Announcement

I4.0 v17.0.0 is under development!
It's a fully re-written version which is fully modular, it allows to import and delete modules very easy, and much more interesting features!

Soon a new repository or branch in this repository will be open.
It will still take a time for it to be 100% usable by both client and server.

v16.0.0

New

• Added KokoroTTS support for the text2audio service.
• Added a chunks parameter in the speech2text service.
• Added a new (server-side) parameter for the speech2text service.
• New (client-side) configuration parameters:
  • SimulatedSensedInUserPrompt: If true this will write the simulated vision and simulated audition results in the user prompt instead of using the system prompt.
  • Chatbot_UseReasoning: Allow the usage of reasoning in the chatbot model or select a reasoning level (if the model and server allows it).
  • Chatbot_AllowMemoriesUsage: Enable or disable the memories.
  • Chatbot_ToolsInSystemPrompt: Write the tools in the system prompt instead of using the special tools argument. Might work better for some models or with some prompts.
  • SimulatedVision_v1_Image2Text_Server: The server to use for the img2text service during the Simulated Vision V1.
  • SimulatedVision_v1_ObjectDetection_Server: The server to use for the od service during the Simulated Vision V1.
  • SimulatedAudition_v1_SpeechToText_Server: The server to use for the speech2text service during the Simulated Audition V1.
  • Image2Text_MaxLength: Max length for the img2text service.
  • Text2Audio_TTS_Voice: Voice to use for the TTS.
  • Text2Audio_TTS_Speed: Voice to use for the TTS.
  • See the documentation for more info!
• New tool: internet-chatbot. It allows the current chatbot to ask a prompt or redirect your prompt to a chatbot in internet, example: gpt-5-nano, deepseek-r1. Be aware that this uses a third-party chatbot, so your prompt will not be encrypted and the third-party can collect your data.
• Added typical_p parameter for the hf chatbot.
• Implemmented Flash Attention 2 for some services besides chatbot.
• Added Simulated Audition V1 support for the videos in the Simulated Vision V2.

Updated / Fixed

• Updated documentation
• Fixed a typo in the speech2text service using the Whisper library.
• Renamed some files.
• Fixed some server-side bugs.
• Updated requirements.

Removed

• Removed SSL support due to problems and not being able to test it. This shouldn't really be a big probles because the prompt (unless you disable it manually) will be sent encrypted to the server and the response of the server will also be encrypted.

Compatibility

All the v16 servers are compatible with v16 clients.
All the v16 clients are compatible with v16 servers.

Older clients/servers should be compatible with newer clients/servers, but new functions will not be available.

Note

The configuration file in the server (if updating the sever from an older version) should be edited manually.

The way that the servers handle the pricing of the services might change in a future version.

v15.2.0

New

• Added multimodal chatbot support with lcpp. You can now run GGUF/Quantized multimodal chatbots (only with image support, audio and video doesn't work for now)!
• hf chatbots (for both text-only and multimodal) now support tools without using the system prompt (experimental; not tested).
• New parameters for chatbots.
• New automatic installation script! See documentation for more info.

Updated / Fixed

• Updated documentation.
• Fixed some bugs with encryption.
• Fixed some client-side bugs.
• Fixed bugs related to Simulated Vision V1 & V2 and Simulated Audition V1.
• Updated client requirements.
• Fixed bugs related to the RVC service, with NVIDIA series 50 GPUs or torch>=2.6. This is not fully fixed yet, please see this issue at the RVC project and manually change the fairseq code following this message.
• Fixed some bugs related to the speech2text service.
• Fixed minor server bugs.
• Now the time it takes for the model to generate it's first token doesn't apply to the queue time (only for the chatbot service).
• Updated some tools for better understanding. Also, now the model can write and see multiples URLs in the internet_url tool.
• Changed the way the server process messages (experimental; testing).
• Updated requirements.
• Modified the way the server handles temporal files.

Removed

• Removed pre-defined models (for both LCPP and SDCPP).
• Removed GPT4All chatbot.
• Removed old automatic installation script.
• The chatbot now doesn't print the response to the server's CLI.
• Removed automatic clear cache, since most services doesn't use temporal files anymore.
• Removed the GlobalConnection class of the server; the server will only have WebSockets support for now.

Compatibility

All the v15 servers are compatible with v15 clients.
All the v15 clients are compatible with v15 servers.

Older clients/servers should be compatible with newer clients/servers, but new functions will not be available.

v15.1.0

Updated

• Improved tool understanding.

Compatibility

All the v15 servers are compatible with v15 clients.
All the v15 clients are compatible with v15 servers.

Older clients/servers should be compatible with newer clients/servers, but new functions will not be available.

Warning

DEVELOPERS USING THE I4_0-Client-PY<15.0.0 PACKAGE MIGHT NEED TO CHANGE SOME THINGS IF THEY UPDATE TO THE NEW VERSION.
Also, this version has not been fully tested.

v15.0.0

New

• New server and client configuration.
• The client can now choose to not encrypt the data.
• New script for parsing GGUF pre-defined models.
• Chatbot models can now use a pre-defined seed. See documentation for details.
• The server can now use SSL for the WebSockets.
• Created a class for managing multiple connection methods in the server. Expect new ways to connect to the server in the future.
• New index -2, to get the first cheapest model of the server.
• MinP and TypicalP configurable parameters for the chatbot, in both client and server.
• I4.0 can now generate XLSX files from a CSV code.
• I4.0 can now understand PDF, CSV, and XLSX files from the user.

Updated / Fixed

• Updated documentation.
• Updated all the services that supports files to save the files in a buffer, not in the disk.
• Updated some tools names.
• Removed that I4.0 generates a prompt for searching the internet. The prompt from the user will be used from now on.
• Updated some tools parameters for the models. Also improved chatbot tools understanding.
• Configuration will not be saved after closing the server. You can now change the configuration while the server is running, but a server restart is required to apply the changes.
• Bug fixes.
• Updated requirements.
• Hybrid encryption is now used for the server database.

Compatibility

All the v15 servers are compatible with v15 clients.
All the v15 clients are compatible with v15 servers.

Older clients/servers should be compatible with newer clients/servers, but new functions will not be available.

Warning

DEVELOPERS USING THE I4_0-Client-PY<15.0.0 PACKAGE MIGHT NEED TO CHANGE SOME THINGS IF THEY UPDATE TO THE NEW VERSION.
Also, this version has not been fully tested.

v14.1.0

New

• New hybrid encryption method, for both server and client. Much faster than RSA.
• text2image (SD-CPP) now uses random seeds for each image.
• New tool: document-creation. I4.0 can now create .docx and .pdf files!
• Google search engine has been integrated. If DuckDuckGo throws and error, it uses Google. This can be customizable in the server configuration (see the documentation for more info).
• Data save can now have a file limit support. If file size > the limit, it doesn't save it.

Updated / Fixed

• Updated documentation.
• Updated client connection to the server.
• text2audio service will now be able to use multiple libraries, other than transformers.
• Bug fixes.
• Updated server code.
• Updated tools description.
• Bug fixes in the automated installation script.
• Updated requirements.

Compatibility

All the v14 servers are compatible with v14 clients.
All the v14 clients are compatible with v14 servers.

Warning

DEVELOPERS USING THE I4_0-Client-PY<14.0.0 PACKAGE WILL NEED TO CHANGE SOME THINGS IF THEY UPDATE TO THE NEW VERSION.

v14.0.0

New

• New system for managing temporal files.
• New queue system.
• Data save.
• Response cache. WE ARE STILL WORKING ON THIS.
• New configuration (both client and server).
• Internet research.
• More documentation.
• New parameters available for the client: TopP, TopK.
• Some tools may require tokens in the API key.
• New response parameter: errors.
• New tools.
• New assets.
• Added a lot of new parameters for the lcpp chatbots.
• More pre-defined models for the lcpp chatbot.
• Official I4.0 client, in CLI. Check out the I4.0 Client CLI!.

