V0.97.4

README.md

@@ -1,11 +1,11 @@
 ## Why `cecli`?
 
-`cecli` (probably pronounced like "Cecily", aka `aider-ce`) is a community-driven fork of the [Aider](https://cecli.dev/) AI pair programming tool.
-Aider is a fantastic piece of software with a wonderful community but it has been painfully slow in receiving updates in the quickly evolving AI tooling space.
+`cecli` (probably pronounced like "Cecily") is yet another cli agent crafted for extensibility and customization. Originally a fork of the [Aider](https://cecli.dev/) AI pair programming tool, we aim to make agentic coding as maximally effective as it can be based on the growing capabilities of large language models.
 
-We aim to foster an open, collaborative ecosystem where new features, experiments, and improvements can be developed and shared rapidly. We believe in genuine FOSS principles and actively welcome contributors of all skill levels.
+We aim to foster an open, collaborative ecosystem where new features, experiments, and improvements can be developed and shared rapidly. We believe in the principles of FOSS and actively welcome contributors of all skill levels.
 
 If you are looking for bleeding-edge features or want to get your hands dirty with the internals of an AI coding agent, here's your sign.
+
 LLMs are a part of our lives from here on out so join us in learning about and crafting the future.
 
 ### Links

cecli/__init__.py

@@ -1,6 +1,6 @@
 from packaging import version
 
-__version__ = "0.97.3.dev"
+__version__ = "0.97.4.dev"
 safe_version = __version__
 
 try:

cecli/args.py

@@ -260,6 +260,12 @@ def get_parser(default_config_files, git_root):
         default=3,
         help="Maximum number of retries a model gets on malformed outputs (default: 3)",
     )
+    group.add_argument(
+        "--cost-limit",
+        type=float,
+        default=None,
+        help="Cost limit per session, exceeding this forces prompt confirmation (default: None)",
+    )
     group.add_argument(
         "--file-diffs",
         action=argparse.BooleanOptionalAction,

cecli/coders/agent_coder.py

@@ -90,9 +90,13 @@ def __init__(self, *args, **kwargs):
         self.skip_cli_confirmations = False
         self.agent_finished = False
         self.agent_config = self._get_agent_config()
+        self._setup_agent()
         ToolRegistry.build_registry(agent_config=self.agent_config)
         super().__init__(*args, **kwargs)
 
+    def _setup_agent(self):
+        os.makedirs(".cecli/workspace", exist_ok=True)
+
     def _get_agent_config(self):
         """
         Parse and return agent configuration from args.agent_config.

cecli/coders/architect_coder.py

@@ -1,7 +1,7 @@
 import asyncio
 
 from ..commands import SwitchCoderSignal
-from ..helpers.conversation import ConversationManager, MessageTag
+from ..helpers.conversation import ConversationManager
 from .ask_coder import AskCoder
 from .base_coder import Coder
 
@@ -61,9 +61,7 @@ async def reply_completed(self):
         editor_coder = await Coder.create(**new_kwargs)
 
         # Re-initialize ConversationManager with editor coder
-        ConversationManager.initialize(
-            editor_coder, reset=True, reformat=True, preserve_tags=[MessageTag.DONE, MessageTag.CUR]
-        )
+        ConversationManager.initialize(editor_coder, reset=True, reformat=True, preserve_tags=True)
 
         if self.verbose:
             editor_coder.show_announcements()
@@ -84,7 +82,7 @@ async def reply_completed(self):
                 original_coder or self,
                 reset=True,
                 reformat=True,
-                preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                preserve_tags=True,
             )
 
             self.total_cost = editor_coder.total_cost
@@ -96,7 +94,7 @@ async def reply_completed(self):
                 original_coder or self,
                 reset=True,
                 reformat=True,
-                preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                preserve_tags=True,
             )
 
         raise SwitchCoderSignal(main_model=self.main_model, edit_format="architect")

cecli/coders/base_coder.py

@@ -32,15 +32,15 @@
 
 import httpx
 from litellm import experimental_mcp_client
-from litellm.types.utils import ChatCompletionMessageToolCall, Function, ModelResponse
+from litellm.types.utils import ModelResponse
 from prompt_toolkit.patch_stdout import patch_stdout
 from rich.console import Console
 
 import cecli.prompts.utils.system as prompts
 from cecli import __version__, models, urls, utils
 from cecli.commands import Commands, SwitchCoderSignal
 from cecli.exceptions import LiteLLMExceptions
-from cecli.helpers import command_parser, coroutines, nested
+from cecli.helpers import command_parser, coroutines, nested, responses
 from cecli.helpers.conversation import (
     ConversationChunks,
     ConversationManager,
@@ -157,6 +157,7 @@ class Coder:
     last_user_message = ""
     uuid = ""
     model_kwargs = {}
+    cost_multiplier = 1
 
