Merge branch 'master' into validation

.github/workflows/autogpt-ci.yml

@@ -22,6 +22,10 @@ defaults:
  run:
  working-directory: autogpts/autogpt
 
+env:
+ TELEMETRY_ENVIRONMENT: autogpt-ci
+ TELEMETRY_OPT_IN: ${{ github.event_name == 'push' && github.ref_name == 'master' }}
+
 jobs:
  lint:
  runs-on: ubuntu-latest

.github/workflows/autogpt-docker-ci.yml

@@ -27,6 +27,9 @@ env:
  DEPLOY_IMAGE_NAME: ${{ secrets.DOCKER_USER }}/auto-gpt
  DEV_IMAGE_TAG: latest-dev
 
+ TELEMETRY_ENVIRONMENT: autogpt-docker-ci
+ TELEMETRY_OPT_IN: ${{ github.event_name == 'push' && github.ref_name == 'master' }}
+
 jobs:
  build:
  runs-on: ubuntu-latest

.github/workflows/autogpts-ci.yml

@@ -25,6 +25,10 @@ on:
  - 'setup.py'
  - '!**/*.md'
 
+env:
+ TELEMETRY_ENVIRONMENT: autogpt-ci
+ TELEMETRY_OPT_IN: ${{ github.event_name == 'push' && github.ref_name == 'master' }}
+
 jobs:
  run-tests:
  runs-on: ubuntu-latest
@@ -56,7 +60,7 @@ jobs:
  run: |
  ./run agent start ${{ matrix.agent-name }}
  cd autogpts/${{ matrix.agent-name }}
- poetry run agbenchmark --mock
+ poetry run agbenchmark --mock --test=BasicRetrieval --test=Battleship --test=WebArenaTask_0
  poetry run agbenchmark --test=WriteFile
  env:
  OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

.github/workflows/benchmark-ci.yml

@@ -16,6 +16,8 @@ on:
 
 env:
  min-python-version: '3.10'
+ TELEMETRY_ENVIRONMENT: autogpt-benchmark-ci
+ TELEMETRY_OPT_IN: ${{ github.event_name == 'push' && github.ref_name == 'master' }}
 
 jobs:
  lint:

autogpts/autogpt/autogpt/app/agent_protocol_server.py

@@ -25,6 +25,7 @@
 from forge.sdk.routes.agent_protocol import base_router
 from hypercorn.asyncio import serve as hypercorn_serve
 from hypercorn.config import Config as HypercornConfig
+from sentry_sdk import set_user
 
 from autogpt.agent_factory.configurators import configure_agent_with_state
 from autogpt.agent_factory.generators import generate_agent_for_task
@@ -121,6 +122,9 @@ async def create_task(self, task_request: TaskRequestBody) -> Task:
  """
  Create a task for the agent.
  """
+ if user_id := (task_request.additional_input or {}).get("user_id"):
+ set_user({"id": user_id})
+
  task = await self.db.create_task(
  input=task_request.input,
  additional_input=task_request.additional_input,
@@ -180,6 +184,9 @@ async def execute_step(self, task_id: str, step_request: StepRequestBody) -> Ste
  llm_provider=self._get_task_llm_provider(task),
  )
 
+ if user_id := (task.additional_input or {}).get("user_id"):
+ set_user({"id": user_id})
+
  # According to the Agent Protocol spec, the first execute_step request contains
  # the same task input as the parent create_task request.
  # To prevent this from interfering with the agent's process, we ignore the input

autogpts/autogpt/autogpt/app/telemetry.py

@@ -2,7 +2,7 @@
 
 import click
 
-from .utils import env_file_exists, set_env_config_value
+from .utils import env_file_exists, get_git_user_email, set_env_config_value
 
 
 def setup_telemetry() -> None:
@@ -16,9 +16,12 @@ def setup_telemetry() -> None:
  allow_telemetry = click.prompt(
  "❓ Do you want to enable telemetry? ❓\n"
  "This means AutoGPT will send diagnostic data to the core development team "
- "when something goes wrong, and will help us to diagnose and fix problems "
- "earlier and faster.\n"
- "Please enter 'yes' or 'no'",
+ "when something goes wrong,\nand will help us to diagnose and fix problems "
+ "earlier and faster.\n\n"
+ "By entering 'yes', you agree that you have read and agree to our Privacy"
+ " Policy available here:\n"
+ "https://www.notion.so/auto-gpt/Privacy-Policy-ab11c9c20dbd4de1a15dcffe84d77984"
+ "\n\nPlease enter 'yes' or 'no'",
  type=bool,
  )
  set_env_config_value("TELEMETRY_OPT_IN", "true" if allow_telemetry else "false")
@@ -38,3 +41,6 @@ def _setup_sentry() -> None:
  dsn="https://dc266f2f7a2381194d1c0fa36dff67d8@o4505260022104064.ingest.sentry.io/4506739844710400", # noqa
  environment=os.getenv("TELEMETRY_ENVIRONMENT"),
  )
+
+ # Allow Sentry to distinguish between users
+ sentry_sdk.set_user({"email": get_git_user_email(), "ip_address": "{{auto}}"})

autogpts/autogpt/autogpt/app/utils.py

@@ -84,6 +84,14 @@ def get_current_git_branch() -> str:
  return ""
 
