revert #include fix for now

Signed-off-by: Jared Van Bortel <jared@nomic.ai>

.clang-tidy

@@ -0,0 +1,126 @@
+# vim: set ft=yaml:
+HeaderFilterRegex: '/gpt4all-(backend|chat)/(src|include)/'
+Checks:
+- 'bugprone-*'
+- 'cert-*'
+- 'clang-analyzer-*'
+- 'cppcoreguidelines-*'
+- 'google-*'
+- 'misc-*'
+- 'modernize-*'
+- 'performance-*'
+- 'portability-*'
+- 'readability-*'
+- '-*-avoid-c-arrays'
+- '-*-macro-to-enum'
+- '-*-magic-numbers'
+- '-*-narrowing-conversions'
+- '-*-non-private-member-variables-in-classes'
+- '-*-use-default-member-init'
+- '-*readability-braces-around-statements'
+- '-bugprone-assignment-in-if-condition'
+- '-bugprone-branch-clone'
+- '-bugprone-crtp-constructor-accessibility'
+- '-bugprone-easily-swappable-parameters'
+- '-bugprone-implicit-widening-of-multiplication-result'
+- '-bugprone-inc-dec-in-conditions'
+- '-bugprone-suspicious-string-compare'
+- '-bugprone-suspicious-stringview-data-usage'
+- '-bugprone-switch-missing-default-case'
+- '-cert-dcl03-c'
+- '-cert-err33-c'
+- '-cert-err58-cpp'
+- '-clang-analyzer-core.CallAndMessage'
+- '-clang-analyzer-cplusplus.Move'
+- '-clang-analyzer-deadcode.DeadStores'
+- '-clang-analyzer-optin.performance.Padding'
+- '-clang-diagnostic-missing-include-dirs'
+- '-clang-diagnostic-unreachable-code-break'
+- '-cppcoreguidelines-avoid-do-while'
+- '-cppcoreguidelines-avoid-goto'
+- '-cppcoreguidelines-avoid-non-const-global-variables'
+- '-cppcoreguidelines-explicit-virtual-functions'
+- '-cppcoreguidelines-init-variables'
+- '-cppcoreguidelines-macro-usage'
+- '-cppcoreguidelines-no-malloc'
+- '-cppcoreguidelines-noexcept-move-operations'
+- '-cppcoreguidelines-owning-memory'
+- '-cppcoreguidelines-prefer-member-initializer'
+- '-cppcoreguidelines-pro-bounds-array-to-pointer-decay'
+- '-cppcoreguidelines-pro-bounds-pointer-arithmetic'
+- '-cppcoreguidelines-pro-type-const-cast'
+- '-cppcoreguidelines-pro-type-cstyle-cast'
+- '-cppcoreguidelines-pro-type-member-init'
+- '-cppcoreguidelines-pro-type-reinterpret-cast'
+- '-cppcoreguidelines-pro-type-static-cast-downcast'
+- '-cppcoreguidelines-pro-type-union-access'
+- '-cppcoreguidelines-pro-type-vararg'
+- '-cppcoreguidelines-rvalue-reference-param-not-moved'
+- '-cppcoreguidelines-special-member-functions'
+- '-cppcoreguidelines-virtual-class-destructor'
+- '-google-build-using-namespace'
+- '-google-default-arguments'
+- '-google-explicit-constructor'
+- '-google-global-names-in-headers'
+- '-google-readability-casting'
+- '-google-readability-namespace-comments'
+- '-google-readability-todo'
+- '-google-runtime-int'
+- '-misc-const-correctness'
+- '-misc-include-cleaner'
+- '-misc-no-recursion'
+- '-misc-static-assert'
+- '-misc-unused-parameters'
+- '-misc-use-anonymous-namespace'
+- '-misc-use-internal-linkage'
+- '-modernize-avoid-bind'
+- '-modernize-deprecated-headers'
+- '-modernize-loop-convert'
+- '-modernize-pass-by-value'
+- '-modernize-raw-string-literal'
+- '-modernize-return-braced-init-list'
+- '-modernize-use-auto'
+- '-modernize-use-designated-initializers'
+- '-modernize-use-equals-default'
+- '-modernize-use-nodiscard'
+- '-modernize-use-override'
+- '-modernize-use-ranges'
+- '-modernize-use-std-print'
+- '-modernize-use-trailing-return-type'
+- '-modernize-use-using'
+- '-performance-avoid-endl'
+- '-performance-enum-size'
+- '-performance-faster-string-find'
+- '-performance-for-range-copy'
+- '-performance-implicit-conversion-in-loop'
+- '-performance-move-const-arg'
+- '-performance-noexcept-move-constructor'
+- '-performance-no-automatic-move'
+- '-performance-unnecessary-copy-initialization'
+- '-performance-unnecessary-value-param'
+- '-readability-avoid-const-params-in-decls'
+- '-readability-avoid-nested-conditional-operator'
+- '-readability-avoid-return-with-void-value'
+- '-readability-avoid-unconditional-preprocessor-if'
+- '-readability-const-return-type'
+- '-readability-container-size-empty'
+- '-readability-convert-member-functions-to-static'
+- '-readability-else-after-return'
+- '-readability-function-cognitive-complexity'
+- '-readability-identifier-length'
+- '-readability-implicit-bool-conversion'
+- '-readability-inconsistent-declaration-parameter-name'
+- '-readability-isolate-declaration'
+- '-readability-math-missing-parentheses'
+- '-readability-named-parameter'
+- '-readability-non-const-parameter'
+- '-readability-qualified-auto'
+- '-readability-redundant-access-specifiers'
+- '-readability-redundant-inline-specifier'
+- '-readability-redundant-member-init'
+- '-readability-redundant-string-init'
+- '-readability-simplify-boolean-expr'
+- '-readability-static-accessed-through-instance'
+- '-readability-suspicious-call-argument'
+- '-readability-uppercase-literal-suffix'
+- '-readability-use-anyofallof'

