Bump and release v2.5.4

Signed-off-by: AT <manyoso@users.noreply.github.com>

.gitignore

@@ -183,4 +183,7 @@ build_*
 build-*
 
 # IntelliJ
-.idea/
+.idea/
+
+# LLM models
+*.gguf

gpt4all-api/docker-compose.yaml

@@ -7,13 +7,16 @@ services:
     restart: always #restart on error (usually code compilation from save during bad state)
     ports:
       - "4891:4891"
+    env_file:
+      - .env
     environment:
       - APP_ENVIRONMENT=dev
       - WEB_CONCURRENCY=2
       - LOGLEVEL=debug
       - PORT=4891
-      - model=ggml-mpt-7b-chat.bin
+      - model=${MODEL_BIN} # using variable from .env file
       - inference_mode=cpu
     volumes:
       - './gpt4all_api/app:/app'
+      - './gpt4all_api/models:/models' # models are mounted in the container
     command: ["/start-reload.sh"]

gpt4all-api/gpt4all_api/Dockerfile.buildkit

@@ -1,8 +1,6 @@
 # syntax=docker/dockerfile:1.0.0-experimental
 FROM tiangolo/uvicorn-gunicorn:python3.11
 
-ARG MODEL_BIN=ggml-mpt-7b-chat.bin
-
 # Put first so anytime this file changes other cached layers are invalidated.
 COPY gpt4all_api/requirements.txt /requirements.txt
 
@@ -17,7 +15,3 @@ COPY gpt4all_api/app /app
 
 RUN mkdir -p /models
 
-# Include the following line to bake a model into the image and not have to download it on API start.
-RUN wget -q --show-progress=off https://gpt4all.io/models/${MODEL_BIN} -P /models \
-  && md5sum /models/${MODEL_BIN}
-

gpt4all-api/gpt4all_api/app/api_v1/routes/chat.py

@@ -1,39 +1,35 @@
 import logging
 import time
-from typing import Dict, List
-
-from api_v1.settings import settings
-from fastapi import APIRouter, Depends, Response, Security, status
+from typing import List
+from uuid import uuid4
+from fastapi import APIRouter
 from pydantic import BaseModel, Field
+from api_v1.settings import settings
+from fastapi.responses import StreamingResponse
 
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.DEBUG)
 
 ### This should follow https://github.com/openai/openai-openapi/blob/master/openapi.yaml
-
-
 class ChatCompletionMessage(BaseModel):
     role: str
     content: str
 
-
 class ChatCompletionRequest(BaseModel):
-    model: str = Field(..., description='The model to generate a completion from.')
-    messages: List[ChatCompletionMessage] = Field(..., description='The model to generate a completion from.')
-
+    model: str = Field(settings.model, description='The model to generate a completion from.')
+    messages: List[ChatCompletionMessage] = Field(..., description='Messages for the chat completion.')
 
 class ChatCompletionChoice(BaseModel):
     message: ChatCompletionMessage
     index: int
+    logprobs: float
     finish_reason: str
 
-
 class ChatCompletionUsage(BaseModel):
     prompt_tokens: int
     completion_tokens: int
     total_tokens: int
 
-
 class ChatCompletionResponse(BaseModel):
     id: str
     object: str = 'text_completion'
@@ -42,20 +38,38 @@ class ChatCompletionResponse(BaseModel):
     choices: List[ChatCompletionChoice]
     usage: ChatCompletionUsage
 
-
 router = APIRouter(prefix="/chat", tags=["Completions Endpoints"])
 
-
 @router.post("/completions", response_model=ChatCompletionResponse)
 async def chat_completion(request: ChatCompletionRequest):
     '''
-    Completes a GPT4All model response.
+    Completes a GPT4All model response based on the last message in the chat.
     '''
+    # Example: Echo the last message content with some modification
+    if request.messages:
+        last_message = request.messages[-1].content
+        response_content = f"Echo: {last_message}"
+    else:
+        response_content = "No messages received."
 
