Files

feat(api): major refactor from TCP socket to HTTP REST via Spring Boot

This commit introduces a large-scale refactor that replaces the existing TCP-based API with a Spring Boot-powered HTTP REST architecture. Due to the size and scope of this change, only essential structural notes are included below.

High-level changes:
- Replaced TCP-based communication with RESTful endpoints
- Introduced Spring Boot for API handling and configuration
- Refactored internal core logic to support REST architecture

New/Updated API components:
- `APIApplication.java`: Main Spring Boot entry point
- `MessageController.java`: Handles LLM-related queries
- `ToolController.java`: Handles adding/removing tools
- `NewToolRequest.java` / `NewToolResponse.java`: Data models for tool addition
- `NewQueryResponseHook.java`: Webhook handler for LLM query results
- `WebhookError.java`: Model for reporting webhook errors
- `EnableIfNotDisplay.java`: Conditional configuration for TTY context
- Other supporting classes (e.g., `ToolArgument`, `ToolRequest`)

Core changes:
- `Core.java`: Removed deprecated `addFunctionTool`, added `removeTool`
- `LaunchOptions.java`: Added `notDisplay` flag for headless operation
- `OllamaObject.java`: Implements tool removal logic

Launcher/display changes:
- `Launcher.java`: Starts `APIApplication` if not in TTY mode
- `Display.java`: Integrates REST API contextually with TTY display

NOTE: Several classes are included but not yet fully utilized; these are placeholders for upcoming features (e.g., `MessageResponse`, `ToolRequest`).

BREAKING CHANGE: This refactors removes all TCP-based API code and replaces it with HTTP REST using Spring Boot. Any clients or modules depending on the old TCP interface will need to be updated.

2025-05-24 18:11:58 +02:00

2.5 KiB

Raw Blame History

Chat thing

(better name pending probably)

What is it?

Well it's basically a font end for Ollama like Open-WebUI, but meant to run a local model and open up a somewhat easy to use API system for tools.

Use cases?

While the primary goal is providing a somewhat more modular frontend for Ollama, you can also use this system to integrate AI into your own application, For example AI could act as a player in a game or interact with external system with the use of tools via the API

simple examples are in the Display module where i gave it the ability to access python through docker and get the current date and time both with the OllamaFunctionTool thru my Ollama framework

API

The documentation for the API is available at the gitea wiki under API docs

How to run?

To run you need to build the launcher module
$ ./gradlew :launcher:shadowJar
This will put the launcher-1.0-all.jar(or similar, based on the version) in ./launcher/build/libs, this can now be copied anywhere and can be ran with
$ java -jar launcher-1.0-all.jar
you can use the -h argument to get a help message.

Launch options for NeuroChat:
-h  --help      Provides this help message
-s  --server    Starts the application as API server
-p  --port      Provides the port number that the API server should use, defaults to 39075
-o  --output    Redirects the API Server output to another file
    --api       Provides API docs

How to build? And general information

If you only want to build a specific module, you can use the same command as listed in How to run? but replace launcher with the desire module.

API: $ ./gradlew :API:shadowJar
- Depends on Core
Display: $ ./gradlew :Display:shadowJar
- Depends on Core and MALAPITool
Core: $ ./gradlew :Core:shadowJar
- However, this one is kinda useless unless you want to directly implement the system into your application
MALAPITool: $ ./gradlew :MALAPITool:shadowJar
- Please read MALAPITool README.md
Launcher: $ ./gradlew :launcher:shadowJar
- Depends on API, Display and Core
- This is the main module that runs the application and starts the API server

2.5 KiB Raw Blame History