 
+def get_git_user_email() -> str:
+ try:
+ repo = Repo(search_parent_directories=True)
+ return repo.config_reader().get_value("user", "email", default="")
+ except InvalidGitRepositoryError:
+ return ""
+
+
 def get_latest_bulletin() -> tuple[str, bool]:
  exists = os.path.exists("data/CURRENT_BULLETIN.md")
  current_bulletin = ""

autogpts/autogpt/autogpt/core/resource/model_providers/openai.py

@@ -423,7 +423,7 @@ async def create_chat_completion(
  tool_calls=(
  [AssistantToolCall(**tc.dict()) for tc in _assistant_msg.tool_calls]
  if _assistant_msg.tool_calls
- else None
+ else list()
  ),
  )
  response = ChatModelResponse(
@@ -466,6 +466,7 @@ async def create_chat_completion(
  extras={"assistant_msg": assistant_msg, "i_attempt": attempts},
  )
  if attempts < self._configuration.fix_failed_parse_tries:
+ model_prompt.append(assistant_msg)
  model_prompt.append(
  ChatMessage.system(f"ERROR PARSING YOUR RESPONSE:\n\n{e}")
  )

autogpts/autogpt/autogpt/core/resource/model_providers/schema.py

@@ -44,16 +44,14 @@ class Role(str, enum.Enum):
  SYSTEM = "system"
  ASSISTANT = "assistant"
 
+ TOOL = "tool"
+ """May be used for the result of tool calls"""
  FUNCTION = "function"
  """May be used for the return value of function calls"""
 
  role: Role
  content: str
 
- @staticmethod
- def assistant(content: str) -> "ChatMessage":
- return ChatMessage(role=ChatMessage.Role.ASSISTANT, content=content)
-
  @staticmethod
  def user(content: str) -> "ChatMessage":
  return ChatMessage(role=ChatMessage.Role.USER, content=content)
@@ -93,7 +91,7 @@ class AssistantToolCallDict(TypedDict):
 class AssistantChatMessage(ChatMessage):
  role: Literal["assistant"] = "assistant"
  content: Optional[str]
- tool_calls: Optional[list[AssistantToolCall]]
+ tool_calls: list[AssistantToolCall] = Field(default_factory=list)
 
 
 class AssistantChatMessageDict(TypedDict, total=False):

autogpts/autogpt/pyproject.toml

@@ -24,7 +24,7 @@ serve = "autogpt.app.cli:serve"
 python = "^3.10"
 auto-gpt-plugin-template = {git = "https://github.com/Significant-Gravitas/Auto-GPT-Plugin-Template", rev = "0.1.0"}
 # autogpt-forge = { path = "../forge" }
-autogpt-forge = {git = "https://github.com/Significant-Gravitas/AutoGPT.git", rev = "6763196d78603808dcbb", subdirectory = "autogpts/forge"}
+autogpt-forge = {git = "https://github.com/Significant-Gravitas/AutoGPT.git", rev = "ab05b7ae70754c063909", subdirectory = "autogpts/forge"}
 beautifulsoup4 = "^4.12.2"
 boto3 = "^1.33.6"
 charset-normalizer = "^3.1.0"

benchmark/agbenchmark/agent_api_interface.py

@@ -1,5 +1,4 @@
 import logging
-import os
 import time
 from pathlib import Path
 from typing import AsyncIterator, Optional
@@ -23,6 +22,8 @@ async def run_api_agent(
  config: AgentBenchmarkConfig,
  timeout: int,
  artifacts_location: Optional[Path] = None,
+ *,
+ mock: bool = False,
 ) -> AsyncIterator[Step]:
  configuration = Configuration(host=config.host)
  async with ApiClient(configuration) as api_client:
@@ -36,26 +37,32 @@ async def run_api_agent(
  task_id = response.task_id
 
  if artifacts_location:
+ logger.debug("Uploading task input artifacts to agent...")
  await upload_artifacts(
  api_instance, artifacts_location, task_id, "artifacts_in"
  )
 
+ logger.debug("Running agent until finished or timeout...")
  while True:
  step = await api_instance.execute_agent_task_step(task_id=task_id)
  yield step
 
  if time.time() - start_time > timeout:
  raise TimeoutError("Time limit exceeded")
+ if step and mock:
+ step.is_last = True
  if not step or step.is_last:
  break
 
  if artifacts_location:
  # In "mock" mode, we cheat by giving the correct artifacts to pass the test
- if os.getenv("IS_MOCK"):
+ if mock:
+ logger.debug("Uploading mock artifacts to agent...")
  await upload_artifacts(
  api_instance, artifacts_location, task_id, "artifacts_out"
  )
 
+ logger.debug("Downloading agent artifacts...")
  await download_agent_artifacts_into_folder(
  api_instance, task_id, config.temp_folder
  )

benchmark/agbenchmark/challenges/base.py

@@ -60,7 +60,7 @@ def test_method(
 
  @classmethod
  async def run_challenge(
- cls, config: AgentBenchmarkConfig, timeout: int
+ cls, config: AgentBenchmarkConfig, timeout: int, *, mock: bool = False
  ) -> AsyncIterator[Step]:
  """
  Runs the challenge on the subject agent with the specified timeout.
@@ -89,7 +89,7 @@ async def run_challenge(
  logger.debug(f"Starting {cls.info.name} challenge run")
  i = 0
  async for step in run_api_agent(
- cls.info.task, config, timeout, cls.info.task_artifacts_dir
+ cls.info.task, config, timeout, cls.info.task_artifacts_dir, mock=mock
  ):
  i += 1
  print(f"[{cls.info.name}] - step {step.name} ({i}. request)")