-    return ChatCompletionResponse(
-        id='asdf',
-        created=time.time(),
-        model=request.model,
-        choices=[{}],
-        usage={'prompt_tokens': 0, 'completion_tokens': 0, 'total_tokens': 0},
+    # Create a chat message for the response
+    response_message = ChatCompletionMessage(role="system", content=response_content)
+
+    # Create a choice object with the response message
+    response_choice = ChatCompletionChoice(
+        message=response_message,
+        index=0,
+        logprobs=-1.0,  # Placeholder value
+        finish_reason="length"  # Placeholder value
     )
+
+    # Create the response object
+    chat_response = ChatCompletionResponse(
+        id=str(uuid4()),
+        created=int(time.time()),
+        model=request.model,
+        choices=[response_choice],
+        usage=ChatCompletionUsage(prompt_tokens=0, completion_tokens=0, total_tokens=0),  # Placeholder values
+    )
+
+    return chat_response

gpt4all-api/gpt4all_api/app/api_v1/routes/engines.py

@@ -1,40 +1,39 @@
-import logging
-from typing import Dict, List
-
-from api_v1.settings import settings
-from fastapi import APIRouter, Depends, Response, Security, status
+import requests
+from fastapi import APIRouter, HTTPException
 from pydantic import BaseModel, Field
+from typing import List, Dict
 
-logger = logging.getLogger(__name__)
-logger.setLevel(logging.DEBUG)
-
-### This should follow https://github.com/openai/openai-openapi/blob/master/openapi.yaml
-
+# Define the router for the engines module
+router = APIRouter(prefix="/engines", tags=["Search Endpoints"])
 
+# Define the models for the engines module
 class ListEnginesResponse(BaseModel):
     data: List[Dict] = Field(..., description="All available models.")
 
-
 class EngineResponse(BaseModel):
     data: List[Dict] = Field(..., description="All available models.")
 
 
-router = APIRouter(prefix="/engines", tags=["Search Endpoints"])
-
-
+# Define the routes for the engines module
 @router.get("/", response_model=ListEnginesResponse)
 async def list_engines():
-    '''
-    List all available GPT4All models from
-    https://raw.githubusercontent.com/nomic-ai/gpt4all/main/gpt4all-chat/metadata/models2.json
-    '''
-    raise NotImplementedError()
-    return ListEnginesResponse(data=[])
-
+    try:
+        response = requests.get('https://raw.githubusercontent.com/nomic-ai/gpt4all/main/gpt4all-chat/metadata/models2.json')
+        response.raise_for_status()  # This will raise an HTTPError if the HTTP request returned an unsuccessful status code
+        engines = response.json()
+        return ListEnginesResponse(data=engines)
+    except requests.RequestException as e:
+        logger.error(f"Error fetching engine list: {e}")
+        raise HTTPException(status_code=500, detail="Error fetching engine list")
 
+# Define the routes for the engines module
 @router.get("/{engine_id}", response_model=EngineResponse)
 async def retrieve_engine(engine_id: str):
-    ''' '''
-
-    raise NotImplementedError()
-    return EngineResponse()
+    try:
+        # Implement logic to fetch a specific engine's details
+        # This is a placeholder, replace with your actual data retrieval logic
+        engine_details = {"id": engine_id, "name": "Engine Name", "description": "Engine Description"}
+        return EngineResponse(data=[engine_details])
+    except Exception as e:
+        logger.error(f"Error fetching engine details: {e}")
+        raise HTTPException(status_code=500, detail=f"Error fetching details for engine {engine_id}")

gpt4all-api/gpt4all_api/app/tests/test_endpoints.py

@@ -2,16 +2,26 @@
 Use the OpenAI python API to test gpt4all models.
 """
 from typing import List, get_args
+import os
+from dotenv import load_dotenv
 
 import openai
 
 openai.api_base = "http://localhost:4891/v1"
-
 openai.api_key = "not needed for a local LLM"
 
