mirror of
https://github.com/tiennm99/DocsGPT.git
synced 2026-10-03 11:11:58 +00:00
Alex
574f96341e
The backend import package is now docsgpt, the name it will carry on PyPI; application was far too generic to install into anyone's site-packages. git mv plus a mechanical rewrite of every import, dotted string and path reference: 734 Python files, the compose files, Dockerfile, workflows, docs, setup scripts, devcontainer, k8s manifests, vscode config, pytest and coverage config, .gitignore. Behaviour is unchanged. Kept for one release: - A top-level application package whose meta-path finder resolves application.x.y to the already-imported docsgpt.x.y object, so old imports and entry points (celery -A application.app.celery, uvicorn application.asgi:asgi_app) keep working with a FutureWarning. - Celery registers every application.* task name as an alias of its docsgpt.* task on start-up, so messages queued by the previous release still run. The redbeat key prefix moves to redbeat:docsgpt:v2: so schedule entries the previous release wrote are left unread instead of firing twice. The backend image builds from the repository root (docker build -f docsgpt/Dockerfile .) so it can ship the alias package; a root .dockerignore allow-lists docsgpt/ and application/ and keeps caches, local data, .env files, the sample index files and the Dockerfile out. Compose and the image workflows point at the new context.
63 lines
2.1 KiB
Python
63 lines
2.1 KiB
Python
from docsgpt.llm.base import BaseLLM
|
|
from docsgpt.core.settings import settings
|
|
import threading
|
|
|
|
|
|
class LlamaSingleton:
|
|
_instances = {}
|
|
_lock = threading.Lock() # Add a lock for thread synchronization
|
|
|
|
@classmethod
|
|
def get_instance(cls, llm_name):
|
|
if llm_name not in cls._instances:
|
|
try:
|
|
from llama_cpp import Llama
|
|
except ImportError:
|
|
raise ImportError(
|
|
"Please install llama_cpp using pip install llama-cpp-python"
|
|
)
|
|
cls._instances[llm_name] = Llama(model_path=llm_name, n_ctx=2048)
|
|
return cls._instances[llm_name]
|
|
|
|
@classmethod
|
|
def query_model(cls, llm, prompt, **kwargs):
|
|
with cls._lock:
|
|
return llm(prompt, **kwargs)
|
|
|
|
|
|
class LlamaCpp(BaseLLM):
|
|
provider_name = "llama_cpp"
|
|
|
|
def __init__(
|
|
self,
|
|
api_key=None,
|
|
user_api_key=None,
|
|
llm_name=settings.LLM_PATH,
|
|
*args,
|
|
**kwargs,
|
|
):
|
|
super().__init__(*args, **kwargs)
|
|
self.api_key = api_key
|
|
self.user_api_key = user_api_key
|
|
self.llama = LlamaSingleton.get_instance(llm_name)
|
|
|
|
def _raw_gen(self, baseself, model, messages, stream=False, **kwargs):
|
|
context = messages[0]["content"]
|
|
user_question = messages[-1]["content"]
|
|
prompt = f"### Instruction \n {user_question} \n ### Context \n {context} \n ### Answer \n"
|
|
result = LlamaSingleton.query_model(
|
|
self.llama, prompt, max_tokens=150, echo=False
|
|
)
|
|
return result["choices"][0]["text"].split("### Answer \n")[-1]
|
|
|
|
def _raw_gen_stream(self, baseself, model, messages, stream=True, **kwargs):
|
|
context = messages[0]["content"]
|
|
user_question = messages[-1]["content"]
|
|
prompt = f"### Instruction \n {user_question} \n ### Context \n {context} \n ### Answer \n"
|
|
result = LlamaSingleton.query_model(
|
|
self.llama, prompt, max_tokens=150, echo=False, stream=stream
|
|
)
|
|
for item in result:
|
|
for choice in item["choices"]:
|
|
yield choice["text"]
|