Merge pull request #13 from manuelseeger/debarcode

Debarcode - unmask barcode players

.gitignore

@@ -17,4 +17,8 @@ mongodb/mongo-seed/*.json
 obs_tools/ttsmodels/
 tests/coverage/
 .coverage
-tests/test-results.xml
+tests/test-results.xml
+*report.html
+mongodb/mongo-seed/*.json
+mongodb/dump/
+playground/external/

.vscode/launch.json

@@ -11,7 +11,6 @@
       "program": "coach.py",
       "console": "integratedTerminal",
       "envFile": "${workspaceFolder}/.env",
-
       "env": {
         "PYDEVD_DISABLE_FILE_VALIDATION": "1"
       }
@@ -23,7 +22,6 @@
       "program": "coach.py",
       "console": "integratedTerminal",
       "envFile": "${workspaceFolder}/.env",
-
       "env": {
         "PYDEVD_DISABLE_FILE_VALIDATION": "1",
         "AICOACH_audiomode": "text",
@@ -37,7 +35,6 @@
       "program": "coach.py",
       "console": "integratedTerminal",
       "envFile": "${workspaceFolder}/.env",
-
       "env": {
         "PYDEVD_DISABLE_FILE_VALIDATION": "1",
         "AICOACH_aibackend": "Mocked"
@@ -50,34 +47,25 @@
       "program": "build.py",
       "console": "integratedTerminal",
       "envFile": "${workspaceFolder}/.env",
-      "args": ["--deploy"]
+      "args": [
+        "--deploy"
+      ]
     },
     {
-      "name": "CLI sync",
+      "name": "CLI full sync",
       "type": "debugpy",
       "request": "launch",
       "program": "repcli.py",
       "console": "integratedTerminal",
       "envFile": "${workspaceFolder}/.env",
-      "args": ["--debug", "--clean", "sync"]
-    },
-    {
-      "name": "CLI add players",
-      "type": "debugpy",
-      "request": "launch",
-      "program": "repcli.py",
-      "console": "integratedTerminal",
-      "envFile": "${workspaceFolder}/.env",
-      "args": ["--debug", "addplayers", "--from=2024-06-07"]
-    },
-    {
-      "name": "CLI deamon",
-      "type": "debugpy",
-      "request": "launch",
-      "program": "repcli.py",
-      "console": "integratedTerminal",
-      "envFile": "${workspaceFolder}/.env",
-      "args": ["--debug", "--clean", "deamon"]
+      "args": [
+        "--debug",
+        "--clean",
+        "sync",
+        "--from=2024-01-01",
+        "players",
+        "replays"
+      ]
     },
     {
       "name": "OBS client",
@@ -89,4 +77,4 @@
       "args": []
     }
   ]
-}
+}

.vscode/tasks.json

@@ -0,0 +1,11 @@
+{
+    "version": "2.0.0",
+    "tasks": [
+        {
+            "label": "Env File Lock",
+            "type": "shell",
+            "command": "powershell -Command 'conda env export --no-builds | findstr -v \"prefix\" > environment.yml'",
+            "problemMatcher": []
+        }
+    ]
+}

README.md

@@ -19,15 +19,15 @@ This is my personal research project to explore the latest in LLM based agents.
 
 Looking up past games when a new game is being played:
 
-![Example new game](archive/aicoach-scanner-example.png "New game started")
+![Example new game](assets/aicoach-scanner-example.png "New game started")
 
 Analyzing a replay after a game just finished:
 
-![Example replay](archive/aicoach-newreplay-manners.png "New Replay, discussing player's manners")
+![Example replay](assets/aicoach-newreplay-manners.png "New Replay, discussing player's manners")
 
 Answering arbitrary questions on SC2:
 
-![Example replay](archive/aicoach-hey-goat.png "Weighing in on the GOAT debate")
+![Example replay](assets/aicoach-hey-goat.png "Weighing in on the GOAT debate")
 
 ## Minimal Setup
 
@@ -63,41 +63,11 @@ The rest of the settings will be taken from `config.yml`.
 
 Secrets are configured with environment variables. Either provide them at runtime or put them in a dotenv file, like [.env.example](.env.example). (But copy to `.env` since the example file is ignored).
 
