updated

Dockerfile

@@ -7,12 +7,15 @@ COPY ./requirements.txt /app/requirements.txt
 RUN pip3 install --no-cache-dir --upgrade pip
 RUN pip3 install --no-cache-dir wheel setuptools build
 RUN pip3 install --no-cache-dir --use-pep517 -r /app/requirements.txt
-
+ 
 # User
 RUN useradd -m -u 1000 user
 USER user
 ENV HOME /home/user
 ENV PATH $HOME/.local/bin:$PATH
+ENV TIKTOKEN_CACHE_DIR $HOME/.cache/tiktoken
+
+RUN mkdir -p $HOME/.cache/tiktoken
 
 WORKDIR $HOME
 RUN mkdir app

agent.py

@@ -1,43 +1,13 @@
+import os
 from pydantic import Field, BaseModel
 
 from vectara_agentic.agent import Agent
 from vectara_agentic.tools import VectaraToolFactory
+from vectara_agentic.types import ModelProvider, AgentType
+from vectara_agentic.agent_config import AgentConfig
 
 initial_prompt = "How can I help you today?"
 
-prompt_old = """
-[
-  {"role": "system", "content": "You are an AI assistant that forms a detailed and comprehensive answer to a user query based on search results that are provided to you." },
-  {"role": "user", "content": " 
-    [INSTRUCTIONS]
-    If the search results are irrelevant to the question respond with *** I do not have enough information to answer this question.***
-    Search results may include tables in a markdown format. 
-    When answering a question using a table be careful about which rows and columns contain the answer and include all relevant information from the relevant rows and columns that the query is asking about.
-    Do not base your response on information or knowledge that is not in the search results.
-    Make sure your response is answering the query asked. If the query is related to an entity (such as a person or place), make sure you use search results related to that entity.
-    Consider that each search result is a partial segment from a bigger text, and may be incomplete.
-    Your output should always be in a single language - the $vectaraLangName language. Check spelling and grammar for the $vectaraLangName language.
-    Search results for the query *** $vectaraQuery***, are listed below, some are text, some MAY be tables in markdown format.
-    #foreach ($qResult in $vectaraQueryResultsDeduped)
-      [$esc.java($foreach.index + 1)]
-      #if($qResult.hasTable())
-        Table Title: $qResult.getTable().title() || Table Description: $qResult.getTable().description() || Table Data:
-        $qResult.getTable().markdown()
-      #else
-        $qResult.getText()
-      #end
-    #end
-    Generate a comprehensive response to the query *** $vectaraQuery *** using information and facts in the search results provided. 
-    Give a slight preference to search results that appear earlier in the list.
-    Include statistical and numerical evidence to support and contextualize your response.
-    Your response should include all relevant information and values from the search results. Do not omit anything relevant.
-    Prioritize a long, detailed, thorough and comprehensive response over a short one.
-    Cite relevant search results in your answer following these specific instructions: $vectaraCitationInstructions
-    Respond always in the $vectaraLangName language, and only in that language."}
-]
-"""
-
-
 prompt = """
 [
   {"role": "system", "content": "
@@ -120,16 +90,12 @@ prompt_new = """
 def create_assistant_tools(cfg):
 
     class QueryPublicationsArgs(BaseModel):
-        query: str = Field(..., description="The user query, always in the form of a question?", 
-                                examples=[
-                                    "what are the risks reported?", 
-                                    "which drug was tested?", 
-                                    "what is the baseline population in the trial?"
-                                ]),
         name: str = Field(..., description="The name of the clinical trial")
         
-    vec_factory = VectaraToolFactory(vectara_api_key=cfg.api_key,
-                                     vectara_corpus_key=cfg.corpus_key)
+    vec_factory = VectaraToolFactory(
+        vectara_api_key=cfg.api_key,
+        vectara_corpus_key=cfg.corpus_key
+      )
     summarizer = 'vectara-summary-table-md-query-ext-jan-2025-gpt-4o'
     ask_publications = vec_factory.create_rag_tool(
         tool_name = "ask_publications",
@@ -138,9 +104,9 @@ def create_assistant_tools(cfg):
         """,
         tool_args_schema = QueryPublicationsArgs,
         reranker = "slingshot", rerank_k = 100, rerank_cutoff = 0.1,
-        n_sentences_before = 1, n_sentences_after = 1, lambda_val = 0.1,
+        n_sentences_before = 2, n_sentences_after = 2, lambda_val = 0.1,
         summary_num_results = 15,
-        max_tokens = 8192,
+        max_tokens = 4096, max_response_chars = 8192,
         vectara_summarizer = summarizer,
         include_citations = True,
         vectara_prompt_text = prompt,
@@ -148,27 +114,18 @@ def create_assistant_tools(cfg):
         verbose = False
     )
 