.github/workflows/cpp-linter.yml

@@ -0,0 +1,62 @@
+name: cpp-linter
+
+on:
+  pull_request:
+    paths: ['.github/workflows/cpp-linter.yml', '**.c', '**.cpp', '**.h', '**.inl', '**CMakeLists.txt', '**.cmake']
+  push:
+    branches: [main]
+    paths: ['.github/workflows/cpp-linter.yml', '**.c', '**.cpp', '**.h', '**.inl', '**CMakeLists.txt', '**.cmake']
+
+jobs:
+  cpp-linter:
+    runs-on: ubuntu-24.04
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          submodules: 'recursive'
+      - name: Install dependencies
+        run: |
+          wget -qO- 'https://apt.llvm.org/llvm-snapshot.gpg.key' | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc >/dev/null
+          sudo add-apt-repository -y 'deb http://apt.llvm.org/noble/  llvm-toolchain-noble-19 main'
+          packages=(
+            clang-19 clang-format-19 clang-tidy-19 cmake git libgl-dev libopengl-dev libx11-6 libx11-xcb1 libxcb-glx0
+            libxcb-icccm4 libxcb-image0 libxcb-keysyms1 libxcb-randr0 libxcb-render-util0 libxcb-render0 libxcb-shape0
+            libxcb-shm0 libxcb-sync1 libxcb-xfixes0 libxcb-xinput0 libxcb-xkb1 libxcb1 libxkbcommon-dev
+            libxkbcommon-x11-0 libxkbcommon0 make
+          )
+          sudo apt-get update
+          sudo apt-get install -y "${packages[@]}" --no-install-recommends
+      - name: Install Qt
+        env:
+          QT_EMAIL: ${{ secrets.QT_EMAIL }}
+          QT_PASSWORD: ${{ secrets.QT_PASSWORD }}
+        run: |
+          wget 'https://qt.mirror.constant.com/archive/online_installers/4.8/qt-online-installer-linux-x64-4.8.0.run'
+          chmod +x qt-online-installer-linux-x64-4.8.0.run
+          ./qt-online-installer-linux-x64-4.8.0.run --no-force-installations --no-default-installations \
+            --no-size-checking --default-answer --accept-licenses --confirm-command --accept-obligations \
+            --email "$QT_EMAIL" --password "$QT_PASSWORD" install \
+            qt.qt6.680.linux_gcc_64 extensions.qtpdf.680.linux_gcc_64 qt.qt6.680.addons.qthttpserver
+      - name: Configure
+        run: |
+          sed -i '/file(DOWNLOAD/,+4 s/^/#/' gpt4all-chat/CMakeLists.txt
+          cmake -S gpt4all-chat -B gpt4all-chat/build \
+            -DCMAKE_PREFIX_PATH=~/Qt/6.8.0/gcc_64/lib/cmake \
+            -DCMAKE_C_COMPILER=clang-19 \
+            -DCMAKE_CXX_COMPILER=clang++-19 \
+            -DLLMODEL_CUDA=OFF \
+            -DLLMODEL_KOMPUTE=OFF
+      - uses: cpp-linter/cpp-linter-action@v2
+        id: linter
+        with:
+          style: ''  # no clang-format
+          extensions: 'c,cpp'
+          tidy-checks: ''  # use .clang-tidy config
+          database: gpt4all-chat/build
+          version: 19
+          files-changed-only: false
+          ignore: 'gpt4all-chat/build/_deps|gpt4all-bindings'
+      - name: Fail Fast
+        if: steps.linter.outputs.clang-tidy-checks-failed > 0
+        run: exit 1