     # Task coordination state variables
     input_running = False
@@ -211,34 +212,14 @@ async def create(
 
             use_kwargs = dict(from_coder.original_kwargs)  # copy orig kwargs
 
-            # If the edit format changes, we can't leave old ASSISTANT
-            # messages in the chat history. The old edit format will
-            # confused the new LLM. It may try and imitate it, disobeying
-            # the system prompt.
-            # Get DONE messages from ConversationManager
-            done_messages = ConversationManager.get_messages_dict(MessageTag.DONE)
-            if edit_format != from_coder.edit_format and done_messages and summarize_from_coder:
-                try:
-                    io.tool_warning("Summarizing messages, please wait...")
-                    done_messages = await from_coder.summarizer.summarize_all(done_messages)
-                except (KeyboardInterrupt, ValueError):
-                    # If summarization fails, keep the original messages and warn the user
-                    io.tool_warning(
-                        "Chat history summarization failed, continuing with full history"
-                    )
-
-            # Bring along context from the old Coder
-            # Get CUR messages from ConversationManager
-            cur_messages = ConversationManager.get_messages_dict(MessageTag.CUR)
-
             update = dict(
                 fnames=list(from_coder.abs_fnames),
                 read_only_fnames=list(from_coder.abs_read_only_fnames),  # Copy read-only files
                 read_only_stubs_fnames=list(
                     from_coder.abs_read_only_stubs_fnames
                 ),  # Copy read-only stubs
-                done_messages=done_messages,
-                cur_messages=cur_messages,
+                done_messages=[],
+                cur_messages=[],
                 coder_commit_hashes=from_coder.coder_commit_hashes,
                 commands=from_coder.commands.clone(),
                 total_cost=from_coder.total_cost,
@@ -415,22 +396,6 @@ def __init__(
         self.add_gitignore_files = add_gitignore_files
         self.abs_read_only_stubs_fnames = set()
 
-        # Always use ConversationManager as the source of truth
-        # Add any provided messages to ConversationManager
-        if done_messages:
-            for msg in done_messages:
-                ConversationManager.add_message(
-                    message_dict=msg,
-                    tag=MessageTag.DONE,
-                )
-
-        if cur_messages:
-            for msg in cur_messages:
-                ConversationManager.add_message(
-                    message_dict=msg,
-                    tag=MessageTag.CUR,
-                )
-
         self.io = io
         self.io.coder = weakref.ref(self)
 
@@ -1616,6 +1581,18 @@ async def run_one(self, user_message, preproc):
             self.reflected_message = None
             self.tool_reflection = False
 
+            if float(self.total_cost) > self.cost_multiplier * (
+                nested.getter(self.args, "cost_limit", float("inf")) or float("inf")
+            ):
+                if await self.io.confirm_ask(
+                    "You have reached your configured cost limit. Continue?",
+                    group_response="Cost Limit",
+                    explicit_yes_required=True,
+                ):
+                    Coder.cost_multiplier += 1
+                else:
+                    return
+
             async for _ in self.send_message(message):
                 pass
 
@@ -2405,7 +2382,7 @@ async def send_message(self, inp):
                 force=True,  # Force update existing message
             )
 
-        if edited and self.auto_test:
+        if edited and self.auto_test and self.test_cmd:
             test_errors = await self.commands.execute("test", self.test_cmd)
             self.test_outcome = not test_errors
             if test_errors:
@@ -3334,66 +3311,16 @@ def consolidate_chunks(self):
         # If no native tool calls, check if the content contains JSON tool calls
         # This handles models that write JSON in text instead of using native calling
         if not self.partial_response_tool_calls and self.partial_response_content:
-            try:
-                # Simple extraction of JSON-like structures that look like tool calls
-                # Only look for tool calls if it looks like JSON
-                if "{" in self.partial_response_content or "[" in self.partial_response_content:
-                    json_chunks = utils.split_concatenated_json(self.partial_response_content)
-                    extracted_calls = []
-                    chunk_index = 0
-
-                    for chunk in json_chunks:
-                        chunk_index += 1
-                        try:
-                            json_obj = json.loads(chunk)
-                            if (
-                                isinstance(json_obj, dict)
-                                and "name" in json_obj
-                                and "arguments" in json_obj
-                            ):
-                                # Create a Pydantic model for the tool call
-                                function_obj = Function(
-                                    name=json_obj["name"],
-                                    arguments=(
-                                        json.dumps(json_obj["arguments"])
-                                        if isinstance(json_obj["arguments"], (dict, list))
-                                        else str(json_obj["arguments"])
-                                    ),
-                                )
-                                tool_call_obj = ChatCompletionMessageToolCall(
-                                    type="function",
-                                    function=function_obj,
-                                    id=f"call_{len(extracted_calls)}_{int(time.time())}_{chunk_index}",
-                                )
-                                extracted_calls.append(tool_call_obj)
-                            elif isinstance(json_obj, list):
-                                for item in json_obj:
-                                    if (
-                                        isinstance(item, dict)
-                                        and "name" in item
-                                        and "arguments" in item
-                                    ):
-                                        function_obj = Function(
-                                            name=item["name"],
-                                            arguments=(
-                                                json.dumps(item["arguments"])
-                                                if isinstance(item["arguments"], (dict, list))
-                                                else str(item["arguments"])
-                                            ),
-                                        )
-                                        tool_call_obj = ChatCompletionMessageToolCall(
-                                            type="function",
-                                            function=function_obj,
-                                            id=f"call_{len(extracted_calls)}_{int(time.time())}_{chunk_index}",
-                                        )
-                                        extracted_calls.append(tool_call_obj)
-                        except json.JSONDecodeError:
-                            continue
-
-                    if extracted_calls:
-                        self.partial_response_tool_calls = extracted_calls
-            except Exception:
-                pass
+            extracted_calls = responses.extract_tools_from_content_json(
+                self.partial_response_content
+            )
+            if not extracted_calls:
+                extracted_calls = responses.extract_tools_from_content_xml(
+                    self.partial_response_content
+                )
+
+            if extracted_calls:
+                self.partial_response_tool_calls = extracted_calls
 