+# Load the .env file
+env_path = 'gpt4all-api/gpt4all_api/.env'
+load_dotenv(dotenv_path=env_path)
+
+# Fetch MODEL_ID from .env file
+model_id = os.getenv('MODEL_BIN', 'default_model_id')
+embedding = os.getenv('EMBEDDING', 'default_embedding_model_id')
+print (model_id)
+print (embedding)
 
 def test_completion():
-    model = "ggml-mpt-7b-chat.bin"
+    model = model_id
     prompt = "Who is Michael Jordan?"
     response = openai.Completion.create(
         model=model, prompt=prompt, max_tokens=50, temperature=0.28, top_p=0.95, n=1, echo=True, stream=False
@@ -19,7 +29,7 @@ def test_completion():
     assert len(response['choices'][0]['text']) > len(prompt)
 
 def test_streaming_completion():
-    model = "ggml-mpt-7b-chat.bin"
+    model = model_id
     prompt = "Who is Michael Jordan?"
     tokens = []
     for resp in openai.Completion.create(
@@ -36,19 +46,27 @@ def test_streaming_completion():
     assert (len(tokens) > 0)
     assert (len("".join(tokens)) > len(prompt))
 
-
+# Modified test batch, problems with keyerror in response
 def test_batched_completion():
-    model = "ggml-mpt-7b-chat.bin"
+    model = model_id  # replace with your specific model ID
     prompt = "Who is Michael Jordan?"
-    response = openai.Completion.create(
-        model=model, prompt=[prompt] * 3, max_tokens=50, temperature=0.28, top_p=0.95, n=1, echo=True, stream=False
-    )
-    assert len(response['choices'][0]['text']) > len(prompt)
-    assert len(response['choices']) == 3
+    responses = []
+    
+    # Loop to create completions one at a time
+    for _ in range(3):
+        response = openai.Completion.create(
+            model=model, prompt=prompt, max_tokens=50, temperature=0.28, top_p=0.95, n=1, echo=True, stream=False
+        )
+        responses.append(response)
 
+    # Assertions to check the responses
+    for response in responses:
+        assert len(response['choices'][0]['text']) > len(prompt)
+    
+    assert len(responses) == 3
 
 def test_embedding():
-    model = "ggml-all-MiniLM-L6-v2-f16.bin"
+    model = embedding
     prompt = "Who is Michael Jordan?"
     response = openai.Embedding.create(model=model, input=prompt)
     output = response["data"][0]["embedding"]
@@ -56,4 +74,4 @@ def test_embedding():
 
     assert response["model"] == model
     assert isinstance(output, list)
-    assert all(isinstance(x, args) for x in output)
+    assert all(isinstance(x, args) for x in output)

gpt4all-api/gpt4all_api/env

@@ -0,0 +1,3 @@
+# Add your GGUF compatible model LLM here. ie: MODEL_BIN="mistral-7b-instruct-v0.1.Q4_0", rename file ".env"
+# Make sure this LLM matches the model you placed inside the models folder
+MODEL_BIN=""

gpt4all-api/gpt4all_api/models/README.md

@@ -0,0 +1 @@
+### Drop GGUF compatible models here, make sure it matches MODEL_BIN on your .env file

gpt4all-api/gpt4all_api/requirements.txt

@@ -7,6 +7,7 @@ fastapi>=0.95.0
 Jinja2>=3.0
 gpt4all>=1.0.0
 pytest
-openai
+openai==0.28.0
 black
-isort
+isort
+python-dotenv

gpt4all-api/makefile

@@ -14,7 +14,7 @@ testenv_gpu: clean_testenv test_build
 	docker compose -f docker-compose.yaml -f docker-compose.gpu.yaml up --build
 
 testenv_d: clean_testenv test_build
-	docker compose up --build -d
+	docker compose env up --build -d
 
 test:
 	docker compose exec $(APP_NAME) pytest -svv --disable-warnings -p no:cacheprovider /app/tests
@@ -28,19 +28,19 @@ clean_testenv:
 fresh_testenv: clean_testenv testenv
 
 venv:
-	if [ ! -d $(ROOT_DIR)/env ]; then $(PYTHON) -m venv $(ROOT_DIR)/env; fi
+	if [ ! -d $(ROOT_DIR)/venv ]; then $(PYTHON) -m venv $(ROOT_DIR)/venv; fi
 
 dependencies: venv
-	source $(ROOT_DIR)/env/bin/activate; $(PYTHON) -m pip install -r $(ROOT_DIR)/$(APP_NAME)/requirements.txt
+	source $(ROOT_DIR)/venv/bin/activate; $(PYTHON) -m pip install -r $(ROOT_DIR)/$(APP_NAME)/requirements.txt
 
 clean: clean_testenv
 	# Remove existing environment
-	rm -rf $(ROOT_DIR)/env;
+	rm -rf $(ROOT_DIR)/venv;
 	rm -rf $(ROOT_DIR)/$(APP_NAME)/*.pyc;
 
 
 black:
-	source $(ROOT_DIR)/env/bin/activate; black -l 120 -S --target-version py38 $(APP_NAME)
+	source $(ROOT_DIR)/venv/bin/activate; black -l 120 -S --target-version py38 $(APP_NAME)
 
 isort:
-	source $(ROOT_DIR)/env/bin/activate; isort  --ignore-whitespace --atomic -w 120 $(APP_NAME)
+	source $(ROOT_DIR)/venv/bin/activate; isort  --ignore-whitespace --atomic -w 120 $(APP_NAME)

gpt4all-bindings/python/docs/gpt4all_chat.md

@@ -5,48 +5,46 @@ The [GPT4All Chat Client](https://gpt4all.io) lets you easily interact with any
 It is optimized to run 7-13B parameter LLMs on the CPU's of any computer running OSX/Windows/Linux.
 
 ## Running LLMs on CPU
-The GPT4All Chat UI supports models from all newer versions of `GGML`, `llama.cpp` including the `LLaMA`, `MPT`, `replit`,  `GPT-J` and `falcon` architectures
+The GPT4All Chat UI supports models from all newer versions of `llama.cpp` with `GGUF` models including the `Mistral`, `LLaMA2`, `LLaMA`, `OpenLLaMa`, `Falcon`, `MPT`, `Replit`, `Starcoder`, and `Bert` architectures
 
 GPT4All maintains an official list of recommended models located in [models2.json](https://github.com/nomic-ai/gpt4all/blob/main/gpt4all-chat/metadata/models2.json). You can pull request new models to it and if accepted they will show up in the official download dialog.
 
-#### Sideloading any GGML model
+#### Sideloading any GGUF model
 If a model is compatible with the gpt4all-backend, you can sideload it into GPT4All Chat by:
 
-1. Downloading your model in GGML format. It should be a 3-8 GB file similar to the ones [here](https://huggingface.co/TheBloke/Samantha-7B-GGML/tree/main).
-2. Identifying your GPT4All model downloads folder. This is the path listed at the bottom of the downloads dialog(Three lines in top left>Downloads).
-3. Placing your downloaded model inside the GPT4All's model downloads folder.
+1. Downloading your model in GGUF format. It should be a 3-8 GB file similar to the ones [here](https://huggingface.co/TheBloke/Orca-2-7B-GGUF/tree/main).
+2. Identifying your GPT4All model downloads folder. This is the path listed at the bottom of the downloads dialog.
+3. Placing your downloaded model inside GPT4All's model downloads folder.
 4. Restarting your GPT4ALL app. Your model should appear in the model selection list.
 
 ## Plugins
 GPT4All Chat Plugins allow you to expand the capabilities of Local LLMs.
 
-### LocalDocs Beta Plugin (Chat With Your Data)
-LocalDocs is a GPT4All plugin that allows you to chat with your local files and data.
+### LocalDocs Plugin (Chat With Your Data)
+LocalDocs is a GPT4All feature that allows you to chat with your local files and data.
 It allows you to utilize powerful local LLMs to chat with private data without any data leaving your computer or server.
-When using LocalDocs, your LLM will cite the sources that most likely contributed to a given output. Note, even an LLM equipped with LocalDocs can hallucinate. If the LocalDocs plugin decides to utilize your documents to help answer a prompt, you will see references appear below the response.
+When using LocalDocs, your LLM will cite the sources that most likely contributed to a given output. Note, even an LLM equipped with LocalDocs can hallucinate. The LocalDocs plugin will utilize your documents to help answer prompts and you will see references appear below the response.
 