gpt4all-backend/include/gpt4all-backend/llmodel_c.h

@@ -312,8 +312,6 @@ int32_t llmodel_count_prompt_tokens(llmodel_model model, const char *prompt, con
 
 void llmodel_model_foreach_special_token(llmodel_model model, llmodel_special_token_callback callback);
 
-const char *llmodel_model_chat_template(const char *model_path, const char **error);
-
 #ifdef __cplusplus
 }
 #endif

gpt4all-backend/src/llmodel_c.cpp

@@ -34,11 +34,11 @@ llmodel_model llmodel_model_create(const char *model_path)
     return fres;
 }
 
-static void llmodel_set_error(const char **errptr, std::string message)
+static void llmodel_set_error(const char **errptr, const char *message)
 {
     thread_local static std::string last_error_message;
     if (errptr) {
-        last_error_message = std::move(message);
+        last_error_message = message;
         *errptr = last_error_message.c_str();
     }
 }
@@ -318,15 +318,3 @@ void llmodel_model_foreach_special_token(llmodel_model model, llmodel_special_to
     for (auto &[name, token] : wrapper->llModel->specialTokens())
         callback(name.c_str(), token.c_str());
 }
-
-const char *llmodel_model_chat_template(const char *model_path, const char **error)
-{
-    static std::string s_chatTemplate;
-    auto res = LLModel::Implementation::chatTemplate(model_path);
-    if (res) {
-        s_chatTemplate = *res;
-        return s_chatTemplate.c_str();
-    }
-    llmodel_set_error(error, std::move(res.error()));
-    return nullptr;
-}

gpt4all-bindings/python/CHANGELOG.md

@@ -9,11 +9,13 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 ### Added
 - Warn on Windows if the Microsoft Visual C++ runtime libraries are not found ([#2920](https://github.com/nomic-ai/gpt4all/pull/2920))
 - Basic cache for faster prefill when the input shares a prefix with previous context ([#3073](https://github.com/nomic-ai/gpt4all/pull/3073))
+- Add ability to modify or replace the history of an active chat session ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
 
 ### Changed
 - Rebase llama.cpp on latest upstream as of September 26th ([#2998](https://github.com/nomic-ai/gpt4all/pull/2998))
 - Change the error message when a message is too long ([#3004](https://github.com/nomic-ai/gpt4all/pull/3004))
 - Fix CalledProcessError on Intel Macs since v2.8.0 ([#3045](https://github.com/nomic-ai/gpt4all/pull/3045))
+- Use Jinja for chat templates instead of per-message QString.arg-style templates ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
 
