chat history feature enable for v1.4 release

EdgeCraftRAG/Dockerfile.server

@@ -46,5 +46,4 @@ WORKDIR /home/user/
 RUN git clone https://github.com/openvinotoolkit/openvino.genai.git genai
 ENV PYTHONPATH="$PYTHONPATH:/home/user/genai/tools/llm_bench"
 
-
 ENTRYPOINT ["python", "-m", "edgecraftrag.server"]

EdgeCraftRAG/docker_image_build/build.yaml

@@ -4,31 +4,31 @@
 services:
   edgecraftrag-server:
     build:
-      context: ../
+      context: ..
       args:
         http_proxy: ${http_proxy}
         https_proxy: ${https_proxy}
       dockerfile: ./Dockerfile.server
     image: ${REGISTRY:-opea}/edgecraftrag-server:${TAG:-latest}
   edgecraftrag-ui:
     build:
-      context: ../
+      context: ..
       args:
         http_proxy: ${http_proxy}
         https_proxy: ${https_proxy}
       dockerfile: ./ui/docker/Dockerfile.ui
     image: ${REGISTRY:-opea}/edgecraftrag-ui:${TAG:-latest}
   edgecraftrag-ui-gradio:
     build:
-      context: ../
+      context: ..
       args:
         http_proxy: ${http_proxy}
         https_proxy: ${https_proxy}
       dockerfile: ./ui/docker/Dockerfile.gradio
     image: ${REGISTRY:-opea}/edgecraftrag-ui-gradio:${TAG:-latest}
   edgecraftrag:
     build:
-      context: ../
+      context: ..
       args:
         http_proxy: ${http_proxy}
         https_proxy: ${https_proxy}

EdgeCraftRAG/edgecraftrag/VERSION

@@ -0,0 +1 @@
+25.05-dev-0421

EdgeCraftRAG/edgecraftrag/api/v1/chatqna.py

@@ -5,6 +5,7 @@
 from comps.cores.proto.api_protocol import ChatCompletionRequest
 from edgecraftrag.api_schema import RagOut
 from edgecraftrag.context import ctx
+from edgecraftrag.utils import serialize_contexts
 from fastapi import FastAPI, File, HTTPException, UploadFile, status
 from fastapi.responses import StreamingResponse
 
@@ -29,14 +30,15 @@ async def retrieval(request: ChatCompletionRequest):
 @chatqna_app.post(path="/v1/chatqna")
 async def chatqna(request: ChatCompletionRequest):
     try:
+        request.messages = convert_message(request.messages)
         generator = ctx.get_pipeline_mgr().get_active_pipeline().generator
         if generator:
             request.model = generator.model_id
         if request.stream:
-            ret, retri_res = ctx.get_pipeline_mgr().run_pipeline(chat_request=request)
+            ret, contexts = ctx.get_pipeline_mgr().run_pipeline(chat_request=request)
             return ret
         else:
-            ret, retri_res = ctx.get_pipeline_mgr().run_pipeline(chat_request=request)
+            ret, contexts = ctx.get_pipeline_mgr().run_pipeline(chat_request=request)
             return str(ret)
     except Exception as e:
         raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=str(e))
@@ -46,7 +48,8 @@ async def chatqna(request: ChatCompletionRequest):
 @chatqna_app.post(path="/v1/ragqna")
 async def ragqna(request: ChatCompletionRequest):
     try:
-        res, retri_res = ctx.get_pipeline_mgr().run_pipeline(chat_request=request)
+        request.messages = convert_message(request.messages)
+        res, contexts = ctx.get_pipeline_mgr().run_pipeline(chat_request=request)
         if isinstance(res, GeneratedDoc):
             res = res.text
         elif isinstance(res, StreamingResponse):
@@ -55,10 +58,9 @@ async def ragqna(request: ChatCompletionRequest):
                 collected_data.append(chunk)
             res = "".join(collected_data)
 
-        ragout = RagOut(query=request.messages, contexts=[], response=str(res))
-        for n in retri_res:
-            origin_text = n.node.get_text()
-            ragout.contexts.append(origin_text.strip())
+        serialized_contexts = serialize_contexts(contexts)
+
+        ragout = RagOut(query=request.messages, contexts=serialized_contexts, response=str(res))
         return ragout
     except Exception as e:
         raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=str(e))
@@ -88,3 +90,34 @@ async def reset_prompt():
             return "Reset LLM Prompt Successfully"
     except Exception as e:
         raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=str(e))
+
+
+def convert_message(messages, history_prompt: str = None):
+    messages_list = []
+    if isinstance(messages, str):
+        str_message = messages
+    else:
+        str_message = ""
+        user_indexs = [i for i, msg in enumerate(messages) if msg.get("role") == "user"]
+        last_user_index = user_indexs[-1] if user_indexs else -1
+
+        for idx, message in enumerate(messages):
+            msg_role = message["role"]
+            if msg_role in ["user", "assistant"]:
+                content = message["content"]
+                if idx == last_user_index and msg_role == "user":
+                    messages_list.append(("system", f"{history_prompt}"))
+                if isinstance(content, str):
+                    messages_list.append((msg_role, content))
+                else:
+                    raise ValueError(
+                        status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail="Only text content is supported."
+                    )
+            else:
+                raise ValueError(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=f"Unknown role: {msg_role}")
+
+        for role, content in messages_list:
+            str_message += f"{role}: {content}\n"
+    if len(str_message) > 8192:
+        str_message = str_message[-8192:]
+    return str_message

EdgeCraftRAG/edgecraftrag/api_schema.py

@@ -1,7 +1,7 @@
 # Copyright (C) 2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
-from typing import Optional
+from typing import Any, Optional
 
 from pydantic import BaseModel
 
@@ -65,5 +65,5 @@ class FilesIn(BaseModel):
 
 class RagOut(BaseModel):
     query: str
-    contexts: Optional[list[str]] = None
+    contexts: Optional[dict[str, Any]] = None
     response: str

EdgeCraftRAG/edgecraftrag/components/data.py

@@ -54,7 +54,7 @@ def convert_text_to_documents(text) -> List[Document]:
 def convert_file_to_documents(file_path) -> List[Document]:
     from llama_index.core import SimpleDirectoryReader
 
-    supported_exts = [".pdf", ".txt", ".doc", ".docx", ".pptx", ".ppt", ".csv", ".md", ".html", ".rst"]
+    supported_exts = [".pdf", ".txt", ".doc", ".docx", ".pptx", ".ppt", ".csv", ".md", ".html", ".rst", ".epub"]
     if file_path.is_dir():
         docs = SimpleDirectoryReader(input_dir=file_path, recursive=True, required_exts=supported_exts).load_data()
     elif file_path.is_file():

EdgeCraftRAG/edgecraftrag/components/pipeline.py

@@ -221,11 +221,13 @@ async def timing_wrapper():
 
 def run_test_generator_ben(pl: Pipeline, chat_request: ChatCompletionRequest) -> Any:
     benchmark_index, benchmark_data = pl.benchmark.init_benchmark_data()
+    contexts = {}
     start = time.perf_counter()
     query = chat_request.messages
     retri_res = pl.retriever.run(query=query)
     query_bundle = QueryBundle(query)
     benchmark_data[CompType.RETRIEVER] = time.perf_counter() - start
+    contexts[CompType.RETRIEVER] = retri_res
 
     start = time.perf_counter()
     if pl.postprocessor:
@@ -236,6 +238,7 @@ def run_test_generator_ben(pl: Pipeline, chat_request: ChatCompletionRequest) ->
             ):
                 processor.top_n = chat_request.top_n
             retri_res = processor.run(retri_res=retri_res, query_bundle=query_bundle)
+            contexts[CompType.POSTPROCESSOR] = retri_res
     benchmark_data[CompType.POSTPROCESSOR] = time.perf_counter() - start
 
     if pl.generator is None:
@@ -260,12 +263,14 @@ def run_test_generator_ben(pl: Pipeline, chat_request: ChatCompletionRequest) ->
         benchmark_data[CompType.GENERATOR] = end - start
         pl.benchmark.insert_llm_data(benchmark_index, input_token_size)
         pl.benchmark.insert_benchmark_data(benchmark_data)
-    return ret, retri_res
+    return ret, contexts
 
 
 def run_test_generator(pl: Pipeline, chat_request: ChatCompletionRequest) -> Any:
     query = chat_request.messages
+    contexts = {}
     retri_res = pl.retriever.run(query=query)
+    contexts[CompType.RETRIEVER] = retri_res
     query_bundle = QueryBundle(query)
 
     if pl.postprocessor:
@@ -276,6 +281,7 @@ def run_test_generator(pl: Pipeline, chat_request: ChatCompletionRequest) -> Any
             ):
                 processor.top_n = chat_request.top_n
             retri_res = processor.run(retri_res=retri_res, query_bundle=query_bundle)
+            contexts[CompType.POSTPROCESSOR] = retri_res
 
     if pl.generator is None:
         raise ValueError("No Generator Specified")
@@ -286,4 +292,4 @@ def run_test_generator(pl: Pipeline, chat_request: ChatCompletionRequest) -> Any
         ret = pl.generator.run_vllm(chat_request, retri_res, np_type)
     else:
         raise ValueError("LLM inference_type not supported")
-    return ret, retri_res
+    return ret, contexts

EdgeCraftRAG/edgecraftrag/utils.py

@@ -29,3 +29,14 @@ def iter_elements(cls, paragraph: Paragraph, opts: DocxPartitionerOptions) -> It
                 image.save(image_path)
                 element_metadata = ElementMetadata(image_path=image_path)
             yield Image(text="IMAGE", metadata=element_metadata)
+
+
+def serialize_node_with_score(node_with_score):
+    return {
+        "node": node_with_score.node.__dict__,
+        "score": node_with_score.score.item() if hasattr(node_with_score.score, "item") else node_with_score.score,
+    }
+
+
+def serialize_contexts(contexts):
+    return {key: [serialize_node_with_score(node) for node in nodes] for key, nodes in contexts.items()}

EdgeCraftRAG/ui/vue/components.d.ts

@@ -32,17 +32,22 @@ declare module 'vue' {
     AModal: typeof import('ant-design-vue/es')['Modal']
     APagination: typeof import('ant-design-vue/es')['Pagination']
     APopover: typeof import('ant-design-vue/es')['Popover']
+    AProgress: typeof import('ant-design-vue/es')['Progress']
     ARadio: typeof import('ant-design-vue/es')['Radio']
     ARadioGroup: typeof import('ant-design-vue/es')['RadioGroup']
+    ARate: typeof import('ant-design-vue/es')['Rate']
     ARow: typeof import('ant-design-vue/es')['Row']
     ASelect: typeof import('ant-design-vue/es')['Select']
     ASelectOption: typeof import('ant-design-vue/es')['SelectOption']
     ASlider: typeof import('ant-design-vue/es')['Slider']
     ASpace: typeof import('ant-design-vue/es')['Space']
     ASteps: typeof import('ant-design-vue/es')['Steps']
+    ASwitch: typeof import('ant-design-vue/es')['Switch']
     ATable: typeof import('ant-design-vue/es')['Table']
     ATag: typeof import('ant-design-vue/es')['Tag']
+    ATextarea: typeof import('ant-design-vue/es')['Textarea']
     ATooltip: typeof import('ant-design-vue/es')['Tooltip']
+    AUpload: typeof import('ant-design-vue/es')['Upload']
     AUploadDragger: typeof import('ant-design-vue/es')['UploadDragger']
     FormTooltip: typeof import('./src/components/FormTooltip.vue')['default']
     RouterLink: typeof import('vue-router')['RouterLink']

