feat: Add task processing API

Signed-off-by: provokateurin <kate@provokateurin.de>

appinfo/info.xml

@@ -29,6 +29,7 @@ See [the nextcloud admin docs](https://docs.nextcloud.com/server/latest/admin_ma
 		</docker-install>
 		<scopes>
 			<value>AI_PROVIDERS</value>
+			<value>TASK_PROCESSING</value>
 		</scopes>
 		<system>false</system>
 	</external-app>

lib/main.py

@@ -1,20 +1,20 @@
 """Tha main module of the llm2 app
 """
 
-import queue
 import threading
-import typing
+import time
 from contextlib import asynccontextmanager
 from time import perf_counter
 
-import pydantic
-from chains import generate_chains
-from fastapi import Depends, FastAPI, responses
+from fastapi import FastAPI
 from nc_py_api import AsyncNextcloudApp, NextcloudApp
-from nc_py_api.ex_app import LogLvl, anc_app, run_app, set_handlers
+from nc_py_api.ex_app import LogLvl, run_app, set_handlers
+
+from chains import generate_chains
 
 chains = generate_chains()
 
+
 @asynccontextmanager
 async def lifespan(_app: FastAPI):
     set_handlers(
@@ -27,57 +27,54 @@ async def lifespan(_app: FastAPI):
 
 
 APP = FastAPI(lifespan=lifespan)
-TASK_LIST: queue.Queue = queue.Queue(maxsize=100)
 
 
 class BackgroundProcessTask(threading.Thread):
     def run(self, *args, **kwargs):  # pylint: disable=unused-argument
+        nc = NextcloudApp()
+
+        provider_ids = set()
+        task_type_ids = set()
+        for chain_name, _ in chains.items():
+            provider_ids.add("llm2:" + chain_name)
+            (model, task) = chain_name.split(":", 2)
+            task_type_ids.add("core:text2text:" + task)
+
         while True:
-            task = TASK_LIST.get(block=True)
+            response = nc.providers.task_processing.next_task(list(provider_ids), list(task_type_ids))
+            if not response:
+                time.sleep(5)
+                continue
+
+            task = response["task"]
+            provider = response["provider"]
+
             try:
-                chain_name = task.get("chain")
+                chain_name = provider["name"][5:]
                 print(f"chain: {chain_name}", flush=True)
                 chain_load = chains.get(chain_name)
                 if chain_load is None:
-                    NextcloudApp().providers.text_processing.report_result(
-                        task["id"], error="Requested model is not available"
+                    NextcloudApp().providers.task_processing.report_result(
+                        task["id"], error_message="Requested model is not available"
                     )
                     continue
                 chain = chain_load()
                 print("Generating reply", flush=True)
                 time_start = perf_counter()
-                print(task.get("prompt"))
-                result = chain.invoke(task.get("prompt")).get("text")
+                print(task.get("input").get("input"))
+                result = chain.invoke(task.get("input").get("input")).get("text")
                 del chain
                 print(f"reply generated: {round(float(perf_counter() - time_start), 2)}s", flush=True)
                 print(result, flush=True)
-                NextcloudApp().providers.text_processing.report_result(
+                NextcloudApp().providers.task_processing.report_result(
                     task["id"],
-                    str(result),
+                    {"output": str(result)},
                 )
             except Exception as e:  # noqa
                 print(str(e), flush=True)
                 nc = NextcloudApp()
                 nc.log(LogLvl.ERROR, str(e))
-                nc.providers.text_processing.report_result(task["id"], error=str(e))
-
-
-class Input(pydantic.BaseModel):
-    prompt: str
-    task_id: int
-
-
-@APP.post("/chain/{chain_name}")
-async def tiny_llama(
-    _nc: typing.Annotated[AsyncNextcloudApp, Depends(anc_app)],
-    req: Input,
-    chain_name=None,
-):
-    try:
-        TASK_LIST.put({"prompt": req.prompt, "id": req.task_id, "chain": chain_name}, block=False)
-    except queue.Full:
-        return responses.JSONResponse(content={"error": "task queue is full"}, status_code=429)
-    return responses.Response()
+                nc.providers.task_processing.report_result(task["id"], error_message=str(e))
 
 
 async def enabled_handler(enabled: bool, nc: AsyncNextcloudApp) -> str:
@@ -86,17 +83,16 @@ async def enabled_handler(enabled: bool, nc: AsyncNextcloudApp) -> str:
         for chain_name, _ in chains.items():
             (model, task) = chain_name.split(":", 2)
             try:
-                await nc.providers.text_processing.register(
-                    "llm2:"+chain_name, "Local Large language Model: " + model, "/chain/" + chain_name, task
+                await nc.providers.task_processing.register(
+                    "llm2:" + chain_name, "Local Large language Model: " + model, "core:text2text:" + task
                 )
-                print(f"Registering {model} - {task}", flush=True)
+                print(f"Registering {chain_name}", flush=True)
             except Exception as e:
                 print(f"Failed to register", f"{model} - {task}", f"Error:", f"{e}\n", flush=True)
     else:
         for chain_name, chain in chains.items():
-            (model, task) = chain_name.split(":", 2)
-            await nc.providers.text_processing.unregister(model)
-            print(f"Unregistering {model} - {task}", flush=True)
+            await nc.providers.task_processing.unregister("llm2:" + chain_name)
+            print(f"Unregistering {chain_name}", flush=True)
     return ""
 
 

pyproject.toml

@@ -11,11 +11,11 @@ package-mode = false
 python = "^3.10"
 pydantic = "^2.8.2"
 fastapi = "^0.111.0"
-nc-py-api = {extras = ["app"], version = "^0.13.0"}
+nc-py-api = {extras = ["app"], version = "^0.14.0"}
 langchain = "^0.1.0"
 llama-cpp-python = "0.2.76"
 
 
 [build-system]
 requires = ["poetry-core"]
-build-backend = "poetry.core.masonry.api"
+build-backend = "poetry.core.masonry.api"