 ## [2.8.2] - 2024-08-14
 

gpt4all-bindings/python/gpt4all/_pyllmodel.py

@@ -9,7 +9,7 @@ import textwrap
 import threading
 from enum import Enum
 from queue import Queue
-from typing import TYPE_CHECKING, Any, Callable, Generic, Iterator, Literal, NoReturn, TypeVar, overload
+from typing import TYPE_CHECKING, Any, Callable, Generic, Iterable, Iterator, Literal, NoReturn, TypeVar, overload
 
 if sys.version_info >= (3, 9):
     import importlib.resources as importlib_resources
@@ -227,9 +227,6 @@ llmodel.llmodel_count_prompt_tokens.restype = ctypes.c_int32
 llmodel.llmodel_model_foreach_special_token.argtypes = [ctypes.c_void_p, SpecialTokenCallback]
 llmodel.llmodel_model_foreach_special_token.restype = None
 
-llmodel.llmodel_model_chat_template.argtypes = [ctypes.c_char_p, ctypes.POINTER(ctypes.c_char_p)]
-llmodel.llmodel_model_chat_template.restype = ctypes.c_char_p
-
 ResponseCallbackType = Callable[[int, str], bool]
 RawResponseCallbackType = Callable[[int, bytes], bool]
 EmbCancelCallbackType: TypeAlias = 'Callable[[list[int], str], bool]'
@@ -293,7 +290,10 @@ class LLModel:
 
             raise RuntimeError(f"Unable to instantiate model: {errmsg}")
         self.model: ctypes.c_void_p | None = model
-        self._special_tokens_map: dict[str, str] | None = None
+        self.special_tokens_map: dict[str, str] = {}
+        llmodel.llmodel_model_foreach_special_token(
+            self.model, lambda n, t: self.special_tokens_map.__setitem__(n.decode(), t.decode()),
+        )
 
     def __del__(self, llmodel=llmodel):
         if hasattr(self, 'model'):
@@ -320,26 +320,6 @@ class LLModel:
         dev = llmodel.llmodel_model_gpu_device_name(self.model)
         return None if dev is None else dev.decode()
 
-    @property
-    def builtin_chat_template(self) -> str:
-        err = ctypes.c_char_p()
-        tmpl = llmodel.llmodel_model_chat_template(self.model_path, ctypes.byref(err))
-        if tmpl is not None:
-            return tmpl.decode()
-        s = err.value
-        raise ValueError('Failed to get chat template', 'null' if s is None else s.decode())
-
-    @property
-    def special_tokens_map(self) -> dict[str, str]:
-        if self.model is None:
-            self._raise_closed()
-        if self._special_tokens_map is None:
-            tokens: dict[str, str] = {}
-            cb = SpecialTokenCallback(lambda n, t: tokens.__setitem__(n.decode(), t.decode()))
-            llmodel.llmodel_model_foreach_special_token(self.model, cb)
-            self._special_tokens_map = tokens
-        return self._special_tokens_map
-
     def count_prompt_tokens(self, prompt: str) -> int:
         if self.model is None:
             self._raise_closed()
@@ -351,6 +331,8 @@ class LLModel:
             raise RuntimeError(f'Unable to count prompt tokens: {errmsg}')
         return n_tok
 
+    llmodel.llmodel_count_prompt_tokens.argtypes = [ctypes.c_void_p, ctypes.c_char_p]
+
     @staticmethod
     def list_gpus(mem_required: int = 0) -> list[str]:
         """

gpt4all-bindings/python/gpt4all/gpt4all.py

@@ -37,9 +37,9 @@ DEFAULT_MODEL_DIRECTORY = Path.home() / ".cache" / "gpt4all"
 
 ConfigType: TypeAlias = "dict[str, Any]"
 
+# Environment setup adapted from HF transformers
 @_operator_call
 def _jinja_env() -> ImmutableSandboxedEnvironment:
-    # Environment setup adapted from HF transformers
     def raise_exception(message: str) -> NoReturn:
         raise jinja2.exceptions.TemplateError(message)
 
@@ -56,17 +56,14 @@ def _jinja_env() -> ImmutableSandboxedEnvironment:
     return env
 
 
-class Message(TypedDict):
-    """A message in a chat with a GPT4All model."""
-
+class MessageType(TypedDict):
     role: str
     content: str
 
 
-class _ChatSession(NamedTuple):
-    template:        jinja2.Template
-    template_source: str
-    history:         list[Message]
+class ChatSession(NamedTuple):
+    template: jinja2.Template
+    history: list[MessageType]
 
 
 class Embed4All:
@@ -196,17 +193,6 @@ class GPT4All:
     Python class that handles instantiation, downloading, generation and chat with GPT4All models.
     """
 
-    RE_LEGACY_SYSPROMPT = re.compile(
-        r"(?:^|\s)(?:### *System\b|S(?:ystem|YSTEM):)|"
-        r"<\|(?:im_(?:start|end)|(?:start|end)_header_id|eot_id|SYSTEM_TOKEN)\|>|<<SYS>>",
-        re.MULTILINE,
-    )
-
-    RE_JINJA_LIKE = re.compile(
-        r"\{%.*%\}.*\{\{.*\}\}.*\{%.*%\}",
-        re.DOTALL,
-    )
-
     def __init__(
         self,
         model_name: str,
@@ -247,7 +233,7 @@ class GPT4All:
         """
 
         self.model_type = model_type
-        self._chat_session: _ChatSession | None = None
+        self._chat_session: ChatSession | None = None
 
         device_init = None
         if sys.platform == "darwin":
@@ -274,7 +260,6 @@ class GPT4All:
 
         # Retrieve model and download if allowed
         self.config: ConfigType = self.retrieve_model(model_name, model_path=model_path, allow_download=allow_download, verbose=verbose)
-        self._was_allow_download = allow_download
         self.model = LLModel(self.config["path"], n_ctx, ngl, backend)
         if device_init is not None:
             self.model.init_gpu(device_init)
@@ -306,20 +291,15 @@ class GPT4All:
         return self.model.device
 
     @property
-    def current_chat_session(self) -> list[Message] | None:
-        """The message history of the current chat session."""
+    def current_chat_session(self) -> list[MessageType] | None:
         return None if self._chat_session is None else self._chat_session.history
 
     @current_chat_session.setter
-    def current_chat_session(self, history: list[Message]) -> None:
+    def current_chat_session(self, history: list[MessageType]) -> None:
         if self._chat_session is None:
             raise ValueError("current_chat_session may only be set when there is an active chat session")
         self._chat_session.history[:] = history
 
-    @property
-    def current_chat_template(self) -> str | None:
-        return None if self._chat_session is None else self._chat_session.template_source
-
     @staticmethod
     def list_models() -> list[ConfigType]:
         """
@@ -589,13 +569,13 @@ class GPT4All:
         last_msg_rendered = prompt
         if self._chat_session is not None:
             session = self._chat_session
-            def render(messages: list[Message]) -> str:
+            def render(messages: list[MessageType]) -> str:
                 return session.template.render(
                     messages=messages,
                     add_generation_prompt=True,
                     **self.model.special_tokens_map,
                 )
-            session.history.append(Message(role="user", content=prompt))
+            session.history.append(MessageType(role="user", content=prompt))
             prompt = render(session.history)
             if len(session.history) > 1:
                 last_msg_rendered = render(session.history[-1:])
@@ -610,73 +590,46 @@ class GPT4All:
             def stream() -> Iterator[str]:
                 yield from self.model.prompt_model_streaming(prompt, _callback_wrapper, **generate_kwargs)
                 if self._chat_session is not None:
-                    self._chat_session.history.append(Message(role="assistant", content=full_response))
+                    self._chat_session.history.append(MessageType(role="assistant", content=full_response))
             return stream()
 
         self.model.prompt_model(prompt, _callback_wrapper, **generate_kwargs)
         if self._chat_session is not None:
-            self._chat_session.history.append(Message(role="assistant", content=full_response))
+            self._chat_session.history.append(MessageType(role="assistant", content=full_response))
         return full_response
 
