Switch to current version of OpenAI.

Introduces several changes with the main objective of enabling OpenAI's streaming API in the chatmastermind application. This allows for the retrieval of AI responses gradually as a stream, which can significantly improve the user experience in interactions that involve large result sets.

* Added tiktoken import in 'openai.py' and modifications to the OpenAI class to support streaming. This includes the addition of a new class OpenAIAnswer to handle streaming API responses.
* Modified request function in the OpenAI class: the stream=True flag is added to the openai.ChatCompletion.create method to enable streaming API.
* Modified 'question.py' to print the answer parts as they are streamed.
* Replaced the Answer class's string data type with a generator which supports str and Generator[str, None, None] data types. Modifications are made to the Answer class methods to handle both data types accordingly.
* Updated the tests in 'test_ais_openai.py' and 'test_message.py' to reflect and validate these changes.

README.md

@@ -65,6 +65,8 @@ cmm question [-t OTAGS]... [-k ATAGS]... [-x XTAGS]... [-o OUTTAGS]... [-A AI_ID
 * `-O, --overwrite`: Overwrite existing messages when repeating them
 * `-s, --source-text FILE`: Add content of a file to the query
 * `-S, --source-code FILE`: Add source code file content to the chat history
+* `-l, --location {cache,db,all}`: Use given location when building the chat history (default: 'db')
+* `-g, --glob GLOB`: Filter message files using the given glob pattern
 
 #### Hist
 
@@ -83,6 +85,8 @@ cmm hist [--print | --convert FORMAT] [-t OTAGS]... [-k ATAGS]... [-x XTAGS]...
 * `-S, --source-code-only`: Only print embedded source code
 * `-A, --answer SUBSTRING`: Filter for answer substring
 * `-Q, --question SUBSTRING`: Filter for question substring
+* `-l, --location {cache,db,all}`: Use given location when building the chat history (default: 'db')
+* `-g, --glob GLOB`: Filter message files using the given glob pattern
 
 #### Tags
 

chatmastermind/ais/openai.py

@@ -2,7 +2,8 @@
 Implements the OpenAI client classes and functions.
 """
 import openai
-from typing import Optional, Union
+import tiktoken
+from typing import Optional, Union, Generator
 from ..tags import Tag
 from ..message import Message, Answer
 from ..chat import Chat
@@ -12,6 +13,52 @@ from ..configuration import OpenAIConfig
 ChatType = list[dict[str, str]]
 
 
+class OpenAIAnswer:
+    def __init__(self,
+                 idx: int,
+                 streams: dict[int, 'OpenAIAnswer'],
+                 response: openai.ChatCompletion,
+                 tokens: Tokens,
+                 encoding: tiktoken.core.Encoding) -> None:
+        self.idx = idx
+        self.streams = streams
+        self.response = response
+        self.position: int = 0
+        self.encoding = encoding
+        self.data: list[str] = []
+        self.finished: bool = False
+        self.tokens = tokens
+
+    def stream(self) -> Generator[str, None, None]:
+        while True:
+            if not self.next():
+                continue
+            if len(self.data) <= self.position:
+                break
+            yield self.data[self.position]
+            self.position += 1
+
+    def next(self) -> bool:
+        if self.finished:
+            return True
+        try:
+            chunk = next(self.response)
+        except StopIteration:
+            self.finished = True
+        if not self.finished:
+            found_choice = False
+            for choice in chunk.choices:
+                if not choice.finish_reason:
+                    self.streams[choice.index].data.append(choice.delta.content)
+                    self.tokens.completion += len(self.encoding.encode(choice.delta.content))
+                    self.tokens.total = self.tokens.prompt + self.tokens.completion
+                if choice.index == self.idx:
+                    found_choice = True
+            if not found_choice:
+                return False
+        return True
+
+
 class OpenAI(AI):
     """
     The OpenAI AI client.
@@ -21,7 +68,10 @@ class OpenAI(AI):
         self.ID = config.ID
         self.name = config.name
         self.config = config
-        openai.api_key = config.api_key
+        self.client = openai.OpenAI(api_key=self.config.api_key)
+
+    def _completions(self, *args, **kw):  # type: ignore
+        return self.client.chat.completions.create(*args, **kw)
 
     def request(self,
                 question: Message,
@@ -33,39 +83,42 @@ class OpenAI(AI):
         chat history. The nr. of requested answers corresponds to the
         nr. of messages in the 'AIResponse'.
         """
-        oai_chat = self.openai_chat(chat, self.config.system, question)
+        self.encoding = tiktoken.encoding_for_model(self.config.model)
-        response = openai.ChatCompletion.create(
+        oai_chat, prompt_tokens = self.openai_chat(chat, self.config.system, question)
+        tokens: Tokens = Tokens(prompt_tokens, 0, prompt_tokens)
+        response = self._completions(
             model=self.config.model,
             messages=oai_chat,
             temperature=self.config.temperature,
             max_tokens=self.config.max_tokens,
             top_p=self.config.top_p,
             n=num_answers,
+            stream=True,
             frequency_penalty=self.config.frequency_penalty,
             presence_penalty=self.config.presence_penalty)
-        question.answer = Answer(response['choices'][0]['message']['content'])
+        streams: dict[int, OpenAIAnswer] = {}
+        for n in range(num_answers):
+            streams[n] = OpenAIAnswer(n, streams, response, tokens, self.encoding)
+        question.answer = Answer(streams[0].stream())
         question.tags = set(otags) if otags is not None else None
         question.ai = self.ID
         question.model = self.config.model
         answers: list[Message] = [question]
-        for choice in response['choices'][1:]:  # type: ignore
+        for idx in range(1, num_answers):
             answers.append(Message(question=question.question,
-                                   answer=Answer(choice['message']['content']),
+                                   answer=Answer(streams[idx].stream()),
                                    tags=otags,
                                    ai=self.ID,
                                    model=self.config.model))
-        return AIResponse(answers, Tokens(response['usage']['prompt_tokens'],
+        return AIResponse(answers, tokens)
-                                          response['usage']['completion_tokens'],
-                                          response['usage']['total_tokens']))
 
     def models(self) -> list[str]:
         """
         Return all models supported by this AI.
         """
         ret = []
-        for engine in sorted(openai.Engine.list()['data'], key=lambda x: x['id']):
+        for engine in sorted(self.client.models.list().data, key=lambda x: x.id):
-            if engine['ready']:
+            ret.append(engine.id)
-                ret.append(engine['id'])
         ret.sort()
         return ret
 
@@ -73,34 +126,30 @@ class OpenAI(AI):
         """
         Print all models supported by the current AI.
         """
-        not_ready = []
+        for model in self.models():
-        for engine in sorted(openai.Engine.list()['data'], key=lambda x: x['id']):
+            print(model)
-            if engine['ready']:
-                print(engine['id'])
-            else:
-                not_ready.append(engine['id'])
-        if len(not_ready) > 0:
-            print('\nNot ready: ' + ', '.join(not_ready))
 
     def openai_chat(self, chat: Chat, system: str,
-                    question: Optional[Message] = None) -> ChatType:
+                    question: Optional[Message] = None) -> tuple[ChatType, int]:
         """
         Create a chat history with system message in OpenAI format.
         Optionally append a new question.
         """
         oai_chat: ChatType = []
+        prompt_tokens: int = 0
 
-        def append(role: str, content: str) -> None:
+        def append(role: str, content: str) -> int:
             oai_chat.append({'role': role, 'content': content.replace("''", "'")})
+            return len(self.encoding.encode(', '.join(['role:', oai_chat[-1]['role'], 'content:', oai_chat[-1]['content']])))
 
-        append('system', system)
+        prompt_tokens += append('system', system)
         for message in chat.messages:
             if message.answer:
-                append('user', message.question)
+                prompt_tokens += append('user', message.question)
-                append('assistant', message.answer)
+                prompt_tokens += append('assistant', str(message.answer))
         if question:
-            append('user', question.question)
+            prompt_tokens += append('user', question.question)
-        return oai_chat
+        return oai_chat, prompt_tokens
 
     def tokens(self, data: Union[Message, Chat]) -> int:
         raise NotImplementedError

chatmastermind/chat.py

@@ -6,7 +6,8 @@ from pathlib import Path
 from pprint import PrettyPrinter
 from pydoc import pager
 from dataclasses import dataclass
-from typing import TypeVar, Type, Optional, Any, Callable, Literal, Union
+from enum import Enum
+from typing import TypeVar, Type, Optional, Any, Callable, Union
 from .configuration import default_config_file
 from .message import Message, MessageFilter, MessageError, MessageFormat, message_in, message_valid_formats
 from .tags import Tag
@@ -16,10 +17,17 @@ ChatDBInst = TypeVar('ChatDBInst', bound='ChatDB')
 
 db_next_file = '.next'
 ignored_files = [db_next_file, default_config_file]
-msg_location = Literal['mem', 'disk', 'cache', 'db', 'all']
 msg_suffix = Message.file_suffix_write
 
 
+class msg_location(Enum):
+    MEM = 'mem'
+    DISK = 'disk'
+    CACHE = 'cache'
+    DB = 'db'
+    ALL = 'all'
+
+
 class ChatError(Exception):
     pass
 
@@ -44,12 +52,12 @@ def read_dir(dir_path: Path,
     Parameters:
     * 'dir_path': source directory
     * 'glob': if specified, files will be filtered using 'path.glob()',
-      otherwise it uses 'path.iterdir()'.
+      otherwise it reads all files with the default message suffix
     * 'mfilter': use with 'Message.from_file()' to filter messages
       when reading them.
     """
     messages: list[Message] = []
-    file_iter = dir_path.glob(glob) if glob else dir_path.iterdir()
+    file_iter = dir_path.glob(glob) if glob else dir_path.glob(f'*{msg_suffix}')
     for file_path in sorted(file_iter):
         if (file_path.is_file()
            and file_path.name not in ignored_files  # noqa: W503
@@ -287,7 +295,7 @@ class ChatDB(Chat):
     # a MessageFilter that all messages must match (if given)
     mfilter: Optional[MessageFilter] = None
     # the glob pattern for all messages
-    glob: Optional[str] = None
+    glob: str = f'*{msg_suffix}'
     # message format (for writing)
     mformat: MessageFormat = Message.default_format
 
@@ -303,20 +311,28 @@ class ChatDB(Chat):
     def from_dir(cls: Type[ChatDBInst],
                  cache_path: Path,
                  db_path: Path,
-                 glob: Optional[str] = None,
+                 glob: str = f'*{msg_suffix}',
-                 mfilter: Optional[MessageFilter] = None) -> ChatDBInst:
+                 mfilter: Optional[MessageFilter] = None,
+                 loc: msg_location = msg_location.DB) -> ChatDBInst:
         """
         Create a 'ChatDB' instance from the given directory structure.
         Reads all messages from 'db_path' into the local message list.
         Parameters:
         * 'cache_path': path to the directory for temporary messages
         * 'db_path': path to the directory for persistent messages
-        * 'glob': if specified, files will be filtered using 'path.glob()',
+        * 'glob': if specified, files will be filtered using 'path.glob()'
-          otherwise it uses 'path.iterdir()'.
         * 'mfilter': use with 'Message.from_file()' to filter messages
           when reading them.
+        * 'loc': read messages from given location instead of 'db_path'
         """
-        messages = read_dir(db_path, glob, mfilter)
+        if loc == msg_location.MEM:
+            raise ChatError(f"Can't build ChatDB from message location '{loc}'")
+        messages: list[Message] = []
+        if loc in [msg_location.DB, msg_location.DISK, msg_location.ALL]:
+            messages.extend(read_dir(db_path, glob, mfilter))
+        if loc in [msg_location.CACHE, msg_location.DISK, msg_location.ALL]:
+            messages.extend(read_dir(cache_path, glob, mfilter))
+        messages.sort(key=lambda x: x.msg_id())
         return cls(messages, cache_path, db_path, mfilter, glob)
 
     @classmethod
@@ -386,7 +402,7 @@ class ChatDB(Chat):
     def msg_gather(self,
                    loc: msg_location,
                    require_file_path: bool = False,
-                   glob: Optional[str] = None,
+                   glob: str = f'*{msg_suffix}',
                    mfilter: Optional[MessageFilter] = None) -> list[Message]:
         """
         Gather and return messages from the given locations:
@@ -399,14 +415,14 @@ class ChatDB(Chat):
         If 'require_file_path' is True, return only files with a valid file_path.
         """
         loc_messages: list[Message] = []
-        if loc in ['mem', 'all']:
+        if loc in [msg_location.MEM, msg_location.ALL]:
             if require_file_path:
                 loc_messages += [m for m in self.messages if (m.file_path is not None and (mfilter is None or m.match(mfilter)))]
             else:
                 loc_messages += [m for m in self.messages if (mfilter is None or m.match(mfilter))]
-        if loc in ['cache', 'disk', 'all']:
+        if loc in [msg_location.CACHE, msg_location.DISK, msg_location.ALL]:
             loc_messages += read_dir(self.cache_path, glob=glob, mfilter=mfilter)
-        if loc in ['db', 'disk', 'all']:
+        if loc in [msg_location.DB, msg_location.DISK, msg_location.ALL]:
             loc_messages += read_dir(self.db_path, glob=glob, mfilter=mfilter)
         # remove_duplicates and sort the list
         unique_messages: list[Message] = []
@@ -422,7 +438,7 @@ class ChatDB(Chat):
 
     def msg_find(self,
                  msg_names: list[str],
-                 loc: msg_location = 'mem',
+                 loc: msg_location = msg_location.MEM,
                  ) -> list[Message]:
         """
         Search and return the messages with the given names. Names can either be filenames
@@ -440,7 +456,7 @@ class ChatDB(Chat):
         return [m for m in loc_messages
                 if any((m.file_path and self.msg_name_matches(m.file_path, mn)) for mn in msg_names)]
 
-    def msg_remove(self, msg_names: list[str], loc: msg_location = 'mem') -> None:
+    def msg_remove(self, msg_names: list[str], loc: msg_location = msg_location.MEM) -> None:
         """
         Remove the messages with the given names. Names can either be filenames
         (with or without suffix), full paths or Message.msg_id(). Also deletes the
@@ -452,7 +468,7 @@ class ChatDB(Chat):
         * 'db'   : messages in the DB directory
         * 'all'  : all messages ('mem' + 'disk')
         """
-        if loc != 'mem':
+        if loc != msg_location.MEM:
             # delete the message files first
             rm_messages = self.msg_find(msg_names, loc=loc)
             for m in rm_messages:
@@ -463,7 +479,7 @@ class ChatDB(Chat):
 
     def msg_latest(self,
                    mfilter: Optional[MessageFilter] = None,
-                   loc: msg_location = 'mem') -> Optional[Message]:
+                   loc: msg_location = msg_location.MEM) -> Optional[Message]:
         """
         Return the last added message (according to the file ID) that matches the given filter.
         Only consider messages with a valid file_path (except if loc is 'mem').
@@ -492,7 +508,7 @@ class ChatDB(Chat):
                     and message.file_path.parent.samefile(self.cache_path)  # noqa: W503
                     and message.file_path.exists())  # noqa: W503
         else:
-            return len(self.msg_find([message], loc='cache')) > 0
+            return len(self.msg_find([message], loc=msg_location.CACHE)) > 0
 
     def msg_in_db(self, message: Union[Message, str]) -> bool:
         """
@@ -504,9 +520,9 @@ class ChatDB(Chat):
                     and message.file_path.parent.samefile(self.db_path)  # noqa: W503
                     and message.file_path.exists())  # noqa: W503
         else:
-            return len(self.msg_find([message], loc='db')) > 0
+            return len(self.msg_find([message], loc=msg_location.DB)) > 0
 
-    def cache_read(self, glob: Optional[str] = None, mfilter: Optional[MessageFilter] = None) -> None:
+    def cache_read(self, glob: str = f'*{msg_suffix}', mfilter: Optional[MessageFilter] = None) -> None:
         """
         Read messages from the cache directory. New ones are added to the internal list,
         existing ones are replaced. A message is determined as 'existing' if a message
@@ -549,7 +565,7 @@ class ChatDB(Chat):
         self.messages += messages
         self.msg_sort()
 
-    def cache_clear(self, glob: Optional[str] = None) -> None:
+    def cache_clear(self, glob: str = f'*{msg_suffix}') -> None:
         """
         Delete all message files from the cache dir and remove them from the internal list.
         """
@@ -569,11 +585,11 @@ class ChatDB(Chat):
         self.cache_write([message])
         # remove the old one (if any)
         if old_path:
-            self.msg_remove([str(old_path)], loc='db')
+            self.msg_remove([str(old_path)], loc=msg_location.DB)
         # (re)add it to the internal list
         self.msg_add([message])
 
-    def db_read(self, glob: Optional[str] = None, mfilter: Optional[MessageFilter] = None) -> None:
+    def db_read(self, glob: str = f'*{msg_suffix}', mfilter: Optional[MessageFilter] = None) -> None:
         """
         Read messages from the DB directory. New ones are added to the internal list,
         existing ones are replaced. A message is determined as 'existing' if a message
@@ -628,6 +644,6 @@ class ChatDB(Chat):
         self.db_write([message])
         # remove the old one (if any)
         if old_path:
-            self.msg_remove([str(old_path)], loc='cache')
+            self.msg_remove([str(old_path)], loc=msg_location.CACHE)
         # (re)add it to the internal list
         self.msg_add([message])

chatmastermind/commands/hist.py

@@ -2,7 +2,7 @@ import sys
 import argparse
 from pathlib import Path
 from ..configuration import Config
-from ..chat import ChatDB
+from ..chat import ChatDB, msg_location
 from ..message import MessageFilter, Message
 
 
@@ -15,9 +15,10 @@ def convert_messages(args: argparse.Namespace, config: Config) -> None:
     ('.txt', '.yaml') to the latest default message file suffix ('.msg').
     """
     chat = ChatDB.from_dir(Path(config.cache),
-                           Path(config.db))
+                           Path(config.db),
+                           glob='*')
     # read all known message files
-    msgs = chat.msg_gather(loc='disk', glob='*.*')
+    msgs = chat.msg_gather(loc=msg_location.DISK, glob='*.*')
     # make a set of all message IDs
     msg_ids = set([m.msg_id() for m in msgs])
     # set requested format and write all messages
@@ -29,14 +30,14 @@ def convert_messages(args: argparse.Namespace, config: Config) -> None:
             m.file_path = m.file_path.with_suffix('')
     chat.msg_write(msgs)
     # read all messages with the current default suffix
-    msgs = chat.msg_gather(loc='disk', glob=f'*{msg_suffix}')
+    msgs = chat.msg_gather(loc=msg_location.DISK, glob=f'*{msg_suffix}')
     # make sure we converted all of the original messages
     for mid in msg_ids:
         if not any(mid == m.msg_id() for m in msgs):
             print(f"Message '{mid}' has not been found after conversion. Aborting.")
             sys.exit(1)
     # delete messages with old suffixes
-    msgs = chat.msg_gather(loc='disk', glob='*.*')
+    msgs = chat.msg_gather(loc=msg_location.DISK, glob='*.*')
     for m in msgs:
         if m.file_path and m.file_path.suffix != msg_suffix:
             m.rm_file()
@@ -55,7 +56,9 @@ def print_chat(args: argparse.Namespace, config: Config) -> None:
                             answer_contains=args.answer)
     chat = ChatDB.from_dir(Path(config.cache),
                            Path(config.db),
-                           mfilter=mfilter)
+                           mfilter=mfilter,
+                           loc=msg_location(args.location),
+                           glob=args.glob)
     chat.print(args.source_code_only,
                args.with_metadata,
                paged=not args.no_paging,

chatmastermind/commands/print.py

@@ -3,7 +3,7 @@ import argparse
 from pathlib import Path
 from ..configuration import Config
 from ..message import Message, MessageError
-from ..chat import ChatDB
+from ..chat import ChatDB, msg_location
 
 
 def print_message(message: Message, args: argparse.Namespace) -> None:
@@ -38,7 +38,7 @@ def print_cmd(args: argparse.Namespace, config: Config) -> None:
     # print latest message
     elif args.latest:
         chat = ChatDB.from_dir(Path(config.cache), Path(config.db))
-        latest = chat.msg_latest(loc='disk')
+        latest = chat.msg_latest(loc=msg_location.DISK)
         if not latest:
             print("No message found!")
             sys.exit(1)

chatmastermind/commands/question.py

@@ -4,7 +4,7 @@ from pathlib import Path
 from itertools import zip_longest
 from copy import deepcopy
 from ..configuration import Config
-from ..chat import ChatDB
+from ..chat import ChatDB, msg_location
 from ..message import Message, MessageFilter, MessageError, Question, source_code
 from ..ai_factory import create_ai
 from ..ai import AI, AIResponse
@@ -101,7 +101,7 @@ def create_message(chat: ChatDB, args: argparse.Namespace) -> Message:
         if code_file is not None and len(code_file) > 0:
             add_file_as_code(question_parts, code_file)
 
-    full_question = '\n\n'.join(question_parts)
+    full_question = '\n\n'.join([str(s) for s in question_parts])
 
     message = Message(question=Question(full_question),
                       tags=args.output_tags,
@@ -129,13 +129,16 @@ def make_request(ai: AI, chat: ChatDB, message: Message, args: argparse.Namespac
                                       args.output_tags)
     # only write the response messages to the cache,
     # don't add them to the internal list
-    chat.cache_write(response.messages)
     for idx, msg in enumerate(response.messages):
-        print(f"=== ANSWER {idx+1} ===")
+        print(f"=== ANSWER {idx+1} ===", flush=True)
-        print(msg.answer)
+        if msg.answer:
+            for piece in msg.answer:
+                print(piece, end='', flush=True)
+            print()
     if response.tokens:
         print("===============")
         print(response.tokens)
+    chat.cache_write(response.messages)
 
 
 def repeat_messages(messages: list[Message], chat: ChatDB, args: argparse.Namespace, config: Config) -> None:
@@ -186,7 +189,9 @@ def question_cmd(args: argparse.Namespace, config: Config) -> None:
                             tags_not=args.exclude_tags)
     chat = ChatDB.from_dir(cache_path=Path(config.cache),
                            db_path=Path(config.db),
-                           mfilter=mfilter)
+                           mfilter=mfilter,
+                           glob=args.glob,
+                           loc=msg_location(args.location))
     # if it's a new question, create and store it immediately
     if args.ask or args.create:
         message = create_message(chat, args)
@@ -202,14 +207,14 @@ def question_cmd(args: argparse.Namespace, config: Config) -> None:
         repeat_msgs: list[Message] = []
         # repeat latest message
         if len(args.repeat) == 0:
-            lmessage = chat.msg_latest(loc='cache')
+            lmessage = chat.msg_latest(loc=msg_location.CACHE)
             if lmessage is None:
                 print("No message found to repeat!")
                 sys.exit(1)
             repeat_msgs.append(lmessage)
         # repeat given message(s)
         else:
-            repeat_msgs = chat.msg_find(args.repeat, loc='disk')
+            repeat_msgs = chat.msg_find(args.repeat, loc=msg_location.DISK)
         repeat_messages(repeat_msgs, chat, args, config)
     # === PROCESS ===
     elif args.process is not None:

chatmastermind/main.py

@@ -14,6 +14,7 @@ from .commands.tags import tags_cmd
 from .commands.config import config_cmd
 from .commands.hist import hist_cmd
 from .commands.print import print_cmd
+from .chat import msg_location
 
 
 def tags_completer(prefix: str, parsed_args: Any, **kwargs: Any) -> list[str]:
@@ -65,6 +66,11 @@ def create_parser() -> argparse.ArgumentParser:
     question_group.add_argument('-c', '--create', nargs='+', help='Create a question', metavar='QUESTION')
     question_group.add_argument('-r', '--repeat', nargs='*', help='Repeat a question', metavar='MESSAGE')
     question_group.add_argument('-p', '--process', nargs='*', help='Process existing questions', metavar='MESSAGE')
+    question_cmd_parser.add_argument('-l', '--location',
+                                     choices=[x.value for x in msg_location if x not in [msg_location.MEM, msg_location.DISK]],
+                                     default='db',
+                                     help='Use given location when building the chat history (default: \'db\')')
+    question_cmd_parser.add_argument('-g', '--glob', help='Filter message files using the given glob pattern')
     question_cmd_parser.add_argument('-O', '--overwrite', help='Overwrite existing messages when repeating them',
                                      action='store_true')
     question_cmd_parser.add_argument('-s', '--source-text', nargs='+', help='Add content of a file to the query', metavar='FILE')
@@ -87,6 +93,11 @@ def create_parser() -> argparse.ArgumentParser:
     hist_cmd_parser.add_argument('-Q', '--question', help='Print only questions with given substring', metavar='SUBSTRING')
     hist_cmd_parser.add_argument('-d', '--tight', help='Print without message separators', action='store_true')
     hist_cmd_parser.add_argument('-P', '--no-paging', help='Print without paging', action='store_true')
+    hist_cmd_parser.add_argument('-l', '--location',
+                                 choices=[x.value for x in msg_location if x not in [msg_location.MEM, msg_location.DISK]],
+                                 default='db',
+                                 help='Use given location when building the chat history (default: \'db\')')
+    hist_cmd_parser.add_argument('-g', '--glob', help='Filter message files using the given glob pattern')
 
     # 'tags' command parser
     tags_cmd_parser = cmdparser.add_parser('tags',

chatmastermind/message.py

@@ -5,7 +5,9 @@ import pathlib
 import yaml
 import tempfile
 import shutil
+import io
 from typing import Type, TypeVar, ClassVar, Optional, Any, Union, Final, Literal, Iterable, Tuple
+from typing import Generator, Iterator
 from typing import get_args as typing_get_args
 from dataclasses import dataclass, asdict, field
 from .tags import Tag, TagLine, TagError, match_tags, rename_tags
@@ -49,7 +51,7 @@ def source_code(text: str, include_delims: bool = False) -> list[str]:
     code_lines: list[str] = []
     in_code_block = False
 
-    for line in text.split('\n'):
+    for line in str(text).split('\n'):
         if line.strip().startswith('```'):
             if include_delims:
                 code_lines.append(line)
@@ -142,30 +144,100 @@ class Answer(str):
     txt_header: ClassVar[str] = '==== ANSWER ===='
     yaml_key: ClassVar[str] = 'answer'
 
-    def __new__(cls: Type[AnswerInst], string: str) -> AnswerInst:
+    def __init__(self, data: Union[str, Generator[str, None, None]]) -> None:
+        # Indicator of whether all of data has been processed
+        self.is_exhausted: bool = False
+
+        # Initialize data
+        self.iterator: Iterator[str] = self._init_data(data)
+
+        # Set up the buffer to hold the 'Answer' content
+        self.buffer: io.StringIO = io.StringIO()
+
+    def _init_data(self, data: Union[str, Generator[str, None, None]]) -> Iterator[str]:
         """
-        Make sure the answer string does not contain the header as a whole line.
+        Process input data (either a string or a string generator)
         """
-        if cls.txt_header in string.split('\n'):
+        if isinstance(data, str):
-            raise MessageError(f"Answer '{string}' contains the header '{cls.txt_header}'")
+            yield data
-        instance = super().__new__(cls, string)
+        else:
-        return instance
+            yield from data
+
+    def __str__(self) -> str:
+        """
+        Output all content when converted into a string
+        """
+        # Ensure all data has been processed
+        for _ in self:
+            pass
+        # Return the 'Answer' content
+        return self.buffer.getvalue()
+
+    def __repr__(self) -> str:
+        return repr(str(self))
+
+    def __iter__(self) -> Generator[str, None, None]:
+        """
+        Allows the object to be iterable
+        """
+        # Generate content if not all data has been processed
+        if not self.is_exhausted:
+            yield from self.generator_iter()
+        else:
+            yield self.buffer.getvalue()
+
+    def generator_iter(self) -> Generator[str, None, None]:
+        """
+        Main generator method to process data
+        """
+        for piece in self.iterator:
+            # Write to buffer and yield piece for the iterator
+            self.buffer.write(piece)
+            yield piece
+        self.is_exhausted = True  # Set the flag that all data has been processed
+        # If the header occurs in the 'Answer' content, raise an error
+        if f'\n{self.txt_header}' in self.buffer.getvalue() or self.buffer.getvalue().startswith(self.txt_header):
+            raise MessageError(f"Answer {repr(self.buffer.getvalue())} contains the header {repr(Answer.txt_header)}")
+
+    def __eq__(self, other: object) -> bool:
+        """
+        Comparing the object to a string or another object
+        """
+        if isinstance(other, str):
+            return str(self) == other  # Compare the string value of this object to the other string
+        # Default behavior for comparing non-string objects
+        return super().__eq__(other)
+
+    def __hash__(self) -> int:
+        """
+        Generate a hash for the object based on its string representation.
+        """
+        return hash(str(self))
+
+    def __format__(self, format_spec: str) -> str:
+        """
+        Return a formatted version of the string as per the format specification.
+        """
+        return str(self).__format__(format_spec)
 
     @classmethod
     def from_list(cls: Type[AnswerInst], strings: list[str]) -> AnswerInst:
         """
-        Build Question from a list of strings. Make sure strings do not contain the header.
+        Build Answer from a list of strings. Make sure strings do not contain the header.
         """
-        if cls.txt_header in strings:
+        def _gen() -> Generator[str, None, None]:
-            raise MessageError(f"Question contains the header '{cls.txt_header}'")
+            if len(strings) > 0:
-        instance = super().__new__(cls, '\n'.join(strings).strip())
+                yield strings[0]
-        return instance
+                for s in strings[1:]:
+                    yield '\n'
+                    yield s
+        return cls(_gen())
 
     def source_code(self, include_delims: bool = False) -> list[str]:
         """
         Extract and return all source code sections.
         """
-        return source_code(self, include_delims)
+        return source_code(str(self), include_delims)
 
 
 class Question(str):
@@ -441,7 +513,7 @@ class Message():
         output.append(self.question)
         if self.answer:
             output.append(Answer.txt_header)
-            output.append(self.answer)
+            output.append(str(self.answer))
         return '\n'.join(output)
 
     def to_file(self, file_path: Optional[pathlib.Path]=None, mformat: MessageFormat = message_default_format) -> None:  # noqa: 11
@@ -491,7 +563,7 @@ class Message():
                 temp_fd.write(f'{ModelLine.from_model(self.model)}\n')
             temp_fd.write(f'{Question.txt_header}\n{self.question}\n')
             if self.answer:
-                temp_fd.write(f'{Answer.txt_header}\n{self.answer}\n')
+                temp_fd.write(f'{Answer.txt_header}\n{str(self.answer)}\n')
         shutil.move(temp_file_path, file_path)
 
     def __to_file_yaml(self, file_path: pathlib.Path) -> None:
@@ -560,7 +632,7 @@ class Message():
            or (mfilter.ai and (not self.ai or mfilter.ai != self.ai))  # noqa: W503
            or (mfilter.model and (not self.model or mfilter.model != self.model))  # noqa: W503
            or (mfilter.question_contains and mfilter.question_contains not in self.question)  # noqa: W503
-           or (mfilter.answer_contains and (not self.answer or mfilter.answer_contains not in self.answer))  # noqa: W503
+           or (mfilter.answer_contains and (not self.answer or mfilter.answer_contains not in str(self.answer)))  # noqa: W503
            or (mfilter.answer_state == 'available' and not self.answer)  # noqa: W503
            or (mfilter.ai_state == 'available' and not self.ai)  # noqa: W503
            or (mfilter.model_state == 'available' and not self.model)  # noqa: W503

requirements.txt

@@ -2,3 +2,4 @@ openai
 PyYAML
 argcomplete
 pytest
+tiktoken

tests/test_ais_openai.py

@@ -9,33 +9,32 @@ from chatmastermind.configuration import OpenAIConfig
 
 class OpenAITest(unittest.TestCase):
 
-    @mock.patch('openai.ChatCompletion.create')
+    @mock.patch('chatmastermind.ais.openai.OpenAI._completions')
     def test_request(self, mock_create: mock.MagicMock) -> None:
         # Create a test instance of OpenAI
         config = OpenAIConfig()
         openai = OpenAI(config)
 
         # Set up the mock response from openai.ChatCompletion.create
-        mock_response = {
+        class mock_obj:
-            'choices': [
+            pass
-                {
+        mock_chunk1 = mock_obj()
-                    'message': {
+        mock_chunk1.choices = [mock_obj(), mock_obj()]  # type: ignore
-                        'content': 'Answer 1'
+        mock_chunk1.choices[0].index = 0  # type: ignore
-                    }
+        mock_chunk1.choices[0].delta = mock_obj()  # type: ignore
-                },
+        mock_chunk1.choices[0].delta.content = 'Answer 1'  # type: ignore
-                {
+        mock_chunk1.choices[0].finish_reason = None  # type: ignore
-                    'message': {
+        mock_chunk1.choices[1].index = 1  # type: ignore
-                        'content': 'Answer 2'
+        mock_chunk1.choices[1].delta = mock_obj()  # type: ignore
-                    }
+        mock_chunk1.choices[1].delta.content = 'Answer 2'  # type: ignore
-                }
+        mock_chunk1.choices[1].finish_reason = None  # type: ignore
-            ],
+        mock_chunk2 = mock_obj()
-            'usage': {
+        mock_chunk2.choices = [mock_obj(), mock_obj()]  # type: ignore
-                'prompt_tokens': 10,
+        mock_chunk2.choices[0].index = 0  # type: ignore
-                'completion_tokens': 20,
+        mock_chunk2.choices[0].finish_reason = 'stop'  # type: ignore
-                'total_tokens': 30
+        mock_chunk2.choices[1].index = 1  # type: ignore
-            }
+        mock_chunk2.choices[1].finish_reason = 'stop'  # type: ignore
-        }
+        mock_create.return_value = iter([mock_chunk1, mock_chunk2])
-        mock_create.return_value = mock_response
 
         # Create test data
         question = Message(Question('Question'))
@@ -57,9 +56,9 @@ class OpenAITest(unittest.TestCase):
         self.assertIsNotNone(response.tokens)
         self.assertIsInstance(response.tokens, Tokens)
         assert response.tokens
-        self.assertEqual(response.tokens.prompt, 10)
+        self.assertEqual(response.tokens.prompt, 53)
-        self.assertEqual(response.tokens.completion, 20)
+        self.assertEqual(response.tokens.completion, 6)
-        self.assertEqual(response.tokens.total, 30)
+        self.assertEqual(response.tokens.total, 59)
 
         # Assert the mock call to openai.ChatCompletion.create
         mock_create.assert_called_once_with(
@@ -76,6 +75,7 @@ class OpenAITest(unittest.TestCase):
             max_tokens=config.max_tokens,
             top_p=config.top_p,
             n=2,
+            stream=True,
             frequency_penalty=config.frequency_penalty,
             presence_penalty=config.presence_penalty
         )

tests/test_chat.py

@@ -7,7 +7,7 @@ from io import StringIO
 from unittest.mock import patch
 from chatmastermind.tags import TagLine
 from chatmastermind.message import Message, Question, Answer, Tag, MessageFilter
-from chatmastermind.chat import Chat, ChatDB, ChatError
+from chatmastermind.chat import Chat, ChatDB, ChatError, msg_location
 
 
 msg_suffix: str = Message.file_suffix_write
@@ -240,7 +240,8 @@ class TestChatDB(TestChatBase):
         msg_to_file_force_suffix(duplicate_message)
         with self.assertRaises(ChatError) as cm:
             ChatDB.from_dir(pathlib.Path(self.cache_path.name),
-                            pathlib.Path(self.db_path.name))
+                            pathlib.Path(self.db_path.name),
+                            glob='*')
         self.assertEqual(str(cm.exception), "Validation failed")
 
     def test_file_path_ID_exists(self) -> None:
@@ -595,92 +596,92 @@ class TestChatDB(TestChatBase):
         chat_db = ChatDB.from_dir(pathlib.Path(self.cache_path.name),
                                   pathlib.Path(self.db_path.name))
         # search for a DB file in memory
-        self.assertEqual(chat_db.msg_find([str(self.message1.file_path)], loc='mem'), [self.message1])
+        self.assertEqual(chat_db.msg_find([str(self.message1.file_path)], loc=msg_location.MEM), [self.message1])
-        self.assertEqual(chat_db.msg_find([self.message1.msg_id()], loc='mem'), [self.message1])
+        self.assertEqual(chat_db.msg_find([self.message1.msg_id()], loc=msg_location.MEM), [self.message1])
-        self.assertEqual(chat_db.msg_find(['0001.msg'], loc='mem'), [self.message1])
+        self.assertEqual(chat_db.msg_find(['0001.msg'], loc=msg_location.MEM), [self.message1])
-        self.assertEqual(chat_db.msg_find(['0001'], loc='mem'), [self.message1])
+        self.assertEqual(chat_db.msg_find(['0001'], loc=msg_location.MEM), [self.message1])
         # and on disk
-        self.assertEqual(chat_db.msg_find([str(self.message2.file_path)], loc='db'), [self.message2])
+        self.assertEqual(chat_db.msg_find([str(self.message2.file_path)], loc=msg_location.DB), [self.message2])
-        self.assertEqual(chat_db.msg_find([self.message2.msg_id()], loc='db'), [self.message2])
+        self.assertEqual(chat_db.msg_find([self.message2.msg_id()], loc=msg_location.DB), [self.message2])
-        self.assertEqual(chat_db.msg_find(['0002.msg'], loc='db'), [self.message2])
+        self.assertEqual(chat_db.msg_find(['0002.msg'], loc=msg_location.DB), [self.message2])
-        self.assertEqual(chat_db.msg_find(['0002'], loc='db'), [self.message2])
+        self.assertEqual(chat_db.msg_find(['0002'], loc=msg_location.DB), [self.message2])
         # now search the cache -> expect empty result
-        self.assertEqual(chat_db.msg_find([str(self.message3.file_path)], loc='cache'), [])
+        self.assertEqual(chat_db.msg_find([str(self.message3.file_path)], loc=msg_location.CACHE), [])
-        self.assertEqual(chat_db.msg_find([self.message3.msg_id()], loc='cache'), [])
+        self.assertEqual(chat_db.msg_find([self.message3.msg_id()], loc=msg_location.CACHE), [])
-        self.assertEqual(chat_db.msg_find(['0003.msg'], loc='cache'), [])
+        self.assertEqual(chat_db.msg_find(['0003.msg'], loc=msg_location.CACHE), [])
-        self.assertEqual(chat_db.msg_find(['0003'], loc='cache'), [])
+        self.assertEqual(chat_db.msg_find(['0003'], loc=msg_location.CACHE), [])
         # search for multiple messages
         # -> search one twice, expect result to be unique
         search_names = ['0001', '0002.msg', self.message3.msg_id(), str(self.message3.file_path)]
         expected_result = [self.message1, self.message2, self.message3]
-        result = chat_db.msg_find(search_names, loc='all')
+        result = chat_db.msg_find(search_names, loc=msg_location.ALL)
         self.assert_messages_equal(result, expected_result)
 
     def test_msg_latest(self) -> None:
         chat_db = ChatDB.from_dir(pathlib.Path(self.cache_path.name),
                                   pathlib.Path(self.db_path.name))
-        self.assertEqual(chat_db.msg_latest(loc='mem'), self.message4)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.MEM), self.message4)
-        self.assertEqual(chat_db.msg_latest(loc='db'), self.message4)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.DB), self.message4)
-        self.assertEqual(chat_db.msg_latest(loc='disk'), self.message4)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.DISK), self.message4)
-        self.assertEqual(chat_db.msg_latest(loc='all'), self.message4)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.ALL), self.message4)
         # the cache is currently empty:
-        self.assertIsNone(chat_db.msg_latest(loc='cache'))
+        self.assertIsNone(chat_db.msg_latest(loc=msg_location.CACHE))
         # add new messages to the cache dir
         new_message = Message(question=Question("New Question"),
                               answer=Answer("New Answer"))
         chat_db.cache_add([new_message])
-        self.assertEqual(chat_db.msg_latest(loc='cache'), new_message)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.CACHE), new_message)
-        self.assertEqual(chat_db.msg_latest(loc='mem'), new_message)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.MEM), new_message)
-        self.assertEqual(chat_db.msg_latest(loc='disk'), new_message)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.DISK), new_message)
-        self.assertEqual(chat_db.msg_latest(loc='all'), new_message)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.ALL), new_message)
         # the DB does not contain the new message
-        self.assertEqual(chat_db.msg_latest(loc='db'), self.message4)
+        self.assertEqual(chat_db.msg_latest(loc=msg_location.DB), self.message4)
 
     def test_msg_gather(self) -> None:
         chat_db = ChatDB.from_dir(pathlib.Path(self.cache_path.name),
                                   pathlib.Path(self.db_path.name))
         all_messages = [self.message1, self.message2, self.message3, self.message4]
-        self.assert_messages_equal(chat_db.msg_gather(loc='all'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.ALL), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='db'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DB), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='mem'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.MEM), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='disk'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DISK), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='cache'), [])
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.CACHE), [])
         # add a new message, but only to the internal list
         new_message = Message(Question("What?"))
         all_messages_mem = all_messages + [new_message]
         chat_db.msg_add([new_message])
-        self.assert_messages_equal(chat_db.msg_gather(loc='mem'), all_messages_mem)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.MEM), all_messages_mem)
-        self.assert_messages_equal(chat_db.msg_gather(loc='all'), all_messages_mem)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.ALL), all_messages_mem)
         # the nr. of messages on disk did not change -> expect old result
-        self.assert_messages_equal(chat_db.msg_gather(loc='db'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DB), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='disk'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DISK), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='cache'), [])
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.CACHE), [])
         # test with MessageFilter
-        self.assert_messages_equal(chat_db.msg_gather(loc='all', mfilter=MessageFilter(tags_or={Tag('tag1')})),
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.ALL, mfilter=MessageFilter(tags_or={Tag('tag1')})),
                                    [self.message1])
-        self.assert_messages_equal(chat_db.msg_gather(loc='disk', mfilter=MessageFilter(tags_or={Tag('tag2')})),
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DISK, mfilter=MessageFilter(tags_or={Tag('tag2')})),
                                    [self.message2])
-        self.assert_messages_equal(chat_db.msg_gather(loc='cache', mfilter=MessageFilter(tags_or={Tag('tag3')})),
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.CACHE, mfilter=MessageFilter(tags_or={Tag('tag3')})),
                                    [])
-        self.assert_messages_equal(chat_db.msg_gather(loc='mem', mfilter=MessageFilter(question_contains="What")),
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.MEM, mfilter=MessageFilter(question_contains="What")),
                                    [new_message])
 
     def test_msg_move_and_gather(self) -> None:
         chat_db = ChatDB.from_dir(pathlib.Path(self.cache_path.name),
                                   pathlib.Path(self.db_path.name))
         all_messages = [self.message1, self.message2, self.message3, self.message4]
-        self.assert_messages_equal(chat_db.msg_gather(loc='db'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DB), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='cache'), [])
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.CACHE), [])
         # move first message to the cache
         chat_db.cache_move(self.message1)
-        self.assert_messages_equal(chat_db.msg_gather(loc='cache'), [self.message1])
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.CACHE), [self.message1])
         self.assertEqual(self.message1.file_path.parent, pathlib.Path(self.cache_path.name))  # type: ignore [union-attr]
-        self.assert_messages_equal(chat_db.msg_gather(loc='db'), [self.message2, self.message3, self.message4])
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DB), [self.message2, self.message3, self.message4])
-        self.assert_messages_equal(chat_db.msg_gather(loc='all'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.ALL), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='disk'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DISK), all_messages)
-        self.assert_messages_equal(chat_db.msg_gather(loc='mem'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.MEM), all_messages)
         # now move first message back to the DB
         chat_db.db_move(self.message1)
-        self.assert_messages_equal(chat_db.msg_gather(loc='cache'), [])
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.CACHE), [])
         self.assertEqual(self.message1.file_path.parent, pathlib.Path(self.db_path.name))  # type: ignore [union-attr]
-        self.assert_messages_equal(chat_db.msg_gather(loc='db'), all_messages)
+        self.assert_messages_equal(chat_db.msg_gather(loc=msg_location.DB), all_messages)

tests/test_hist_cmd.py

@@ -4,7 +4,7 @@ import tempfile
 import yaml
 from pathlib import Path
 from chatmastermind.message import Message, Question
-from chatmastermind.chat import ChatDB, ChatError
+from chatmastermind.chat import ChatDB, ChatError, msg_location
 from chatmastermind.configuration import Config
 from chatmastermind.commands.hist import convert_messages
 