         return response, func_err, content_err
 

cecli/commands/agent_model.py

@@ -3,7 +3,7 @@
 import cecli.models as models
 from cecli.commands.utils.base_command import BaseCommand
 from cecli.commands.utils.helpers import format_command_result
-from cecli.helpers.conversation import ConversationManager, MessageTag
+from cecli.helpers.conversation import ConversationManager
 
 
 class AgentModelCommand(BaseCommand):
@@ -69,7 +69,7 @@ async def execute(cls, io, coder, args, **kwargs):
                 temp_coder,
                 reset=True,
                 reformat=True,
-                preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                preserve_tags=True,
             )
 
             verbose = kwargs.get("verbose", False)
@@ -86,7 +86,7 @@ async def execute(cls, io, coder, args, **kwargs):
                     original_coder,
                     reset=True,
                     reformat=True,
-                    preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                    preserve_tags=True,
                 )
 
                 # Restore the original model configuration

cecli/commands/editor_model.py

@@ -3,7 +3,7 @@
 import cecli.models as models
 from cecli.commands.utils.base_command import BaseCommand
 from cecli.commands.utils.helpers import format_command_result
-from cecli.helpers.conversation import ConversationManager, MessageTag
+from cecli.helpers.conversation import ConversationManager
 
 
 class EditorModelCommand(BaseCommand):
@@ -69,7 +69,7 @@ async def execute(cls, io, coder, args, **kwargs):
                 temp_coder,
                 reset=True,
                 reformat=True,
-                preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                preserve_tags=True,
             )
 
             verbose = kwargs.get("verbose", False)
@@ -86,7 +86,7 @@ async def execute(cls, io, coder, args, **kwargs):
                     original_coder,
                     reset=True,
                     reformat=True,
-                    preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                    preserve_tags=True,
                 )
 
                 # Restore the original model configuration

cecli/commands/model.py

@@ -3,7 +3,7 @@
 import cecli.models as models
 from cecli.commands.utils.base_command import BaseCommand
 from cecli.commands.utils.helpers import format_command_result
-from cecli.helpers.conversation import ConversationManager, MessageTag
+from cecli.helpers.conversation import ConversationManager
 
 
 class ModelCommand(BaseCommand):
@@ -73,7 +73,7 @@ async def execute(cls, io, coder, args, **kwargs):
                 temp_coder,
                 reset=True,
                 reformat=True,
-                preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                preserve_tags=True,
             )
 
             verbose = kwargs.get("verbose", False)
@@ -90,7 +90,7 @@ async def execute(cls, io, coder, args, **kwargs):
                     original_coder,
                     reset=True,
                     reformat=True,
-                    preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+                    preserve_tags=True,
                 )
 
                 # Restore the original model configuration

cecli/commands/tokens.py

@@ -64,6 +64,7 @@ async def execute(cls, io, coder, args, **kwargs):
         tokens_done = 0
         tokens_cur = 0
         tokens_diffs = 0
+        tokens_file_contexts = 0
 
         if msgs_done:
             tokens_done = coder.main_model.token_count(msgs_done)

cecli/commands/utils/base_command.py

@@ -1,7 +1,7 @@
 from abc import ABC, ABCMeta, abstractmethod
 from typing import List
 
-from cecli.helpers.conversation import ConversationManager, MessageTag
+from cecli.helpers.conversation import ConversationManager
 
 
 class CommandMeta(ABCMeta):
@@ -152,9 +152,7 @@ async def _generic_chat_command(cls, io, coder, args, edit_format, placeholder=N
         new_coder = await Coder.create(**kwargs)
 
         # Re-initialize ConversationManager with new coder
-        ConversationManager.initialize(
-            new_coder, reset=True, reformat=True, preserve_tags=[MessageTag.DONE, MessageTag.CUR]
-        )
+        ConversationManager.initialize(new_coder, reset=True, reformat=True, preserve_tags=True)
 
         await new_coder.generate(user_message=user_msg, preproc=False)
         coder.coder_commit_hashes = new_coder.coder_commit_hashes
@@ -164,7 +162,7 @@ async def _generic_chat_command(cls, io, coder, args, edit_format, placeholder=N
             original_coder,
             reset=True,
             reformat=True,
-            preserve_tags=[MessageTag.DONE, MessageTag.CUR],
+            preserve_tags=True,
         )
 
         from cecli.commands import SwitchCoderSignal