Remove legacy install extras

Drop AbstractGateway's compatibility-only extras (http/server/etc) so the supported install surface is just base, apple, and gpu. Update docs + install-profile tests and regenerate llms-full.

Author: lpalbou

CHANGELOG.md

@@ -7,6 +7,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Changed
+
+- Removed legacy compatibility install extras (`abstractgateway[http]`, `[server]`, `[multimodal]`, `[memory]`, `[voice]`, `[vision]`, `[telegram]`, `[visualflow]`, `[all]`, `[all-apple]`, `[all-gpu]`, `[server-nvidia]`). The supported install surface is now:
+  - `pip install abstractgateway`
+  - `pip install "abstractgateway[apple]"`
+  - `pip install "abstractgateway[gpu]"`
+
 ## [0.2.20] - 2026-05-26
 
 ### Added

README.md

@@ -200,9 +200,6 @@ pip install abstractgateway
 
 - `abstractgateway[apple]`: full native macOS Python profile with Apple-local engines and all non-NVIDIA framework capabilities
 - `abstractgateway[gpu]`: full local GPU profile with vLLM/HuggingFace, local Diffusers image generation, local voice engines, music, and KG memory; this is also the NVIDIA Docker install profile
-- `abstractgateway[server]`, `[http]`, `[multimodal]`, `[memory]`, `[voice]`, `[vision]`, and `[all]`: compatibility aliases; the base package already includes the remote-light server stack
-- `abstractgateway[server-nvidia]`: compatibility alias for the GPU profile used by older NVIDIA Docker commands
-- `abstractgateway[telegram]` and `[visualflow]`: compatibility aliases; Telegram bridge and VisualFlow publishing use base install dependencies
 - `abstractgateway[docs]`: MkDocs site tooling
 - `abstractgateway[dev]`: local test/dev deps
 

docs/configuration.md

@@ -12,10 +12,6 @@ compatibility (runs bundles produced by AbstractFlow; does not require the `abst
 Optional extras (see `pyproject.toml`):
 - `abstractgateway[apple]`: full native macOS Python profile with Apple-local engines and all non-NVIDIA framework capabilities; this is for native macOS, not Docker
 - `abstractgateway[gpu]`: full native/container GPU profile with local GPU engines and all relevant framework capabilities; the NVIDIA Docker image uses this profile
-- `abstractgateway[http]`, `[server]`, `[multimodal]`, `[memory]`, `[voice]`, `[vision]`, `[all]`: compatibility aliases because the base install now includes the remote-light server stack
-- `abstractgateway[server-nvidia]`: compatibility alias for the GPU profile used by older NVIDIA Docker commands
-- `abstractgateway[visualflow]`: deprecated compatibility alias (VisualFlow directory mode was removed; does nothing)
-- `abstractgateway[telegram]`: compatibility alias; Telegram bridge support uses base install dependencies
 - `abstractgateway[docs]`: MkDocs site tooling
 - `abstractgateway[dev]`: local dev/test deps
 
@@ -200,14 +196,14 @@ and hybrid runtimes report `prompt_cache_local_only`.
 
 ### Multimodal provider/plugin controls
 
-The `server`, `multimodal`, and `all` extras install the lightweight AbstractCore
-plugin surface for generated images, generated videos, generated voice, STT, and music
-capabilities. Voice generation, transcription, generated images, and generated
-music have direct Gateway endpoints; generated images are also exposed through
-Runtime/Core workflows and through `POST /api/gateway/runs/{run_id}/images/generate`
-and `POST /api/gateway/runs/{run_id}/images/edit` when a Runtime/Core image
-backend is configured or provider/model are supplied on the request. Local
-heavy engines remain explicit opt-ins in the provider packages.
+The base install already includes the Gateway HTTP/SSE server and the Runtime
+multimodal integration layer. Direct Gateway routes for voice/audio, image/video,
+and music become available when the corresponding lower-layer capability
+packages are installed on the gateway host (or when Gateway is configured to
+proxy to a remote AbstractCore server).
+
+Local heavy engines remain explicit opt-ins in the provider packages; Gateway
+does not implicitly install them.
 