     @contextmanager
     def chat_session(
         self,
-        *,
         system_message: str | Literal[False] | None = None,
         chat_template: str | None = None,
-        warn_legacy: bool = True,
     ):
         """
         Context manager to hold an inference optimized chat session with a GPT4All model.
 
         Args:
-            system_message: An initial instruction for the model, None to use the model default, or False to disable.
-                Defaults to None.
+            system_message: An initial instruction for the model, None to use the model default, or False to disable. Defaults to None.
             chat_template: Jinja template for the conversation, or None to use the model default. Defaults to None.
-            warn_legacy: Whether to warn about legacy system prompts or prompt templates. Defaults to True.
-
-        Raises:
-            ValueError: If no valid chat template was found.
         """
+
         if system_message is None:
             system_message = self.config.get("systemMessage", False)
-        elif system_message is not False and warn_legacy and (m := self.RE_LEGACY_SYSPROMPT.search(system_message)):
-            print(
-                "Warning: chat_session() was passed a system message that is not plain text. System messages "
-                f"containing {m.group()!r} or with any special prefix/suffix are no longer supported.\nTo disable this "
-                "warning, pass warn_legacy=False.",
-                file=sys.stderr,
-            )
 
         if chat_template is None:
-            if "chatTemplate" in self.config:
-                if (tmpl := self.config["chatTemplate"]) is None:
-                    raise ValueError(f"The model {self.config['name']!r} does not support chat.")
-                chat_template = tmpl
-            else:
-                try:
-                    chat_template = self.model.builtin_chat_template
-                except ValueError as e:
-                    if len(e.args) >= 2 and isinstance(err := e.args[1], str):
-                        msg = (f"Failed to load default chat template from model: {err}\n"
-                               "Please pass a template to chat_session() directly.")
-                        if not self._was_allow_download:
-                            msg += " If this is a built-in model, consider setting allow_download to True."
-                        raise ValueError(msg) from None
-                    raise
-        elif warn_legacy and self._is_legacy_chat_template(chat_template):
-            print(
-                "Warning: chat_session() was passed a chat template that is not in Jinja format. Old-style prompt "
-                "templates are no longer supported.\nTo disable this warning, pass warn_legacy=False.",
-                file=sys.stderr,
-            )
+            if "name" not in self.config:
+                raise ValueError("For sideloaded models or with allow_download=False, you must specify a chat template.")
+            if "chatTemplate" not in self.config:
+                raise NotImplementedError("This model appears to have a built-in chat template, but loading it is not "
+                                          "currently implemented. Please pass a template to chat_session() directly.")
+            if (tmpl := self.config["chatTemplate"]) is None:
+                raise ValueError(f"The model {self.config['name']!r} does not support chat.")
+            chat_template = tmpl
 
         history = []
         if system_message is not False:
-            history.append(Message(role="system", content=system_message))
-        self._chat_session = _ChatSession(
+            history.append(MessageType(role="system", content=system_message))
+        self._chat_session = ChatSession(
             template=_jinja_env.from_string(chat_template),
-            template_source=chat_template,
             history=history,
         )
         try:
@@ -694,12 +647,6 @@ class GPT4All:
         """
         return LLModel.list_gpus()
 