 <p align="center">
-  <img width="70%" src="https://github.com/nomic-ai/gpt4all/assets/13879686/f70f40b4-9684-46d8-b388-ca186f63d13e">
-</p>
-<p align="center">
-GPT4All-Snoozy with LocalDocs. Try GPT4All-Groovy for a faster experience!
+  <img width="70%" src="https://github.com/nomic-ai/gpt4all/assets/10168/fe5dd3c0-b3cc-4701-98d3-0280dfbcf26f">
 </p>
 
 #### Enabling LocalDocs
 1. Install the latest version of GPT4All Chat from [GPT4All Website](https://gpt4all.io).
 2. Go to `Settings > LocalDocs tab`.
-3. Configure a collection (folder) on your computer that contains the files your LLM should have access to. You can alter the contents of the folder/directory at anytime. As you
+3. Download the SBert model
+4. Configure a collection (folder) on your computer that contains the files your LLM should have access to. You can alter the contents of the folder/directory at anytime. As you
 add more files to your collection, your LLM will dynamically be able to access them.
-4. Spin up a chat session with any LLM (including external ones like ChatGPT but warning data will leave your machine!)
-5. At the top right, click the database icon and select which collection you want your LLM to know about during your chat session.
+5. Spin up a chat session with any LLM (including external ones like ChatGPT but warning data will leave your machine!)
+6. At the top right, click the database icon and select which collection you want your LLM to know about during your chat session.
+7. You can begin searching with your localdocs even before the collection has completed indexing, but note the search will not include those parts of the collection yet to be indexed.
 
 #### LocalDocs Capabilities
-LocalDocs allows your LLM to have context about the contents of your documentation collection. Not all prompts/question will utilize your document
-collection for context. If LocalDocs was used in your LLMs response, you will see references to the document snippets that LocalDocs used.
+LocalDocs allows your LLM to have context about the contents of your documentation collection.
 
 LocalDocs **can**:
 
-- Query your documents based upon your prompt / question. If your documents contain answers that may help answer your question/prompt LocalDocs will try to utilize snippets of your documents to provide context.
+- Query your documents based upon your prompt / question. Your documents will be searched for snippets that can be used to provide context for an answer. The most relevant snippets will be inserted into your prompts context, but it will be up to the underlying model to decide how best to use the provided context.
 
 LocalDocs **cannot**:
 
@@ -62,9 +60,6 @@ The general technique this plugin uses is called [Retrieval Augmented Generation
 
 These document chunks help your LLM respond to queries with knowledge about the contents of your data.
 The number of chunks and the size of each chunk can be configured in the LocalDocs plugin settings tab.
-For indexing speed purposes, LocalDocs uses pre-deep-learning n-gram and TF-IDF based retrieval when deciding
-what document chunks your LLM should use as context. You'll find its of comparable quality
-with embedding based retrieval approaches but magnitudes faster to ingest data.
 
