonImportsCheck` proves the modules load; it does not prove the script
# Nix wrapped can run, and the script is what an MCP client spawns.
# `--list-tools` and `--list-resources` are the two code paths that exercise
# argument parsing and both whole tables without opening stdio, so together
# they are the honest smoke test. `--toolsets` is in it because a selection
# that silently served everything would be the failure nobody notices.
#
# A `postPhases` entry and not `postInstall`, because the script is wrapped in
# `fixupPhase`: run any earlier and it has no PYTHONPATH, so `import mcp`
# fails and the check tests the wrong thing. `installCheckPhase` is no use
# either -- `buildPythonPackage` does not include that phase at all.
postPhases = [ "entryPointCheckPhase" ];
entryPointCheckPhase = ''
echo "checking the wrapped entry point..."
tools=$($out/bin/hhstream-mcp --list-tools)
echo "$tools"
test "$(echo "$tools" | wc -l)" -eq 21
resources=$($out/bin/hhstream-mcp --list-resources)
echo "$resources"
test "$(echo "$resources" | wc -l)" -eq 8
narrowed=$($out/bin/hhstream-mcp --list-tools --toolsets keyboard | wc -l)
echo "--toolsets keyboard offers $narrowed tools"
test "$narrowed" -eq 4
# An unknown toolset must be argparse's exit 2, not a server with no tools.
if $out/bin/hhstream-mcp --list-tools --toolsets nonsense; then
echo "an unknown toolset was accepted" >&2
exit 1
fi
'';
meta = {
description = "Drive a headless Hyprland session over the Model Context Protocol";
license = lib.licenses.mit;
mainProgram = "hhstream-mcp";
};
}