-### OpenAI
-
-Prerequisites:
-
-- Setup an OpenAI account and fund with credits
-- Create an OpenAI Assistant
-- Create an API key.
-
-Add your OpenAI organization, Assistant ID, and API key to the env variables, `AICOACH_ASSISTANT_ID`, `AICOACH_OPENAI_API_KEY`, `AICOACH_OPENAI_ORG_ID`.
-
-Note on cost: Long conversations can cost up to one dollar ($1.00) in OpenAI API usage. AICoach will not incur API costs until one of the wake events is triggered - see below.
-
-If you just want a database with your replays you can skip this step and the next or do it later.
-
-### Build and deploy assistant
-
-```sh
-> python build.py
-```
-
-to build the assistant. You should have a new file [aicoach/assistant.json](aicoach/assistant.json).
-
-```sh
-> python build.py --deploy
-```
-
-to deploy the assistant to OpenAI. Check on https://platform.openai.com/playground if the assistant is initialized with tools and instructions.
-
 ### Database
 
 Any MongoDB > 4.5 will do. Either setup one by yourself, or follow the instructions in [mongodb/](mongodb/README.md) on how to setup a local database for dev/testing.
 
-If you setup your own MongoDB, create a database, and add 3 collections `replays`, `replays.meta`, `replays.players`.
-
-Add the DB name to settings:
+If you setup your own MongoDB, create a database and add the DB name to settings:
 
 ```yaml
 # config.yourname.yml
@@ -123,35 +93,66 @@ Use the tool [repcli.py](repcli.py) to populate your DB with replays. The tools
 Usage: repcli.py [OPTIONS] COMMAND [ARGS]...
 
 Options:
-  --clean       Delete replays from instant-leave games
-  --debug       Print debug messages, including replay parser
-  --simulation  Run in simulation mode, don't actually insert to DB
-  --help        Show this message and exit.
+  --clean        Delete replays from instant-leave games
+  --debug        Print debug messages, including replay parser
+  --simulation   Run in simulation mode, don't actually insert to DB
+  -v, --verbose  Print verbose output
+  --help         Show this message and exit.
 
 Commands:
-  deamon  Monitor replay folder, add new replays to MongoDB
-  echo    Echo pretty-printed parsed replay data from a .SC2Replay file
-  sync    Sync replays from replay folder to MongoDB
+  echo   Echo pretty-printed parsed replay data from a .SC2Replay file
+  query  Query the DB for replays and players
+  sync   Sync replays and players from replay folder to MongoDB
 ```
 
 Run
 
 ```sh
-> python repcli.py --simulation sync --from=2024-01-01
+> python repcli.py --simulation sync players replays --from=2024-01-01 
 ```
-
-to read all 1v1 ladder replays from beginning of 2024. With the `--simulation` flag the replays will not actually be commited to DB. Remove the `--simulation` flag and run again to store all replay in DB.
+to read all 1v1 ladder replays from beginning of 2024, and add the replays and the players from the replays to the DB. With the `--simulation` flag the replays will not actually be commited to DB. Remove the `--simulation` flag and run again to store all replay in DB.
 
 The `replays` collection of the DB should now be populated with replay documents.
 
 See `python repcli.py sync --help` for more options. You can always repopulate the DB from replay files without destroying anything. AICoach does not change anything on the replay data in the DB.
 
+### OpenAI
+
+Prerequisites:
+
+- Setup an OpenAI account and fund with credits
+- Create an OpenAI Assistant
+- Create an API key.
+
+Add your OpenAI organization, Assistant ID, and API key to the env variables, `AICOACH_ASSISTANT_ID`, `AICOACH_OPENAI_API_KEY`, `AICOACH_OPENAI_ORG_ID`.
+
+Note on cost: Long conversations can cost up to one dollar ($1.00) in OpenAI API usage. Typically interactions stay below $0.10 however. AICoach will not incur API costs until one of the wake events is triggered - see below.
+
+If you just want a database with your replays you can skip this step and the next or do it later.
+
+### Build and deploy assistant
+
+```sh
+> python build.py
+```
+
+to build the assistant. You should have a new file [aicoach/assistant.json](aicoach/assistant.json).
+
+```sh
+> python build.py --deploy
+```
+
+to deploy the assistant to OpenAI. Check on https://platform.openai.com/playground if the assistant is initialized with tools and instructions.
+
+
 ### (Optional) Additional settings
 
 Configure a wake hotkey. On pressing this key (combination) AICoach will wake up and ask for input. Default: `ctrl+alt+w`.
 
 Configure student.emoji if you want to show a [different icon](./playground/emojis.txt) in the terminal output.
 