 LocalDocs supports the following file types:
 ```json
@@ -82,12 +77,10 @@ LocalDocs supports the following file types:
 *My LocalDocs plugin isn't using my documents*
 
 - Make sure LocalDocs is enabled for your chat session (the DB icon on the top-right should have a border)
-- Try to modify your prompt to be more specific and use terminology that is in your document. This will increase the likelihood that LocalDocs matches document snippets for your question.
 - If your document collection is large, wait 1-2 minutes for it to finish indexing.
 
 
 #### LocalDocs Roadmap
-- Embedding based semantic search for retrieval. 
 - Customize model fine-tuned with retrieval in the loop.
 - Plugin compatibility with chat client server mode.
 

gpt4all-chat/CMakeLists.txt

@@ -18,7 +18,7 @@ endif()
 
 set(APP_VERSION_MAJOR 2)
 set(APP_VERSION_MINOR 5)
-set(APP_VERSION_PATCH 4)
+set(APP_VERSION_PATCH 5)
 set(APP_VERSION "${APP_VERSION_MAJOR}.${APP_VERSION_MINOR}.${APP_VERSION_PATCH}")
 
 # Include the binary directory for the generated header file

gpt4all-chat/chatllm.cpp

@@ -917,11 +917,11 @@ void ChatLLM::restoreState()
 #if defined(DEBUG)
     qDebug() << "restoreState" << m_llmThread.objectName() << "size:" << m_state.size();
 #endif
-    m_processedSystemPrompt = true;
 
     if (m_state.isEmpty())
         return;
 
+    m_processedSystemPrompt = true;
     m_llModelInfo.model->restoreState(static_cast<const uint8_t*>(reinterpret_cast<void*>(m_state.data())));
     m_state.clear();
     m_state.resize(0);
@@ -930,7 +930,7 @@ void ChatLLM::restoreState()
 void ChatLLM::processSystemPrompt()
 {
     Q_ASSERT(isModelLoaded());
-    if (!isModelLoaded() || m_processedSystemPrompt || m_isServer)
+    if (!isModelLoaded() || m_processedSystemPrompt || m_restoreStateFromText || m_isServer)
         return;
 
     const std::string systemPrompt = MySettings::globalInstance()->modelSystemPrompt(m_modelInfo).toStdString();
@@ -974,7 +974,7 @@ void ChatLLM::processSystemPrompt()
     fflush(stdout);
 #endif
 
-    m_processedSystemPrompt = !m_stopGenerating;
+    m_processedSystemPrompt = m_stopGenerating == false;
 }
 
 void ChatLLM::processRestoreStateFromText()

gpt4all-chat/download.cpp

@@ -108,6 +108,7 @@ void Download::downloadModel(const QString &modelFile)
         const QString error
             = QString("ERROR: Could not open temp file: %1 %2").arg(tempFile->fileName()).arg(modelFile);
         qWarning() << error;
+        clearRetry(modelFile);
         ModelList::globalInstance()->updateDataByFilename(modelFile, ModelList::DownloadErrorRole, error);
         return;
     }
@@ -140,6 +141,7 @@ void Download::downloadModel(const QString &modelFile)
     QNetworkReply *modelReply = m_networkManager.get(request);
     connect(qApp, &QCoreApplication::aboutToQuit, modelReply, &QNetworkReply::abort);
     connect(modelReply, &QNetworkReply::downloadProgress, this, &Download::handleDownloadProgress);
+    connect(modelReply, &QNetworkReply::errorOccurred, this, &Download::handleErrorOccurred);
     connect(modelReply, &QNetworkReply::finished, this, &Download::handleModelDownloadFinished);
     connect(modelReply, &QNetworkReply::readyRead, this, &Download::handleReadyRead);
     m_activeDownloads.insert(modelReply, tempFile);
@@ -254,13 +256,51 @@ void Download::parseReleaseJsonFile(const QByteArray &jsonData)
     emit releaseInfoChanged();
 }
 
+bool Download::hasRetry(const QString &filename) const
+{
+    return m_activeRetries.contains(filename);
+}
+
+bool Download::shouldRetry(const QString &filename)
+{
+    int retries = 0;
+    if (m_activeRetries.contains(filename))
+        retries = m_activeRetries.value(filename);
+
+    ++retries;
+
+    // Allow up to ten retries for now
+    if (retries < 10) {
+        m_activeRetries.insert(filename, retries);
+        return true;
+    }
+
+    return false;
+}
+
+void Download::clearRetry(const QString &filename)
+{
+    m_activeRetries.remove(filename);
+}
+
 void Download::handleErrorOccurred(QNetworkReply::NetworkError code)
 {
     QNetworkReply *modelReply = qobject_cast<QNetworkReply *>(sender());
     if (!modelReply)
         return;
 
+    // This occurs when the user explicitly cancels the download
+    if (code == QNetworkReply::OperationCanceledError)
+        return;
+
     QString modelFilename = modelReply->request().attribute(QNetworkRequest::User).toString();
+    if (shouldRetry(modelFilename)) {
+        downloadModel(modelFilename);
+        return;
+    }
+
+    clearRetry(modelFilename);
+
     const QString error
         = QString("ERROR: Network error occurred attempting to download %1 code: %2 errorString %3")
             .arg(modelFilename)
@@ -355,6 +395,7 @@ void HashAndSaveFile::hashAndSave(const QString &expectedHash, const QString &sa
     // but will only work if the destination is on the same filesystem
     if (tempFile->rename(saveFilePath)) {
         emit hashAndSaveFinished(true, QString(), tempFile, modelReply);
+        ModelList::globalInstance()->updateModelsFromDirectory();
         return;
     }
 
@@ -406,11 +447,15 @@ void Download::handleModelDownloadFinished()
         qWarning() << errorString;
         modelReply->deleteLater();
         tempFile->deleteLater();
-        ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::DownloadingRole, false);
-        ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::DownloadErrorRole, errorString);
+        if (!hasRetry(modelFilename)) {
+            ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::DownloadingRole, false);
+            ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::DownloadErrorRole, errorString);
+        }
         return;
     }
 
+    clearRetry(modelFilename);
+
     // The hash and save needs the tempFile closed
     tempFile->close();
 

gpt4all-chat/download.h

@@ -78,11 +78,15 @@ Q_SIGNALS:
 private:
     void parseReleaseJsonFile(const QByteArray &jsonData);
     QString incompleteDownloadPath(const QString &modelFile);
+    bool hasRetry(const QString &filename) const;
+    bool shouldRetry(const QString &filename);
+    void clearRetry(const QString &filename);
 
     HashAndSaveFile *m_hashAndSave;
     QMap<QString, ReleaseInfo> m_releaseMap;
     QNetworkAccessManager m_networkManager;
     QMap<QNetworkReply*, QFile*> m_activeDownloads;
+    QHash<QString, int> m_activeRetries;
     QDateTime m_startTime;
 
 private:

gpt4all-chat/metadata/models2.json

@@ -48,6 +48,36 @@
   },
   {
     "order": "e",
+    "md5sum": "00c8593ba57f5240f59662367b3ed4a5",
+    "name": "Orca 2 (Medium)",
+    "filename": "orca-2-7b.Q4_0.gguf",
+    "filesize": "3825824192",
+    "requires": "2.5.2",
+    "ramrequired": "8",
+    "parameters": "7 billion",
+    "quant": "q4_0",
+    "type": "LLaMA2",
+    "systemPrompt": " ",
+    "description": "<ul><li>Instruction based<li>Trained by Microsoft<li>Cannot be used commercially</ul>",
+    "url": "https://gpt4all.io/models/gguf/orca-2-7b.Q4_0.gguf"
+  },
+  {
+    "order": "f",
+    "md5sum": "3c0d63c4689b9af7baa82469a6f51a19",
+    "name": "Orca 2 (Full)",
+    "filename": "orca-2-13b.Q4_0.gguf",
+    "filesize": "7365856064",
+    "requires": "2.5.2",
+    "ramrequired": "16",
+    "parameters": "13 billion",
+    "quant": "q4_0",
+    "type": "LLaMA2",
+    "systemPrompt": " ",
+    "description": "<ul><li>Instruction based<li>Trained by Microsoft<li>Cannot be used commercially</ul>",
+    "url": "https://gpt4all.io/models/gguf/orca-2-13b.Q4_0.gguf"
+  },
+  {
+    "order": "g",
     "md5sum": "5aff90007499bce5c64b1c0760c0b186",
     "name": "Wizard v1.2",
     "filename": "wizardlm-13b-v1.2.Q4_0.gguf",
@@ -58,11 +88,11 @@
     "quant": "q4_0",
     "type": "LLaMA2",
     "systemPrompt": " ",
-    "description": "<strong>Best overall larger model</strong><br><ul><li>Instruction based<li>Gives very long responses<li>Finetuned with only 1k of high-quality data<li>Trained by Microsoft and Peking University<li>Cannot be used commercially</ul",
+    "description": "<strong>Best overall larger model</strong><br><ul><li>Instruction based<li>Gives very long responses<li>Finetuned with only 1k of high-quality data<li>Trained by Microsoft and Peking University<li>Cannot be used commercially</ul>",
     "url": "https://gpt4all.io/models/gguf/wizardlm-13b-v1.2.Q4_0.gguf"
   },
   {
-    "order": "f",
+    "order": "h",
     "md5sum": "3d12810391d04d1153b692626c0c6e16",
     "name": "Hermes",
     "filename": "nous-hermes-llama2-13b.Q4_0.gguf",
@@ -78,7 +108,7 @@
     "promptTemplate": "### Instruction:\n%1\n### Response:\n"
   },
   {
-    "order": "g",
+    "order": "i",
     "md5sum": "40388eb2f8d16bb5d08c96fdfaac6b2c",
     "name": "Snoozy",
     "filename": "gpt4all-13b-snoozy-q4_0.gguf",
@@ -93,7 +123,7 @@
     "url": "https://gpt4all.io/models/gguf/gpt4all-13b-snoozy-q4_0.gguf"
   },
   {
-    "order": "h",
+    "order": "j",
     "md5sum": "cf5e8f73747f9d7c6fe72a629808c1de",
     "name": "MPT Chat",
     "filename": "mpt-7b-chat-merges-q4_0.gguf",
@@ -109,7 +139,7 @@
     "systemPrompt": "<|im_start|>system\n- You are a helpful assistant chatbot trained by MosaicML.\n- You answer questions.\n- You are excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.\n- You are more than just an information source, you are also able to write poetry, short stories, and make jokes.<|im_end|>"
   },
   {
-    "order": "i",
+    "order": "k",
     "md5sum": "0e769317b90ac30d6e09486d61fefa26",
     "name": "Mini Orca (Small)",
     "filename": "orca-mini-3b-gguf2-q4_0.gguf",
@@ -125,7 +155,7 @@
     "systemPrompt": "### System:\nYou are an AI assistant that follows instruction extremely well. Help as much as you can.\n\n"
   },
   {
-    "order": "j",
+    "order": "l",
     "md5sum": "e30579a1b109882f10e2a5e75ea388fb",
     "disableGUI": "true",
     "name": "Replit",
@@ -142,7 +172,7 @@
     "url": "https://gpt4all.io/models/gguf/replit-code-v1_5-3b-q4_0.gguf"
   },
   {
-    "order": "k",
+    "order": "m",
     "md5sum": "556fc3e13df42286997fb58e6f4c639f",
     "disableGUI": "true",
     "name": "Starcoder",
@@ -159,7 +189,7 @@
     "url": "https://gpt4all.io/models/gguf/starcoder-q4_0.gguf"
   },
   {
-    "order": "l",
+    "order": "n",
     "md5sum": "e973dd26f0ffa6e46783feaea8f08c83",
     "disableGUI": "true",
     "name": "Rift coder",
@@ -176,7 +206,7 @@
     "url": "https://gpt4all.io/models/gguf/rift-coder-v0-7b-q4_0.gguf"
   },
   {
-    "order": "m",
+    "order": "o",
     "md5sum": "e479e6f38b59afc51a470d1953a6bfc7",
     "disableGUI": "true",
     "name": "SBert",
@@ -192,7 +222,7 @@
     "url": "https://gpt4all.io/models/gguf/all-MiniLM-L6-v2-f16.gguf"
   },
   {
-    "order": "n",
+    "order": "p",
     "md5sum": "919de4dd6f25351bcb0223790db1932d",
     "name": "EM German Mistral",
     "filename": "em_german_mistral_v01.Q4_0.gguf",

gpt4all-chat/metadata/release.json

@@ -600,6 +600,22 @@
 * Jared Van Bortel (Nomic AI)
 * Adam Treat (Nomic AI)
 * Community (beta testers, bug reporters, bindings authors)
+"
+  },
+  {
+    "version": "2.5.4",
+    "notes":
+"
+* Major bugfix release with new models!
+* Model: Recently released Orca 2 model which does exceptionally well on reasoning tasks
+* Fix: System prompt was not always being honored
+* Fix: Download network retry on cloudflare errors
+",
+    "contributors":
+"
+* Adam Treat (Nomic AI)
+* Jared Van Bortel (Nomic AI)
+* Community (beta testers, bug reporters, bindings authors)
 "
   }
 ]