Revert "rag (#7)"

This reverts commit 9c703bd.

Author: david20571015

.dockerignore

@@ -1,4 +1,6 @@
 .DS_Store
+configs/*
+!configs/example.toml
 llm_backend/protos/
 
 # Byte-compiled / optimized / DLL files

.gitignore

@@ -11,7 +11,7 @@ uv run gen-protos
 
 ## Usage
 
-Please configure the `configs/config.toml` file.
+Please configure the `configs/config.toml` file (refer to `configs/example.toml` for the options).
 The following environment variables are required (`export` them or place them in a `.env` file):
 
 - `OPENAI_API_KEY`: Your ChatGPT API key.
@@ -23,12 +23,6 @@ The following environment variables are required (`export` them or place them in
 python3 scripts/serve.py --config configs/config.toml
 ```
 
-You can refer to `scripts/client.py` for an example implementation of a client:
-
-```shell
-python3 scripts/client.py
-```
-
 ## Features
 
 Refer to the protobuf files in the `protos/` directory for the features provided by the server.

Dockerfile

@@ -0,0 +1,43 @@
+[server]
+host = 'localhost'
+port = 50051
+max_workers = 10
+
+[service.search.embeddings]
+# Name of embedding model. All available models can be found [here](https://huggingface.co/models?language=zh)
+model = 'moka-ai/m3e-base'
+
+[service.search.query]
+# The template must contain the `{keywords}` placeholder.
+prompt_template = 'Please search for the content related to the following keywords: {keywords}.'
+similarity_top_k = 3
+
+[service.summarize.chatgpt]
+model = 'o3-mini'
+
+[service.summarize.query]
+system_template = """
+You are an expert Q&A system that is trusted around the world.
+Always answer the query using the provided context information, and not prior knowledge.
+Some rules to follow:
+1. Never directly reference the given context in your answer.
+2. Avoid statements like 'Based on the context, ...' or The context information ...' or anything along those lines."""
+
+# The template must contain the `{context_str}` and `{query_str}` placeholders.
+user_template = """
+Context information from multiple sources is below.
+---------------------
+{context_str}
+---------------------
+Given the information from multiple sources and not prior knowledge, answer the query.
+Query: {query_str}
+Answer: """
+
+# The content of `{query_str}` placeholder in the user template.
+query_str = '請用繁體中文總結這幾篇新聞。'
+
+# The transform function from the request strings to the query strings.
+# Must be one of:
+# - 'plain': The query string is the same as the request string.
+# - 'numbered': Add a number (1., 2., ...) to the beginning of each request string.
+content_format = 'plain'

README.md

@@ -1,9 +1,18 @@
 import os
 
-import grpc
+from grpc._server import _Server
 from pydantic import BaseModel
 
-from .rag import RagConfig, RagService, add_RagServiceServicer_to_server
+from .search import (
+    SearchService,
+    add_SearchServiceServicer_to_server,
+)
+from .search.config import SearchConfig
+from .summarize import (
+    SummarizeService,
+    add_SummarizeServiceServicer_to_server,
+)
+from .summarize.config import SummarizeConfig
 
 
 class ServerConfig(BaseModel):
@@ -12,11 +21,21 @@ class ServerConfig(BaseModel):
     max_workers: int = (os.cpu_count() or 1) * 5
 
 
+class ServiceConfig(BaseModel):
+    search: SearchConfig
+    summarize: SummarizeConfig
+
+
 class Config(BaseModel):
     server: ServerConfig
-    service: RagConfig
+    service: ServiceConfig
+
+
+def setup_search_service(config: Config, server: _Server):
+    search_service = SearchService(config.service.search)
+    add_SearchServiceServicer_to_server(search_service, server)
 
 
-def setup_rag_service(config: Config, server: grpc.aio.Server):
-    rag_service = RagService(config.service)
-    add_RagServiceServicer_to_server(rag_service, server)
+def setup_summarize_service(config: Config, server: _Server):
+    summarize_service = SummarizeService(config.service.summarize)
+    add_SummarizeServiceServicer_to_server(summarize_service, server)