+You can disable interactions with the `interactive` flag. If off, AI coach will speak, but won't listen for input. 
+
 ```yaml
 # config.yourname.yml
 
@@ -162,6 +163,7 @@ student:
   emoji: ":woman_student:"
 db_name: "YOURDB"
 wake_key: "ctrl+alt+w"
+interactive: False
 ```
 
 ## Run AICoach

aicoach/aicoach.py

@@ -1,7 +1,7 @@
 import json
 import logging
 from datetime import datetime
-from typing import Callable, Dict, Generator
+from typing import Callable, Dict, Generator, Generic, Type, TypeVar
 
 from openai import APIError, AssistantEventHandler, OpenAI
 from openai.lib.streaming import AssistantStreamManager
@@ -17,6 +17,7 @@
     Run,
 )
 from openai.types.beta.threads.run import Usage
+from pydantic import BaseModel
 
 from config import config
 
@@ -27,6 +28,8 @@
 
 client = OpenAI(api_key=config.openai_api_key, organization=config.openai_org_id)
 
+TBaseModel = TypeVar("T", bound=BaseModel)
+
 
 class AICoach:
     threads: Dict[str, Thread] = {}
@@ -120,6 +123,29 @@ def get_response(self, message) -> str:
             buffer += response
         return buffer
 
+    def get_structured_response(self, message, schema: Type[TBaseModel]) -> TBaseModel:
+        """Get the structured response from the assistant for a given message."""
+        message = client.beta.threads.messages.create(
+            thread_id=self.thread.id,
+            role="user",
+            content=message,
+        )
+        new_run = client.beta.threads.runs.create_and_poll(
+            thread_id=self.thread.id,
+            assistant_id=self.assistant.id,
+            tools=[],  # structured output requires no tools
+            response_format={
+                "type": "json_schema",
+                "json_schema": {
+                    "name": schema.__name__,
+                    "schema": schema.model_json_schema(),
+                },
+            },
+        )
+
+        if new_run.status == "completed":
+            return schema(**json.loads(self.get_most_recent_message()))
+
     def chat(self, text) -> Generator[str, None, None]:
         """Send a message to the assistant and stream the response.
 
@@ -135,10 +161,13 @@ def chat(self, text) -> Generator[str, None, None]:
             assistant_id=self.assistant.id,
             additional_instructions=self.additional_instructions,
         ) as stream:
-
-            for event in stream:
-                for token in self._process_event(event):
-                    yield token
+            try:
+                for event in stream:
+                    for token in self._process_event(event):
+                        yield token
+            except APIError as e:
+                log.error(f"API Error: {e}")
+                yield ""
 
     def _process_event(self, event) -> Generator[str, None, None]:
         if isinstance(event, ThreadMessageDelta):
@@ -172,9 +201,6 @@ def _handle_tool_calls(self, run: Run) -> dict[str, str]:
     def _handle_tool_call(
         self, tool_call: RequiredActionFunctionToolCall
     ) -> tuple[str, str]:
-        if tool_call.type != "function":
-            return (None, None)
-
         args = json.loads(tool_call.function.arguments)
         name = tool_call.function.name
         log.info(f"Calling function {name} with args: {args}")

aicoach/aicoach_mock.py

@@ -1,15 +1,15 @@
 import random
 from threading import Thread
 from time import sleep
-from typing import Callable, Dict, Generator
+from typing import Callable, Dict, Generator, Type
 from uuid import uuid4
 
 import tiktoken
 from openai.types.beta.threads import Message, Text, TextContentBlock
 from openai.types.beta.threads.run import Usage
 from typing_extensions import override
 
-from .aicoach import AICoach
+from .aicoach import AICoach, TBaseModel
 
 data = [
     "The current supply count is not part of the game information provided. I can give insights based on replays on record but not from live games or current replays in progress. Would you like me to look up a recent replay instead?",
@@ -100,6 +100,10 @@ def chat(self, text: str) -> Generator[str, None, None]:
         for token in self.generate_stream():
             yield token
 
+    @override
+    def get_structured_response(self, message, schema: type[TBaseModel]) -> TBaseModel:
+        raise NotImplementedError
+
     @override
     def get_conversation(self) -> list[Message]:
         return [

aicoach/functions/LookupPlayer.py

@@ -1,9 +1,12 @@
-from .base import AIFunction
-from typing import Annotated
-import requests
 import logging
+from typing import Annotated
+
+import httpx
+
 from config import config
 
+from .base import AIFunction
+
 log = logging.getLogger(f"{config.name}.{__name__}")
 
 PROFILE_BASE = "https://starcraft2.com/en-us/profile/"
@@ -24,13 +27,14 @@ def LookupPlayer(
 
     Only call this function with valid toon handles in the format 2-S2-1-1849098.
     """
+    raise NotImplementedError("This function is not implemented yet.")
     summary = {}
     try:
         handle = toon_handle.replace("-S2-", "/").replace("-", "/")
 
         api_url = API_BASE + handle
 
-        with requests.Session() as s:
+        with httpx.Client() as s:
             r = s.get(api_url, timeout=10)
 
             if r.status_code != 200: