refactor: remove playground proxy

.dockerignore

@@ -6,7 +6,7 @@
 
 credentials.json
 
-credentials/
+**/credentials
 
 **/bin
 

Makefile

@@ -1,3 +1,18 @@
+NPI_CMD_ROOT = $(shell pwd)
+GIT_COMMIT = $(shell git log -1 --format='%h' | awk '{print $0}')
+DOCKER_REGISTRY ?= 992297059634.dkr.ecr.us-west-2.amazonaws.com
+DOCKER_PLATFORM ?= linux/arm64,linux/amd64
+IMAGE_TAG ?= ${GIT_COMMIT}
+
+build-docker-python:
+	docker buildx build --platform ${DOCKER_PLATFORM} \
+		-t npiai/python:3.12 -f build/base.Dockerfile build --push
+
+build-docker-playground:
+	docker buildx build --platform ${DOCKER_PLATFORM} \
+		-t ${DOCKER_REGISTRY}/playground:${IMAGE_TAG} \
+		-f build/playground.Dockerfile . --push
+
 release:
 	poetry publish --build -u __token__ -p ${PYPI_TOKEN}
 	rm -rf dist

build/base.Dockerfile

@@ -1,4 +1,4 @@
-FROM --platform=$BUILDPLATFORM python:3.10
+FROM --platform=$BUILDPLATFORM python:3.12
 
 LABEL maintainer="Wenfeng Wang <[email protected]>"
 

build/playground.Dockerfile

@@ -0,0 +1,14 @@
+FROM --platform=$BUILDPLATFORM npiai/python:3.12
+
+LABEL maintainer="Wenfeng Wang <[email protected]>"
+
+COPY . /npi/
+
+WORKDIR /npi
+
+RUN poetry install
+ENV PATH="/root/.cache/pypoetry/virtualenvs/npiai-Gwe6qz7c-py3.11/bin:${PATH}"
+RUN playwright install-deps chromium
+RUN playwright install chromium
+
+ENTRYPOINT ["python", "playground/main.py"]

build/playwright/pyproject.toml

@@ -13,4 +13,4 @@ setuptools = "^69.2.0"
 
 [build-system]
 requires = ["poetry-core"]
-build-backend = "poetry.core.masonry.api"
+build-backend = "poetry.core.masonry.api"

npiai/context/context.py

@@ -80,7 +80,7 @@ class Context:
     __active_tool: Union['BaseTool', None] = None
     __last_screenshot: str | None = None
 
-    def __init__(self, instruction: str) -> None:
+    def __init__(self, instruction: str | None = None) -> None:
         self.id = str(uuid.uuid4())
         self.born_at = datetime.datetime.now()
         self.history: List[ThreadMessage | str] = []

npiai/core/base.py

@@ -38,12 +38,12 @@ def server(self):
         ...
 
     @abstractmethod
-    async def start(self):
+    async def start(self, ctx: Context | None = None):
         """Start the tools"""
         ...
 
     @abstractmethod
-    async def end(self):
+    async def end(self, ctx: Context | None = None):
         """Stop and dispose the tools"""
         ...
 

npiai/core/browser/navigator.py

@@ -170,10 +170,10 @@ def __init__(
         self.max_steps = max_steps
 
     # navigator uses shared playwright context, so we don't need to start it again here
-    async def start(self):
+    async def start(self, ctx: Context | None = None):
         pass
 
-    async def end(self):
+    async def end(self, ctx: Context | None = None):
         pass
 
     async def generate_user_prompt(self, task: str, history: List[Response]):

npiai/core/tool/agent.py

@@ -4,12 +4,10 @@
 import sys
 from typing import List, overload
 
-from litellm.types.completion import ChatCompletionMessageParam
 from fastapi import FastAPI, Request
 
 import uvicorn
 from npiai.llm import LLM, OpenAI
-from npiai.utils import logger
 from npiai.types import FunctionRegistration
 from npiai.context import Context, ThreadMessage
 from npiai.core import callback
@@ -33,7 +31,6 @@ def unpack_functions(self) -> List[FunctionRegistration]:
         # Wrap the chat function of this agent to FunctionRegistration
 
         def chat(message: str, ctx: Context):
-            # TODO: pass thread down
             return self.chat(message, ctx)
 
         fn_reg = FunctionRegistration(
@@ -59,11 +56,11 @@ def use_hitl(self, hitl: HITL):
         # super().use_hitl(hitl)
         self._tool.use_hitl(hitl)
 
-    async def start(self):
-        await self._tool.start()
+    async def start(self, ctx: Context | None = None):
+        await self._tool.start(ctx)
 
-    async def end(self):
-        await self._tool.end()
+    async def end(self, ctx: Context | None = None):
+        await self._tool.end(ctx)
 
     async def server(self):
         """Start the server"""
@@ -91,12 +88,12 @@ def signal_handler(sig, frame):
     async def chat(
             self,
             message: str,
-            thread: Context = None,
+            ctx: Context = None,
     ) -> str:
-        if thread is None:
-            thread = Context('')
+        if ctx is None:
+            ctx = Context()
 