Updated / Fixed

• Fixed a bug related to the memories.
• Updated websockets library.
• Fixed bugs related with the queue.
• Changed some client-side configuration parameters.
• Updated automated installation script.
• Updated requirements.
• Fixed bugs related with the API key creation.
• Fixed bugs related with the hf chatbot (both text-only and multimodal).
• From now on, gpt4all chatbots are deprecated.
• Updated a lot of services to use temporal files.
• Updated README.md.
• Updated low-level client.
• Updated high-level client.
• Updated client config to be in a class, so it can be more easy to save into a JSON or dictionary.

Compatibility

All the v14 servers are compatible with v14 clients.
All the v14 clients are compatible with v14 servers.

Warning

DEVELOPERS USING THE I4_0-Client-PY<14.0.0 PACKAGE WILL NEED TO CHANGE SOME THINGS IF THEY UPDATE TO THE NEW VERSION.

v13.2.1

Fixed

• Fixed a small bug when using a different dtype in the service de.

Compatibility

• All the v13 servers are compatible with v13 clients.
• All the v13 clients are compatible with v13 servers.

Warning

DEVELOPERS USING THE I4_0-Client-PY<13.2.0 PACKAGE WILL NEED TO CHANGE SOME THINGS IF THEY UPDATE TO THE NEW VERSION.

v13.2.0

New

• Temporal conversations: conversations can now be automatically deleted after any amount of years, months, days, hours, or minutes.
  • For clients without a valid API key (or no API key at all), the conversation will be automatically deleted based on the server configuration.
  • For clients with a valid API key, conversations will not be deleted by default unless specified by the key.
• New tools for the chatbot:
  • memory-edit: edit previously created memories.
  • memory-delete: delete previously created memories.
• Client-customizable maximum response length.
• Client-customizable model temperature.
• The server can now use dml for the RVC service.
• New client configuration parameters:
  • MaxLength: changes the maximum length of the response.
  • Temperature: adjusts the model's temperature.
• Simulated Vision V2: text-only chatbots can now understand videos! The old simulated vision (for images) has been renamed to Simulated Vision V1.
  • WARNING: This feature is experimental and may contain bugs.
• Simulated Audition V1: text-only chatbots can now understand audio files!
  • WARNING: This feature is experimental and may contain bugs.

Updated / Fixed

• Updated chatbot tools: improved understanding and tool invocation.
  • Clients can now add custom tools that the chatbot can call.
• Updated multimodal parameter:
  • Previously known as allows_files. If set to true, clients could send images, videos, audios, etc., even if the model didn’t support them. Now, the server owner can specify supported file types in the configuration, separated by spaces.
  • Examples:
    • "multimodal": "" → text only.
    • "multimodal": "image" → supports images and text.
    • "multimodal": "image video" → supports images, videos, and text.
    • "multimodal": "video audio" → supports videos, audios, and text.
  • Supported file types for now: image, video, audio.
• Updated encryption: now faster.
• Updated automatic server installation script with minor bug fixes.
• Updated requirements.
• Fixed a bug in several services when the server used a custom dtype:
  • Affected services: qa, chatbot (hf), translation (translators), nsfw_filter (text), multimodal chatbot (hf), de, od, text2audio.
• Updated documentation.
• Split client-server connection logic into separate files:
  • ServerConnection.py: basic and low-level connection to the server.
  • Utils/Recorder.py: records audio from microphone or system.
  • Utils/ServerUtilities.py: high-level server connection utilities.

Compatibility

• All v13 servers are compatible with v13 clients.
• All v13 clients are compatible with v13 servers.

Warning

DEVELOPERS USING THE I4_0-Client-PY<13.2.0 PACKAGE WILL NEED TO MAKE SOME CHANGES IF UPDATING TO THIS VERSION.

v13.1.0

Fixed

• Fixed a small bug in the client.
• Fixed a bug that saves the conversation files.
• Fixed API key bugs.

Updated

• New predefined models for Llama-CPP.
• Replaced the default internet_chat in the configuration.
• Updated database server.
• Updated the automatic requirements installation script.
• Updated requirements.

Compatibility

• All the v13 servers are compatible with v13 clients.
• All the v13 clients are compatible with v13 servers.