-    @classmethod
-    def _is_legacy_chat_template(cls, tmpl: str) -> bool:
-        # check if tmpl does not look like a Jinja template
-        return bool(re.search(r"%[12]\b", tmpl) or not cls.RE_JINJA_LIKE.search(tmpl)
-                    or not re.search(r"\bcontent\b", tmpl))
-
 
 def append_extension_if_missing(model_name):
     if not model_name.endswith((".bin", ".gguf")):

gpt4all-bindings/python/setup.py

@@ -68,7 +68,7 @@ def get_long_description():
 
 setup(
     name=package_name,
-    version="3.0.0.dev0",
+    version="2.8.3.dev0",
     description="Python bindings for GPT4All",
     long_description=get_long_description(),
     long_description_content_type="text/markdown",

gpt4all-chat/CHANGELOG.md

@@ -4,15 +4,30 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
+## [3.5.0-rc2] - 2024-12-06
+
+### Changed
+- Fade messages out with an animation when they are removed from the chat view ([#3227](https://github.com/nomic-ai/gpt4all/pull/3227))
+- Tweak wording of edit/redo confirmation dialogs ([#3228](https://github.com/nomic-ai/gpt4all/pull/3228))
+- Make edit/redo buttons disabled instead of invisible when they are temporarily unavailable ([#3228](https://github.com/nomic-ai/gpt4all/pull/3228))
+
 ## [3.5.0-rc1] - 2024-12-04
 
 ### Added
 - Add ability to attach text, markdown, and rst files to chat ([#3135](https://github.com/nomic-ai/gpt4all/pull/3135))
 - Add feature to minimize to system tray (by [@bgallois](https://github.com/bgallois) in ([#3109](https://github.com/nomic-ai/gpt4all/pull/3109))
 - Basic cache for faster prefill when the input shares a prefix with previous context ([#3073](https://github.com/nomic-ai/gpt4all/pull/3073))
+- Add ability to edit prompts and regenerate any response ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
 
 ### Changed
 - Implement Qt 6.8 compatibility ([#3121](https://github.com/nomic-ai/gpt4all/pull/3121))
+- Use Jinja for chat templates instead of per-message QString.arg-style templates ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
+- API server: Use system message(s) from client instead of settings ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
+- API server: Accept messages in any order supported by the model instead of requiring user/assistant pairs ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
+- Remote models: Pass system message with "system" role instead of joining with user message ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
+
+### Removed
+- Remove option to save binary model state to disk ([#3147](https://github.com/nomic-ai/gpt4all/pull/3147))
 
 ### Fixed
 - Fix bug in GUI when localdocs encounters binary data ([#3137](https://github.com/nomic-ai/gpt4all/pull/3137))
@@ -179,6 +194,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 - Fix several Vulkan resource management issues ([#2694](https://github.com/nomic-ai/gpt4all/pull/2694))
 - Fix crash/hang when some models stop generating, by showing special tokens ([#2701](https://github.com/nomic-ai/gpt4all/pull/2701))
 
+[3.5.0-rc2]: https://github.com/nomic-ai/gpt4all/compare/v3.5.0-rc1...v3.5.0-rc2
 [3.5.0-rc1]: https://github.com/nomic-ai/gpt4all/compare/v3.4.2...v3.5.0-rc1
 [3.4.2]: https://github.com/nomic-ai/gpt4all/compare/v3.4.1...v3.4.2
 [3.4.1]: https://github.com/nomic-ai/gpt4all/compare/v3.4.0...v3.4.1

gpt4all-chat/CMakeLists.txt

@@ -6,7 +6,7 @@ set(APP_VERSION_MAJOR 3)
 set(APP_VERSION_MINOR 5)
 set(APP_VERSION_PATCH 0)
 set(APP_VERSION_BASE "${APP_VERSION_MAJOR}.${APP_VERSION_MINOR}.${APP_VERSION_PATCH}")
-set(APP_VERSION "${APP_VERSION_BASE}-rc1")
+set(APP_VERSION "${APP_VERSION_BASE}-rc2")
 
 project(gpt4all VERSION ${APP_VERSION_BASE} LANGUAGES CXX C)
 