-        msg = thread.fork(message)
+        msg = ctx.fork(message)
         if self._tool.system_prompt:
             msg.append(
                 {
@@ -112,7 +109,7 @@ async def chat(
             }
         )
 
-        return await self._call_llm(thread, msg)
+        return await self._call_llm(ctx, msg)
 
     async def _call_llm(self, thread: Context, message: ThreadMessage) -> str:
         while True:
@@ -158,7 +155,7 @@ async def chat(
             ctx: Context = None,
     ) -> str:
         if ctx is None:
-            ctx = Context('')
+            ctx = Context()
 
         if not self._tool.use_screenshot:
             return await super().chat(message, ctx)

npiai/core/tool/browser.py

@@ -5,6 +5,7 @@
 
 from npiai.core.tool.function import FunctionTool, function
 from npiai.core.browser.playwright import PlaywrightContext
+from npiai.context import Context
 
 
 class MdConverter(MarkdownConverter):
@@ -49,15 +50,15 @@ async def get_text(self):
         html = await self.playwright.page.evaluate('() => document.body.innerHTML')
         return MdConverter().convert(html)
 
-    async def start(self):
+    async def start(self, ctx: Context | None = None):
         """Start the Browser App"""
         if not self._started:
-            await super().start()
+            await super().start(ctx)
             await self.playwright.start()
 
-    async def end(self):
+    async def end(self, ctx: Context | None = None):
         """Dispose the chrome tools"""
-        await super().end()
+        await super().end(ctx)
         await self.playwright.stop()
 
     async def goto_blank(self):

npiai/core/tool/function.py

@@ -138,12 +138,12 @@ def use_hitl(self, hitl: HITL):
         for app in self._sub_tools:
             app.use_hitl(hitl)
 
-    async def start(self):
+    async def start(self, ctx: Context | None = None):
         """Start the tools"""
         if not self._started:
             self._started = True
             for tool in self._sub_tools:
-                await tool.start()
+                await tool.start(ctx)
 
     def server(self):
         """Start the server"""
@@ -161,14 +161,15 @@ def convert_camel_to_snake(name: str) -> str:
         @fapp.api_route("/{full_path:path}", methods=["GET", "POST", "PUT", "DELETE", "PATCH"])
         async def root(full_path: str, request: Request):
             method = convert_camel_to_snake(full_path)
+            ctx = Context(instruction="")
             try:
                 match request.method:
                     case "POST":
                         args = await request.json()
-                        res = await self._exec(method, args)
+                        res = await self._exec(ctx, method, args)
                     case "GET":
                         args = {k: v for k, v in request.query_params.items()}
-                        res = await self._exec(method, args)
+                        res = await self._exec(ctx, method, args)
                     case _:
                         return JSONResponse({'error': 'Method not allowed'}, status_code=405)
                 try:
@@ -198,12 +199,12 @@ def signal_handler(sig, frame):
 
         uvicorn.run(fapp, host="0.0.0.0", port=18000)
 
-    async def end(self):
+    async def end(self, ctx: Context | None = None):
         """Stop and dispose the tools"""
         if self._started:
             self._started = False
             for app in self._sub_tools:
-                await app.end()
+                await app.end(ctx)
 
     def add_tool(
             self,
@@ -223,8 +224,8 @@ def add_tool(
     async def debug(self, app_name: str = None, fn_name: str = None, args: Dict[str, Any] = None) -> str:
         if app_name:
             fn_name = f'{app_name}_{fn_name}'
-        ctx = Context('')
-        return await self._exec(ctx, fn_name, args)
+
+        return await self._exec(Context(), fn_name, args)
 
     async def call(
             self,
@@ -254,7 +255,8 @@ async def call(
                 res = await self._exec(ctx, fn_name, args)
             except Exception as e:
                 logger.error(e)
-                await ctx.send_msg(callback.Callable(f'Exception while executing {fn_name}: {e}'))
+                res = f'Exception while executing {fn_name}: {e}'
+                await ctx.send_msg(callback.Callable(res))
 
             logger.debug(f'[{self.name}]: function `{fn_name}` returned: {res}')
 
@@ -282,7 +284,6 @@ async def _exec(self, ctx: Context, fn_name: str, args: Dict[str, Any] = None) -
                 args = {tool.ctx_param_name: ctx}
             else:
                 args[tool.ctx_param_name] = ctx
-
         if args is None:
             return str(await tool.fn())
         return str(await tool.fn(**args))

npiai/tools/discord/app.py

@@ -4,6 +4,7 @@
 
 import discord
 
+from npiai.context import Context
 from npiai.utils import logger
 from npiai import FunctionTool, function
 from npiai.error.auth import UnauthorizedError
@@ -54,12 +55,12 @@ def __init__(self, access_token: str = None):
         self.client = discord.Client(intents=discord.Intents.default())
         self._access_token = token
 
-    async def start(self):
-        await super().start()
+    async def start(self, ctx: Context | None = None):
+        await super().start(ctx)
         await self.client.login(self._access_token)
 
-    async def end(self):
-        await super().end()
+    async def end(self, ctx: Context | None = None):
+        await super().end(ctx)
         await self.client.close()
 
     def parse_user(self, user: discord.User):