 - `output.text` capability route: default text model for bundle LLM nodes
 - `OPENAI_COMPATIBLE_BASE_URL` / `OPENAI_COMPATIBLE_API_KEY`: OpenAI-compatible text endpoint for AbstractCore providers

docs/getting-started.md

@@ -38,10 +38,6 @@ pip install "abstractgateway[apple]"
 pip install "abstractgateway[gpu]"
 ```
 
-Compatibility extras such as `abstractgateway[http]`, `[server]`, `[memory]`,
-`[multimodal]`, `[voice]`, `[vision]`, and `[all]` remain accepted, but the
-base install already contains the remote-light server stack.
-
 With the base install and a configured provider stack, Gateway can surface
 run-scoped direct TTS, STT, image generation, image edit, and music generation
 for higher apps through one shared capability contract.

llms-full.txt

@@ -210,9 +210,6 @@ pip install abstractgateway
 
 - `abstractgateway[apple]`: full native macOS Python profile with Apple-local engines and all non-NVIDIA framework capabilities
 - `abstractgateway[gpu]`: full local GPU profile with vLLM/HuggingFace, local Diffusers image generation, local voice engines, music, and KG memory; this is also the NVIDIA Docker install profile
-- `abstractgateway[server]`, `[http]`, `[multimodal]`, `[memory]`, `[voice]`, `[vision]`, and `[all]`: compatibility aliases; the base package already includes the remote-light server stack
-- `abstractgateway[server-nvidia]`: compatibility alias for the GPU profile used by older NVIDIA Docker commands
-- `abstractgateway[telegram]` and `[visualflow]`: compatibility aliases; Telegram bridge and VisualFlow publishing use base install dependencies
 - `abstractgateway[docs]`: MkDocs site tooling
 - `abstractgateway[dev]`: local test/dev deps
 
@@ -357,10 +354,6 @@ pip install "abstractgateway[apple]"
 pip install "abstractgateway[gpu]"
 ```
 
-Compatibility extras such as `abstractgateway[http]`, `[server]`, `[memory]`,
-`[multimodal]`, `[voice]`, `[vision]`, and `[all]` remain accepted, but the
-base install already contains the remote-light server stack.
-
 With the base install and a configured provider stack, Gateway can surface
 run-scoped direct TTS, STT, image generation, image edit, and music generation
 for higher apps through one shared capability contract.
@@ -1736,10 +1729,6 @@ compatibility (runs bundles produced by AbstractFlow; does not require the `abst
 Optional extras (see `pyproject.toml`):
 - `abstractgateway[apple]`: full native macOS Python profile with Apple-local engines and all non-NVIDIA framework capabilities; this is for native macOS, not Docker
 - `abstractgateway[gpu]`: full native/container GPU profile with local GPU engines and all relevant framework capabilities; the NVIDIA Docker image uses this profile
-- `abstractgateway[http]`, `[server]`, `[multimodal]`, `[memory]`, `[voice]`, `[vision]`, `[all]`: compatibility aliases because the base install now includes the remote-light server stack
-- `abstractgateway[server-nvidia]`: compatibility alias for the GPU profile used by older NVIDIA Docker commands
-- `abstractgateway[visualflow]`: deprecated compatibility alias (VisualFlow directory mode was removed; does nothing)
-- `abstractgateway[telegram]`: compatibility alias; Telegram bridge support uses base install dependencies
 - `abstractgateway[docs]`: MkDocs site tooling
 - `abstractgateway[dev]`: local dev/test deps
 
@@ -1924,14 +1913,14 @@ and hybrid runtimes report `prompt_cache_local_only`.
 
 ### Multimodal provider/plugin controls
 
-The `server`, `multimodal`, and `all` extras install the lightweight AbstractCore
-plugin surface for generated images, generated videos, generated voice, STT, and music
-capabilities. Voice generation, transcription, generated images, and generated
-music have direct Gateway endpoints; generated images are also exposed through
-Runtime/Core workflows and through `POST /api/gateway/runs/{run_id}/images/generate`
-and `POST /api/gateway/runs/{run_id}/images/edit` when a Runtime/Core image
-backend is configured or provider/model are supplied on the request. Local
-heavy engines remain explicit opt-ins in the provider packages.
+The base install already includes the Gateway HTTP/SSE server and the Runtime
+multimodal integration layer. Direct Gateway routes for voice/audio, image/video,
+and music become available when the corresponding lower-layer capability
+packages are installed on the gateway host (or when Gateway is configured to
+proxy to a remote AbstractCore server).
+
+Local heavy engines remain explicit opt-ins in the provider packages; Gateway
+does not implicitly install them.
 
 - `output.text` capability route: default text model for bundle LLM nodes
 - `OPENAI_COMPATIBLE_BASE_URL` / `OPENAI_COMPATIBLE_API_KEY`: OpenAI-compatible text endpoint for AbstractCore providers

pyproject.toml

@@ -25,15 +25,14 @@ dependencies = [
 ]
 