-    class SearchPublicationsArgs(BaseModel):
-        query: str = Field(..., description="The user query, always in the form of a question?", 
-                                examples=[
-                                    "what are the risks reported?", 
-                                    "which drug was tested?", 
-                                    "what is the baseline population in the trial?"
-                                ]),
     search_publications = vec_factory.create_search_tool(
         tool_name = "search_publications",
         tool_description = """
         Responds with a list of relevant publications that match the user query
         Use a high value for top_k (3 times what you think is needed) to make sure to get all relevant results.
         """,
-        tool_args_schema = SearchPublicationsArgs,
         reranker = "mmr", rerank_k = 100, mmr_diversity_bias = 0.5,
         n_sentences_before = 2, n_sentences_after = 2, lambda_val = 0.3,
         save_history = True,
         verbose = False
     )
 
-
     return (
         [ask_publications, search_publications]
     )
@@ -200,11 +157,31 @@ def initialize_agent(_cfg, agent_progress_callback=None):
     7) Be consistent and comprehensive in your responses, ensuring that all relevant information is included.    
     """
 
+    agent_config = AgentConfig(
+        agent_type = os.getenv("VECTARA_AGENTIC_AGENT_TYPE", AgentType.OPENAI.value),
+        main_llm_provider = os.getenv("VECTARA_AGENTIC_MAIN_LLM_PROVIDER", ModelProvider.OPENAI.value),
+        main_llm_model_name = os.getenv("VECTARA_AGENTIC_MAIN_MODEL_NAME", ""),
+        tool_llm_provider = os.getenv("VECTARA_AGENTIC_TOOL_LLM_PROVIDER", ModelProvider.OPENAI.value),
+        tool_llm_model_name = os.getenv("VECTARA_AGENTIC_TOOL_MODEL_NAME", ""),
+        observer = os.getenv("VECTARA_AGENTIC_OBSERVER_TYPE", "NO_OBSERVER")
+    )
+    fallback_agent_config = AgentConfig(
+        agent_type = os.getenv("VECTARA_AGENTIC_FALLBACK_AGENT_TYPE", AgentType.OPENAI.value),
+        main_llm_provider = os.getenv("VECTARA_AGENTIC_FALLBACK_MAIN_LLM_PROVIDER", ModelProvider.OPENAI.value),
+        main_llm_model_name = os.getenv("VECTARA_AGENTIC_FALLBACK_MAIN_MODEL_NAME", ""),
+        tool_llm_provider = os.getenv("VECTARA_AGENTIC_FALLBACK_TOOL_LLM_PROVIDER", ModelProvider.OPENAI.value),
+        tool_llm_model_name = os.getenv("VECTARA_AGENTIC_FALLBACK_TOOL_MODEL_NAME", ""),
+        observer = os.getenv("VECTARA_AGENTIC_OBSERVER_TYPE", "NO_OBSERVER")
+    )
+
+
     agent = Agent(
         tools=create_assistant_tools(_cfg),
         topic="Drug trials publications",
         custom_instructions=menarini_bot_instructions,
         agent_progress_callback=agent_progress_callback,
+        agent_config=agent_config,
+        fallback_agent_config=fallback_agent_config,
     )
     agent.report()
     return agent

requirements.txt

@@ -1,9 +1,9 @@
 omegaconf==2.3.0
 python-dotenv==1.0.1
-streamlit==1.43.2
+streamlit==1.45.0
 streamlit_feedback==0.1.3
 uuid==1.30
 langdetect==1.0.9
 langcodes==3.4.0
-vectara-agentic==0.2.11
+vectara-agentic==0.2.15
 torch==2.6.0

st_app.py

@@ -89,7 +89,7 @@ async def launch_bot():
     if st.session_state.prompt:
         with st.chat_message("assistant", avatar='🤖'):
             st.session_state.status = st.status('Processing...', expanded=False)
-            res = st.session_state.agent.chat(st.session_state.prompt)
+            res = await st.session_state.agent.achat(st.session_state.prompt)
             #res = escape_dollars_outside_latex(res)
             res = str(res)
             message = {"role": "assistant", "content": res, "avatar": '🤖'}