Introducing Llama Stack distributions (meta-llama#37)

* A single distro url to rule them all * fix typo * gut out the llama_agentic_system implementation and move it to toolchain there is `llama_agentic_system` sub-directory anymore * custom tool naming * updated var names to refer to distribution; updated requirements * make test run consistently * Updated README * Remove FP8 section, since it is rolled in now * Show safety configuration * Shorter subtitle lol * kill the dev dependencies * added ollama distro details to readme * Clarify Ollama instructions --------- Co-authored-by: Hardik Shah <[email protected]> Co-authored-by: dltn <[email protected]>
radreports · Aug 8, 2024 · afa0630 · afa0630
1 parent ce0c902
commit afa0630
Show file tree

Hide file tree

Showing 46 changed files with 290 additions and 3,494 deletions.
diff --git a/MANIFEST.in b/MANIFEST.in
@@ -1,2 +1 @@
 include requirements.txt
-include llama_agentic_system/data/*.yaml
diff --git a/README.md b/README.md
diff --git a/app/chat_moderation_with_llama_guard.py b/app/chat_moderation_with_llama_guard.py
@@ -8,15 +8,15 @@
 from utils.chat import chat, State
 from utils.client import ClientManager
 
-from utils.common import DISABLE_SAFETY, INFERENCE_HOST, INFERENCE_PORT, on_attach
+from utils.common import DISABLE_SAFETY, DISTRIBUTION_HOST, DISTRIBUTION_PORT, on_attach
 
 from utils.transform import transform
 
 
 client_manager = ClientManager()
 client_manager.init_client(
-    inference_port=INFERENCE_PORT,
-    host=INFERENCE_HOST,
+    inference_port=DISTRIBUTION_PORT,
+    host=DISTRIBUTION_HOST,
     custom_tools=[],
     disable_safety=DISABLE_SAFETY,
 )

diff --git a/app/chat_with_custom_tools.py b/app/chat_with_custom_tools.py
@@ -9,14 +9,14 @@
 from examples.custom_tools.ticker_data import TickerDataTool
 from utils.chat import chat, State
 from utils.client import ClientManager
-from utils.common import DISABLE_SAFETY, INFERENCE_HOST, INFERENCE_PORT, on_attach
+from utils.common import DISABLE_SAFETY, DISTRIBUTION_HOST, DISTRIBUTION_PORT, on_attach
 from utils.transform import transform
 
 
 client_manager = ClientManager()
 client_manager.init_client(
-    inference_port=INFERENCE_PORT,
-    host=INFERENCE_HOST,
+    inference_port=DISTRIBUTION_PORT,
+    host=DISTRIBUTION_HOST,
     custom_tools=[TickerDataTool()],
     disable_safety=DISABLE_SAFETY,
 )

diff --git a/app/main.py b/app/main.py
@@ -8,14 +8,14 @@
 
 from utils.chat import chat, State
 from utils.client import ClientManager
-from utils.common import DISABLE_SAFETY, INFERENCE_HOST, INFERENCE_PORT, on_attach
+from utils.common import DISABLE_SAFETY, DISTRIBUTION_HOST, DISTRIBUTION_PORT, on_attach
 from utils.transform import transform
 
 
 client_manager = ClientManager()
 client_manager.init_client(
-    inference_port=INFERENCE_PORT,
-    host=INFERENCE_HOST,
+    inference_port=DISTRIBUTION_PORT,
+    host=DISTRIBUTION_HOST,
     custom_tools=[],
     disable_safety=DISABLE_SAFETY,
 )

diff --git a/app/utils/chat.py b/app/utils/chat.py
@@ -17,7 +17,7 @@
 
 from llama_toolchain.inference.api import *  # noqa: F403
 from dotenv import load_dotenv
-from llama_agentic_system.api.datatypes import StepType
+from llama_toolchain.agentic_system.api.datatypes import StepType
 from llama_toolchain.safety.api.datatypes import ShieldResponse
 