gpt4all-chat/qml/ChatItemView.qml

@@ -529,8 +529,8 @@ GridLayout {
 
     ConfirmationDialog {
         id: editPromptDialog
-        dialogTitle: qsTr("Edit this prompt?")
-        description: qsTr("The existing response and all later messages will be permanently erased.")
+        dialogTitle: qsTr("Edit this message?")
+        description: qsTr("All following messages will be permanently erased.")
         onAccepted: {
             const msg = currentChat.popPrompt(index);
             if (msg !== null)
@@ -541,7 +541,7 @@ GridLayout {
     ConfirmationDialog {
         id: redoResponseDialog
         dialogTitle: qsTr("Redo this response?")
-        description: qsTr("The existing response and all later messages will be permanently erased.")
+        description: qsTr("All following messages will be permanently erased.")
         onAccepted: currentChat.regenerateResponse(index)
     }
 
@@ -556,33 +556,48 @@ GridLayout {
         visible: !isCurrentResponse || !currentChat.responseInProgress
         enabled: opacity > 0
         opacity: hoverArea.hovered
-        readonly property var canModify: !currentChat.isServer && currentChat.isModelLoaded && !currentChat.responseInProgress
 
         Behavior on opacity {
             OpacityAnimator { duration: 30 }
         }
 
         ChatMessageButton {
-            visible: parent.canModify && model.name === "Prompt: "
+            readonly property var editingDisabledReason: {
+                if (!currentChat.isModelLoaded)
+                    return qsTr("Cannot edit chat without a loaded model.");
+                if (currentChat.responseInProgress)
+                    return qsTr("Cannot edit chat while the model is generating.");
+                return null;
+            }
+            visible: !currentChat.isServer && model.name === "Prompt: "
+            enabled: editingDisabledReason === null
             Layout.maximumWidth: 24
             Layout.maximumHeight: 24
             Layout.alignment: Qt.AlignVCenter
             Layout.fillWidth: false
+            name: editingDisabledReason ?? qsTr("Edit")
             source: "qrc:/gpt4all/icons/edit.svg"
             onClicked: {
                 if (inputBoxText === "")
                     editPromptDialog.open();
             }
-            name: qsTr("Edit")
         }
 
         ChatMessageButton {
-            visible: parent.canModify && model.name === "Response: "
+            readonly property var editingDisabledReason: {
+                if (!currentChat.isModelLoaded)
+                    return qsTr("Cannot redo response without a loaded model.");
+                if (currentChat.responseInProgress)
+                    return qsTr("Cannot redo response while the model is generating.");
+                return null;
+            }
+            visible: !currentChat.isServer && model.name === "Response: "
+            enabled: editingDisabledReason === null
             Layout.maximumWidth: 24
             Layout.maximumHeight: 24
             Layout.alignment: Qt.AlignVCenter
             Layout.fillWidth: false
-            name: qsTr("Redo")
+            name: editingDisabledReason ?? qsTr("Redo")
             source: "qrc:/gpt4all/icons/regenerate.svg"
             onClicked: redoResponseDialog.open()
         }

gpt4all-chat/qml/ChatView.qml

@@ -793,7 +793,7 @@ Rectangle {
 
                     ColumnLayout {
                         anchors.fill: parent
-                        visible: ModelList.selectableModels.count !== 0 && chatModel.count !== 0
+                        visible: ModelList.selectableModels.count !== 0
                         ListView {
                             id: listView
                             Layout.maximumWidth: 1280
@@ -825,6 +825,10 @@ Rectangle {
                                 }
                             }
 
+                            remove: Transition {
+                                OpacityAnimator { to: 0; duration: 500 }
+                            }
+
                             function scrollToEnd() {
                                 listView.positionViewAtEnd()
                             }

gpt4all-chat/qml/MyToolButton.qml

@@ -49,7 +49,7 @@ Button {
         ColorOverlay {
             anchors.fill: image
             source: image
-            color: myButton.hovered ? backgroundColorHovered : backgroundColor
+            color: !myButton.enabled ? theme.mutedTextColor : myButton.hovered ? backgroundColorHovered : backgroundColor
         }
     }
     Accessible.role: Accessible.Button