EdgeCraftRAG/ui/vue/package.json

@@ -12,8 +12,10 @@
     "@vueuse/i18n": "^4.0.0-beta.12",
     "ant-design-vue": "^4.0.0-rc.6",
     "axios": "^1.7.9",
+    "clipboard": "^2.0.11",
     "echarts": "^5.5.1",
     "event-source-polyfill": "^1.0.31",
+    "highlight.js": "^11.11.1",
     "http": "^0.0.1-security",
     "js-cookie": "^3.0.5",
     "lodash": "^4.17.21",

EdgeCraftRAG/ui/vue/src/assets/iconFont/iconfont.css

@@ -1,9 +1,9 @@
 @font-face {
   font-family: "iconfont"; /* Project id 4784207 */
   src:
-    url("iconfont.woff2?t=1739238081968") format("woff2"),
-    url("iconfont.woff?t=1739238081968") format("woff"),
-    url("iconfont.ttf?t=1739238081968") format("truetype");
+    url("iconfont.woff2?t=1744699312006") format("woff2"),
+    url("iconfont.woff?t=1744699312006") format("woff"),
+    url("iconfont.ttf?t=1744699312006") format("truetype");
 }
 
 .iconfont {
@@ -14,6 +14,42 @@
   -moz-osx-font-smoothing: grayscale;
 }
 
+.icon-export:before {
+  content: "\e619";
+}
+
+.icon-rename:before {
+  content: "\e618";
+}
+
+.icon-delete:before {
+  content: "\e664";
+}
+
+.icon-setting1:before {
+  content: "\e61b";
+}
+
+.icon-upload:before {
+  content: "\e617";
+}
+
+.icon-clear:before {
+  content: "\e765";
+}
+
+.icon-copy-success:before {
+  content: "\e666";
+}
+
+.icon-copy:before {
+  content: "\e660";
+}
+
+.icon-subway:before {
+  content: "\e6ed";
+}
+
 .icon-stop:before {
   content: "\e904";
 }

EdgeCraftRAG/ui/vue/src/assets/iconFont/iconfont.json

@@ -5,6 +5,69 @@
   "css_prefix_text": "icon-",
   "description": "",
   "glyphs": [
+    {
+      "icon_id": "43924556",
+      "name": "export",
+      "font_class": "export",
+      "unicode": "e619",
+      "unicode_decimal": 58905
+    },
+    {
+      "icon_id": "43924554",
+      "name": "rename",
+      "font_class": "rename",
+      "unicode": "e618",
+      "unicode_decimal": 58904
+    },
+    {
+      "icon_id": "2570142",
+      "name": "delete",
+      "font_class": "delete",
+      "unicode": "e664",
+      "unicode_decimal": 58980
+    },
+    {
+      "icon_id": "13253937",
+      "name": "setting",
+      "font_class": "setting1",
+      "unicode": "e61b",
+      "unicode_decimal": 58907
+    },
+    {
+      "icon_id": "43796752",
+      "name": "upload",
+      "font_class": "upload",
+      "unicode": "e617",
+      "unicode_decimal": 58903
+    },
+    {
+      "icon_id": "42194548",
+      "name": "clear",
+      "font_class": "clear",
+      "unicode": "e765",
+      "unicode_decimal": 59237
+    },
+    {
+      "icon_id": "1198529",
+      "name": "copy-success",
+      "font_class": "copy-success",
+      "unicode": "e666",
+      "unicode_decimal": 58982
+    },
+    {
+      "icon_id": "9080698",
+      "name": "copy",
+      "font_class": "copy",
+      "unicode": "e660",
+      "unicode_decimal": 58976
+    },
+    {
+      "icon_id": "796912",
+      "name": "地铁",
+      "font_class": "subway",
+      "unicode": "e6ed",
+      "unicode_decimal": 59117
+    },
     {
       "icon_id": "42853460",
       "name": "stop",