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

@@ -11,11 +11,13 @@ load_dotenv(override=True)
 
 from vectara_agentic.agent import Agent
 from vectara_agentic.tools import ToolsFactory, VectaraToolFactory
+from vectara_agentic.types import ModelProvider, AgentType
+from vectara_agentic.agent_config import AgentConfig
+
 
 def create_assistant_tools(cfg):    
 
     class QueryCFPBComplaints(BaseModel):
-        query: str = Field(description="The user query.")
         company: Optional[str] = Field(
             default=None,
             description="The company that the complaint is about.",
@@ -32,7 +34,7 @@ def create_assistant_tools(cfg):
         vectara_corpus_key=cfg.corpus_keys
     )
     
-    summarizer = 'vectara-experimental-summary-ext-2023-12-11-med-omni'
+    summarizer = 'vectara-summary-table-md-query-ext-jan-2025-gpt-4o'
     ask_complaints = vec_factory.create_rag_tool(
         tool_name = "ask_complaints",
         tool_description = """
@@ -40,7 +42,7 @@ def create_assistant_tools(cfg):
         returns a response to a user question about customer complaints for bank services.
         """,
         tool_args_schema = QueryCFPBComplaints,
-        reranker = "chain", rerank_k = 100, 
+        reranker = "chain", rerank_k = 100,
         rerank_chain = [
             {
                 "type": "slingshot",
@@ -49,46 +51,62 @@ def create_assistant_tools(cfg):
             {
                 "type": "mmr",
                 "diversity_bias": 0.2,
-                "limit": 30
             }
         ],
-        n_sentences_before = 2, n_sentences_after = 3, lambda_val = 0.005,
+        n_sentences_before = 2, n_sentences_after = 2, lambda_val = 0.005,
+        summary_num_results = 10,
+        max_tokens = 4096, max_response_chars = 8192,
         vectara_summarizer = summarizer,
         include_citations = True,
-        verbose=False
+        verbose = True
     )
 
     tools_factory = ToolsFactory()
-
     db_tools = tools_factory.database_tools(
                 tool_name_prefix = "cfpb",
-                content_description = 'Customer complaints about five banks (Bank of America, Wells Fargo, Capital One, Chase, and CITI Bank) and geographic information (counties and zip codes)',
+                content_description = "Customer complaints about five banks (Bank of America, Wells Fargo, Capital One, Chase, and CITI Bank) and geographic information (counties and zip codes)",
                 sql_database = SQLDatabase(create_engine('sqlite:///cfpb_database.db')),
             )
 
     return (tools_factory.standard_tools() + 
-            tools_factory.guardrail_tools() +
             db_tools +
             [ask_complaints]
     )
 
 def initialize_agent(_cfg, agent_progress_callback=None):
     cfpb_complaints_bot_instructions = """
-    - You are a helpful research assistant, 
-      with expertise in finance and complaints from the CFPB (Consumer Financial Protection Bureau), 
-      in conversation with a user.
-    - For questions about customers' complaints (the text of the complaint), use the ask_complaints tool.
-      You only need the query parameter to use this tool, but you can supply other parameters if provided.
-      Do not include the "References" section in your response.
-    - When using a company name with a tool, if the tool returns no information, try a difference variation of that company names, as well as different capitalization.
+    - You are a helpful research assistant in conversation with a user.
+    - You are in expert in the domain of complaints recorded by the CFPB (Consumer Financial Protection Bureau).
+    - For informational questions about customer complaints, use the 'ask_complaints' tool.
+    - For analytical questions, use the database tools: cfpb_load_data, cfpb_load_sample_data, cfpb_list_tables, cfpb_describe_tables and cfpb_load_unique_values.
     - Never discuss politics, and always respond politely.
     """
+    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="Customer complaints from the Consumer Financial Protection Bureau (CFPB)",
         custom_instructions=cfpb_complaints_bot_instructions,
-        agent_progress_callback=agent_progress_callback
+        agent_progress_callback=agent_progress_callback,
+        validate_tools=True,
+        verbose=True,
+        agent_config=agent_config,
+        fallback_agent_config=fallback_agent_config,
     )
     agent.report(detailed=False)
     return agent

requirements.txt

@@ -1,10 +1,10 @@
 omegaconf==2.3.0
 python-dotenv==1.0.1
-streamlit==1.43.2
+streamlit==1.45.0
 streamlit-feedback==0.1.3
 langdetect==1.0.9
 langcodes==3.4.0
 datasets==2.19.2
 uuid==1.30
-vectara-agentic==0.2.9
-torch==2.6.0
+vectara-agentic==0.2.15
+torch==2.6.0

st_app.py

@@ -131,7 +131,7 @@ async def launch_bot():
     if st.session_state.prompt:
         with st.chat_message("assistant", avatar='🤖'):
             st.session_state.status = st.status('Processing...', expanded=False)
-            response = st.session_state.agent.chat(st.session_state.prompt)
+            response = await st.session_state.agent.achat(st.session_state.prompt)
             res = escape_dollars_outside_latex(response.response)
             message = {"role": "assistant", "content": res, "avatar": '🤖'}
             st.session_state.messages.append(message)