 MAX_VIOLATIONS = 3
@@ -431,7 +431,7 @@ def render_tool(op_uuid: str, op: RenderableOutputType):
                     key=f"{len(state.output)}",
                     style=me.Style(
                         color=_COLOR_BUTTON,
-                    )
+                    ),
                 )
 
                 me.slide_toggle(label="Debug Mode", on_change=on_debug_mode_change)
diff --git a/app/utils/client.py b/app/utils/client.py
@@ -6,7 +6,7 @@
 
 import asyncio
 
-from llama_agentic_system.utils import get_agent_system_instance
+from llama_toolchain.agentic_system.utils import get_agent_system_instance
 
 global CLIENT
 CLIENT = None

diff --git a/app/utils/common.py b/app/utils/common.py
@@ -22,8 +22,8 @@
 
 UPLOADS_DIR = "app/uploads/"
 CHUNK_SIZE = 1024
-INFERENCE_PORT = os.environ.get("INFERENCE_PORT", 5000)
-INFERENCE_HOST = os.environ.get("INFERENCE_HOST", "localhost")
+DISTRIBUTION_PORT = os.environ.get("DISTRIBUTION_PORT", 5000)
+DISTRIBUTION_HOST = os.environ.get("DISTRIBUTION_HOST", "localhost")
 DISABLE_SAFETY = bool(int(os.environ.get("DISABLE_SAFETY", "0")))
 
 

diff --git a/app/utils/transform.py b/app/utils/transform.py
@@ -8,8 +8,8 @@
 import uuid
 
 import mesop as me
-from llama_agentic_system.api.datatypes import (
-    AgenticSystemTurnResponseEventType,
+from llama_toolchain.agentic_system.api import (
+    AgenticSystemTurnResponseEventType as EventType,
     StepType,
 )
 
@@ -24,9 +24,6 @@
 EVENT_LOOP = asyncio.new_event_loop()
 
 
-EventType = AgenticSystemTurnResponseEventType
-
-
 def transform(content: InterleavedTextAttachment):
     state = me.state(State)
 

diff --git a/examples/custom_tools/ticker_data.py b/examples/custom_tools/ticker_data.py
@@ -7,8 +7,10 @@
 from typing import Dict
 
 import yfinance as yf
-from llama_agentic_system.tools.custom import SingleMessageCustomTool
 from llama_models.llama3_1.api.datatypes import ToolParamDefinition
+from llama_toolchain.agentic_system.tools.custom.datatypes import (
+    SingleMessageCustomTool,
+)
 
 
 class TickerDataTool(SingleMessageCustomTool):

diff --git a/examples/notebooks/Cybersecurity_demo.ipynb b/examples/notebooks/Cybersecurity_demo.ipynb
@@ -32,7 +32,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from llama_agentic_system.import CodeInterpreterTool, with_safety\n",
+    "from llama_toolchain.agentic_system import CodeInterpreterTool, with_safety\n",
     "\n",
     "from llama_models.llama3_1.api import Message\n",
     "\n",

diff --git a/examples/notebooks/LlamaGuard.ipynb b/examples/notebooks/LlamaGuard.ipynb
@@ -46,7 +46,7 @@
     "\n",
     "from llama_toolchain.spec import *\n",
     "\n",
-    "from llama_agentic_system.utils import get_agent_system_instance\n",
+    "from llama_toolchain.agentic_system.utils import get_agent_system_instance\n",
     "\n",
     "\n",
     "async def get_assistant(host: str, port: int, disable_safety: bool = False):\n",

diff --git a/examples/scripts/multi_turn.py b/examples/scripts/multi_turn.py
@@ -10,14 +10,16 @@
 
 from dotenv import load_dotenv
 
-from llama_agentic_system.client import execute_with_custom_tools
-from llama_agentic_system.event_logger import EventLogger
+from llama_models.llama3_1.api.datatypes import *  # noqa: F403
 
-from llama_agentic_system.tools.custom import CustomTool
-from llama_agentic_system.utils import get_agent_system_instance
+from llama_toolchain.agentic_system.event_logger import EventLogger
+from llama_toolchain.agentic_system.tools.custom.datatypes import CustomTool
 
-from llama_models.llama3_1.api.datatypes import *  # noqa: F403
-from llama_agentic_system.api import *  # noqa: F403
+from llama_toolchain.agentic_system.tools.custom.execute import (
+    execute_with_custom_tools,
+)
+from llama_toolchain.agentic_system.utils import get_agent_system_instance
+from llama_toolchain.agentic_system.api import *  # noqa: F403
 
 from termcolor import cprint
 

diff --git a/llama_agentic_system/__init__.py b/llama_agentic_system/__init__.py
Original file line number	Diff line number	Diff line change
		@@ -1,2 +1 @@
		include requirements.txt
		include llama_agentic_system/data/*.yaml