 [project.optional-dependencies]
-# The base package is the remote-light server install. These extras remain as
-# compatibility aliases for existing install commands.
-http = []
-server = []
-multimodal = []
-memory = []
-
-# Native Apple Gateway profile. This is for direct macOS Python installs, not
-# Docker, because MLX/Metal-backed engines need the host Apple runtime. As the
+# Supported install profiles.
+# - Base install (`pip install abstractgateway`) is the remote-light HTTP/SSE server.
+# - `apple` / `gpu` select hardware-local stacks for the gateway host.
+#
+# Tooling extras for contributors: `dev`, `docs`.
+
+# Native Apple Gateway profile. This is for direct macOS Python installs (not
+# Docker), because MLX/Metal-backed engines need the host Apple runtime. As the
 # deployment entry point, Gateway's `apple` profile installs the full native
 # Apple stack and all relevant non-NVIDIA framework capabilities.
 apple = [
@@ -49,41 +48,6 @@ gpu = [
   "abstractagent[all-gpu]>=0.3.8",
   "AbstractMemory[all-gpu]>=0.2.6",
 ]
-
-# Native macOS aggregate profile for full local+remote Gateway installs.
-all-apple = [
-  "AbstractRuntime[multimodal,mcp-worker,all-apple]>=0.4.24",
-  "abstractagent[all-apple]>=0.3.8",
-  "AbstractMemory[all-apple]>=0.2.6",
-]
-
-# Full GPU aggregate profile for CUDA/ROCm-style local engine hosts.
-all-gpu = [
-  "AbstractRuntime[multimodal,mcp-worker,all-gpu]>=0.4.24",
-  "abstractagent[all-gpu]>=0.3.8",
-  "AbstractMemory[all-gpu]>=0.2.6",
-]
-
-# Compatibility alias for historical Docker builds. New NVIDIA Docker builds use
-# `abstractgateway[gpu]`.
-server-nvidia = [
-  "AbstractRuntime[multimodal,mcp-worker,all-gpu]>=0.4.24",
-  "abstractagent[all-gpu]>=0.3.8",
-  "AbstractMemory[all-gpu]>=0.2.6",
-]
-
-# Voice/audio capabilities (TTS + STT) via AbstractVoice.
-voice = []
-
-# Generative vision capabilities via AbstractCore's AbstractVision plugin.
-vision = []
-
-# Compatibility alias (deprecated): VisualFlow directory mode was removed.
-visualflow = []
-
-# Telegram's AbstractCore tool integration is already present in the base
-# remote-light server install.
-telegram = []
 dev = [
   "pytest>=7.0.0",
   "httpx>=0.27.0",
@@ -94,11 +58,6 @@ docs = [
   "mkdocs-material>=9.0.0",
 ]
 
-# Compatibility alias. The base install is already the remote-light
-# batteries-included server; hardware-local engines are selected by `apple` or
-# `gpu`.
-all = []
-
 [project.scripts]
 abstractgateway = "abstractgateway.cli:main"
 abstractgateway-config = "abstractgateway.config_cli:main"

tests/test_gateway_install_profiles.py

@@ -43,19 +43,29 @@ def test_base_install_is_remote_light_server() -> None:
 def test_entrypoint_profiles_cascade_lower_package_extras() -> None:
     extras = _pyproject()["project"]["optional-dependencies"]
 
-    assert extras["http"] == []
-    assert extras["server"] == []
-    assert extras["multimodal"] == []
-    assert extras["memory"] == []
-    assert extras["voice"] == []
-    assert extras["vision"] == []
-    assert extras["visualflow"] == []
-    assert extras["telegram"] == []
-    assert extras["all"] == []
+    # Legacy compatibility aliases were removed; keep the user-facing install
+    # surface minimal and unambiguous.
+    for legacy in (
+        "http",
+        "server",
+        "multimodal",
+        "memory",
+        "voice",
+        "vision",
+        "visualflow",
+        "telegram",
+        "all",
+        "all-apple",
+        "all-gpu",
+        "server-nvidia",
+    ):
+        assert legacy not in extras
+
     assert "apple" in extras
     assert "gpu" in extras
-    assert "all-apple" in extras
-    assert "all-gpu" in extras
+    # Tooling extras remain for contributors/CI.
+    assert "dev" in extras
+    assert "docs" in extras
 
     apple = "\n".join(extras["apple"])
     assert "AbstractRuntime[multimodal,mcp-worker,all-apple]>=0.4.24" in apple
@@ -75,16 +85,6 @@ def test_entrypoint_profiles_cascade_lower_package_extras() -> None:
     assert "abstractvoice" not in gpu
     assert "abstractmusic" not in gpu
 
-    assert extras["all-apple"] == extras["apple"]
-    assert extras["all-gpu"] == extras["gpu"]
-
-    nvidia = "\n".join(extras["server-nvidia"])
-    assert nvidia == gpu
-
-    assert "AbstractMemory[all-apple]>=0.2.6" in extras["all-apple"]
-    assert "AbstractMemory[all-gpu]>=0.2.6" in extras["all-gpu"]
-    assert "AbstractMemory[all-gpu]>=0.2.6" in extras["server-nvidia"]
-
 
 def test_config_entrypoint_is_published() -> None:
     scripts = _pyproject()["project"]["scripts"]