@@ -41,7 +41,7 @@ class TestConvertMessages(unittest.TestCase):
     def test_convert_messages(self) -> None:
         self.args.convert = 'yaml'
         convert_messages(self.args, self.config)
-        msgs = self.chat.msg_gather(loc='disk', glob='*.*')
+        msgs = self.chat.msg_gather(loc=msg_location.DISK, glob='*.*')
         # Check if the number of messages is the same as before
         self.assertEqual(len(msgs), len(self.messages))
         # Check if all messages have the requested suffix

tests/test_message.py

@@ -91,7 +91,7 @@ class QuestionTestCase(unittest.TestCase):
 class AnswerTestCase(unittest.TestCase):
     def test_answer_with_header(self) -> None:
         with self.assertRaises(MessageError):
-            Answer(f"{Answer.txt_header}\nno")
+            str(Answer(f"{Answer.txt_header}\nno"))
 
     def test_answer_with_legal_header(self) -> None:
         answer = Answer(f"This is a line contaning '{Answer.txt_header}'\nIt is what it is.")

tests/test_question_cmd.py

@@ -9,7 +9,7 @@ from chatmastermind.configuration import Config
 from chatmastermind.commands.question import create_message, question_cmd
 from chatmastermind.tags import Tag
 from chatmastermind.message import Message, Question, Answer
-from chatmastermind.chat import Chat, ChatDB
+from chatmastermind.chat import Chat, ChatDB, msg_location
 from chatmastermind.ai import AIError
 from .test_common import TestWithFakeAI
 
@@ -234,6 +234,8 @@ class TestQuestionCmd(TestWithFakeAI):
         # create a mock argparse.Namespace
         self.args = argparse.Namespace(
             ask=['What is the meaning of life?'],
+            glob=None,
+            location='db',
             num_answers=1,
             output_tags=['science'],
             AI='FakeAI',
@@ -279,7 +281,7 @@ class TestQuestionCmdAsk(TestQuestionCmd):
         # check for the expected message files
         chat = ChatDB.from_dir(Path(self.cache_dir.name),
                                Path(self.db_dir.name))
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         self.assertEqual(len(self.message_list(self.cache_dir)), 1)
         self.assert_msgs_equal_except_file_path(cached_msg, expected_responses)
 
@@ -337,7 +339,7 @@ class TestQuestionCmdAsk(TestQuestionCmd):
         # check for the expected message files
         chat = ChatDB.from_dir(Path(self.cache_dir.name),
                                Path(self.db_dir.name))
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         self.assertEqual(len(self.message_list(self.cache_dir)), 1)
         self.assert_msgs_equal_except_file_path(cached_msg, [expected_question])
 
@@ -375,7 +377,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
         # we expect the original message + the one with the new response
         expected_responses = [message] + [expected_response]
         question_cmd(self.args, self.config)
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         print(self.message_list(self.cache_dir))
         self.assertEqual(len(self.message_list(self.cache_dir)), 2)
         self.assert_msgs_equal_except_file_path(cached_msg, expected_responses)
@@ -396,7 +398,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                           model=self.args.model,
                           file_path=Path(self.cache_dir.name) / f'0001{msg_suffix}')
         chat.msg_write([message])
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         assert cached_msg[0].file_path
         cached_msg_file_id = cached_msg[0].file_path.stem
 
@@ -412,7 +414,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                                     tags=message.tags,
                                     file_path=Path('<NOT COMPARED>'))
         question_cmd(self.args, self.config)
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         self.assertEqual(len(self.message_list(self.cache_dir)), 1)
         self.assert_msgs_equal_except_file_path(cached_msg, [expected_response])
         # also check that the file ID has not been changed
@@ -435,7 +437,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                           model=self.args.model,
                           file_path=Path(self.cache_dir.name) / f'0001{msg_suffix}')
         chat.msg_write([message])
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         assert cached_msg[0].file_path
         cached_msg_file_id = cached_msg[0].file_path.stem
 
@@ -452,7 +454,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                                     tags=message.tags,
                                     file_path=Path('<NOT COMPARED>'))
         question_cmd(self.args, self.config)
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         self.assertEqual(len(self.message_list(self.cache_dir)), 1)
         self.assert_msgs_equal_except_file_path(cached_msg, [expected_response])
         # also check that the file ID has not been changed
@@ -475,7 +477,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                           model=self.args.model,
                           file_path=Path(self.cache_dir.name) / f'0001{msg_suffix}')
         chat.msg_write([message])
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         assert cached_msg[0].file_path
 
         # repeat the last question with new arguments (without overwriting)
@@ -493,7 +495,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                                         tags={Tag('newtag')},
                                         file_path=Path('<NOT COMPARED>'))
         question_cmd(self.args, self.config)
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         self.assertEqual(len(self.message_list(self.cache_dir)), 2)
         self.assert_msgs_equal_except_file_path(cached_msg, [message] + [new_expected_response])
 
@@ -513,7 +515,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                           model=self.args.model,
                           file_path=Path(self.cache_dir.name) / f'0001{msg_suffix}')
         chat.msg_write([message])
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         assert cached_msg[0].file_path
 
         # repeat the last question with new arguments
@@ -530,7 +532,7 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
                                         tags={Tag('newtag')},
                                         file_path=Path('<NOT COMPARED>'))
         question_cmd(self.args, self.config)
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         self.assertEqual(len(self.message_list(self.cache_dir)), 1)
         self.assert_msgs_equal_except_file_path(cached_msg, [new_expected_response])
 
@@ -586,8 +588,8 @@ class TestQuestionCmdRepeat(TestQuestionCmd):
         self.assertEqual(len(self.message_list(self.cache_dir)), 4)
         self.assertEqual(len(self.message_list(self.db_dir)), 1)
         expected_cache_messages = [expected_responses[0], message2, expected_responses[1], expected_responses[2]]
-        cached_msg = chat.msg_gather(loc='cache')
+        cached_msg = chat.msg_gather(loc=msg_location.CACHE)
         self.assert_msgs_equal_except_file_path(cached_msg, expected_cache_messages)
         # check that the DB message has not been modified at all
-        db_msg = chat.msg_gather(loc='db')
+        db_msg = chat.msg_gather(loc=msg_location.DB)
         self.assert_msgs_all_equal(db_msg, [message3])