Add eval tab (#214)

src/pai_rag/app/web/rag_client.py

@@ -468,7 +468,7 @@ def evaluate_for_response_stage(self):
         response = dotdict(json.loads(r.text))
         if r.status_code != HTTPStatus.OK:
             raise RagApiError(code=r.status_code, msg=response.message)
-        print("evaluate_for_response_stage response", response)
+        return response
 
 
 rag_client = RagWebClient()

src/pai_rag/app/web/tabs/eval_tab.py

@@ -44,7 +44,6 @@ def eval_response_stage():
         response_res = rag_client.evaluate_for_response_stage()
     except RagApiError as api_error:
         raise gr.Error(f"HTTP {api_error.code} Error: {api_error.msg}")
-
     formatted_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
     pd_results = {
         "Metrics": ["Faithfulness", "Correctness", "Similarity", "LastModified"],

src/pai_rag/app/web/webui.py

@@ -8,6 +8,7 @@
 from pai_rag.app.web.tabs.chat_tab import create_chat_tab
 from pai_rag.app.web.tabs.agent_tab import create_agent_tab
 from pai_rag.app.web.tabs.data_analysis_tab import create_data_analysis_tab
+from pai_rag.app.web.tabs.eval_tab import create_evaluation_tab
 from pai_rag.app.web.element_manager import elem_manager
 from pai_rag.app.web.ui_constants import (
     DEFAULT_CSS_STYPE,
@@ -56,15 +57,15 @@ def make_homepage():
         with gr.Tab("\N{fire} Chat"):
             chat_elements = create_chat_tab()
             elem_manager.add_elems(chat_elements)
-        # with gr.Tab("\N{rocket} Evaluation"):
-        #     eval_elements = create_evaluation_tab()
-        # elem_manager.add_elems(eval_elements)
         with gr.Tab("\N{rocket} Agent"):
             agent_elements = create_agent_tab()
             elem_manager.add_elems(agent_elements)
         with gr.Tab("\N{bar chart} Data Analysis"):
             analysis_elements = create_data_analysis_tab()
             elem_manager.add_elems(analysis_elements)
+        with gr.Tab("\N{rocket} Evaluation"):
+            eval_elements = create_evaluation_tab()
+            elem_manager.add_elems(eval_elements)
         homepage.load(
             resume_ui, outputs=elem_manager.get_elem_list(), concurrency_limit=None
         )

src/pai_rag/config/settings.toml

@@ -69,6 +69,10 @@ name = "qwen-turbo"
 [rag.llm.function_calling_llm]
 source = ""
 
+[rag.llm.multi_modal]
+enable = false
+source = ""
+
 [rag.llm_chat_engine]
 type = "SimpleChatEngine"
 

src/pai_rag/modules/evaluation/batch_eval_runner.py

@@ -7,7 +7,6 @@
 from llama_index.core.base.response.schema import RESPONSE_TYPE, Response
 from llama_index.core.evaluation.base import BaseEvaluator, EvaluationResult
 from pai_rag.integrations.evaluation.retrieval.evaluator import MyRetrievalEvalResult
-from fastapi.concurrency import run_in_threadpool
 
 
 async def eval_response_worker(
@@ -277,14 +276,12 @@ async def aevaluate_queries(
             response_jobs.append(response_worker(self.semaphore, query_engine, query))
         responses = await self.asyncio_mod.gather(*response_jobs)
 
-        return await run_in_threadpool(
-            lambda: self.aevaluate_responses(
-                queries=queries,
-                node_ids=node_ids,
-                responses=responses,
-                references=reference_answers,
-                **eval_kwargs_lists,
-            )
+        return await self.aevaluate_responses(
+            queries=queries,
+            node_ids=node_ids,
+            responses=responses,
+            references=reference_answers,
+            **eval_kwargs_lists,
         )
 
     async def aevaluate_queries_for_retrieval(