updated

agent.py

@@ -1,6 +1,7 @@
 import os
 import pandas as pd
 import requests
+from functools import lru_cache
 from pydantic import Field, BaseModel
 
 from omegaconf import OmegaConf
@@ -11,7 +12,6 @@ from vectara_agentic.tools import ToolsFactory, VectaraToolFactory
 from dotenv import load_dotenv
 load_dotenv(override=True)
 
-
 tickers = {
     "C": "Citigroup",
     "COF": "Capital One",
@@ -51,9 +51,10 @@ def create_assistant_tools(cfg):
         return years
 
     # Tool to get the income statement for a given company and year using the FMP API
+    @lru_cache(maxsize=128)
     def fmp_income_statement(
-        ticker: str = Field(description="the ticker symbol of the company."),
-        year: int = Field(description="the year for which to get the income statement."),
+        ticker: str = Field(description="the ticker symbol of the company.", examples=["AAPL", "GOOG", "AMZN"]),
+        year: int = Field(description="the year for which to get the income statement.", examples=[2020, 2021, 2022]),
     ) -> str:
         """
         Get the income statement for a given company and year using the FMP (https://financialmodelingprep.com) API.
@@ -80,7 +81,11 @@ def create_assistant_tools(cfg):
 
     class QueryTranscriptsArgs(BaseModel):
         query: str = Field(..., description="The user query, always in the form of a question", examples=["what are the risks reported?", "who are the competitors?"])
-        year: int | str = Field(..., description=f"The year this query relates to. An integer between {min(years)} and {max(years)} or a string specifying a condition on the year (example: '>2020').")
+        year: int | str = Field(
+            default=None,
+            description=f"The year this query relates to. An integer between {min(years)} and {max(years)} or a string specifying a condition on the year",
+            examples=[2020, '>2021', '<2023', '>=2021', '<=2023', '[2021, 2023]', '[2021, 2023)']
+        )
         ticker: str = Field(..., description=f"The company ticker this query relates to. Must be a valid ticket symbol from the list {list(tickers.keys())}.")
 
     vec_factory = VectaraToolFactory(vectara_api_key=cfg.api_key,
@@ -99,6 +104,27 @@ def create_assistant_tools(cfg):
         summary_num_results = 10,
         vectara_summarizer = summarizer,
         include_citations = True,
+        verbose=True,
+    )
+
+    class SearchTranscriptsArgs(BaseModel):
+        query: str = Field(..., description="The user query, always in the form of a question", examples=["what are the risks reported?", "who are the competitors?"])
+        top_k: int = Field(..., description="The number of results to return.")
+        year: int | str = Field(
+            default=None,
+            description=f"The year this query relates to. An integer between {min(years)} and {max(years)} or a string specifying a condition on the year",
+            examples=[2020, '>2021', '<2023', '>=2021', '<=2023', '[2021, 2023]', '[2021, 2023)']
+        )
+        ticker: str = Field(..., description=f"The company ticker this query relates to. Must be a valid ticket symbol from the list {list(tickers.keys())}.")
+    search_transcripts = vec_factory.create_search_tool(
+        tool_name = "search_transcripts",
+        tool_description = """
+        Given a company name and year, and a user query, retrieves the most relevant text from analyst call transcripts about the company related to the user query.
+        """,
+        tool_args_schema = QueryTranscriptsArgs,
+        reranker = "multilingual_reranker_v1", rerank_k = 100,
+        lambda_val = 0.005,
+        verbose=True
     )
 
     tools_factory = ToolsFactory()
@@ -111,7 +137,7 @@ def create_assistant_tools(cfg):
                 ]
             ] +
             tools_factory.financial_tools() +
-            [ask_transcripts]
+            [ask_transcripts, search_transcripts]
     )
 
 def initialize_agent(_cfg, agent_progress_callback=None):
@@ -129,12 +155,15 @@ def initialize_agent(_cfg, agent_progress_callback=None):
     - When querying a tool for a numeric value or KPI, use a concise and non-ambiguous description of what you are looking for. 
     - If you calculate a metric, make sure you have all the necessary information to complete the calculation. Don't guess.
     """
+    def query_logging(query: str, response: str):
+        print(f"Logging query={query}, response={response}")
 
     agent = Agent(
         tools=create_assistant_tools(_cfg),
         topic="Financial data, annual reports and 10-K filings",
         custom_instructions=financial_bot_instructions,
         agent_progress_callback=agent_progress_callback,
+        query_logging_callback=query_logging,
     )
     agent.report()
     return agent

app.py

@@ -12,7 +12,10 @@ if 'device_id' not in st.session_state:
 if "feedback_key" not in st.session_state:
     st.session_state.feedback_key = 0
 
+async def main():
+    await launch_bot()
+
 if __name__ == "__main__":
    st.set_page_config(page_title="Financial Assistant", layout="wide")
    nest_asyncio.apply()
-   asyncio.run(launch_bot())
+   asyncio.run(main())

requirements.txt

@@ -6,4 +6,4 @@ streamlit_feedback==0.1.3
 uuid==1.30
 langdetect==1.0.9
 langcodes==3.4.0
-vectara-agentic==0.1.22
+vectara-agentic==0.1.27

st_app.py

@@ -13,15 +13,6 @@ from agent import initialize_agent, get_agent_config
 
 initial_prompt = "How can I help you today?"
 
-def show_example_questions():        
-    if len(st.session_state.example_messages) > 0 and st.session_state.first_turn:            
-        selected_example = pills("Queries to Try:", st.session_state.example_messages, index=None)
-        if selected_example:
-            st.session_state.ex_prompt = selected_example
-            st.session_state.first_turn = False
-            return True
-    return False
-
 def format_log_msg(log_msg: str):
     max_log_msg_size = 500
     return log_msg if len(log_msg) <= max_log_msg_size else log_msg[:max_log_msg_size]+'...'
@@ -37,7 +28,13 @@ def agent_progress_callback(status_type: AgentStatusType, msg: str):
             latest_message = f"Calling tool {tool_name}..."
         elif status_type == AgentStatusType.TOOL_OUTPUT:
             latest_message = "Analyzing tool output..."
+        elif status_type == AgentStatusType.AGENT_UPDATE:
+            if "Thought:" in msg:
+                latest_message = "Thinking..."
+            else:
+                latest_message = "Updating agent..."
         else:
+            print(f"callback with {status_type} and {msg}")
             return
         
         st.session_state.status.update(label=latest_message)
@@ -46,6 +43,16 @@ def agent_progress_callback(status_type: AgentStatusType, msg: str):
             for log_msg in st.session_state.log_messages:
                 st.markdown(format_log_msg(log_msg), unsafe_allow_html=True)
 
+
+def show_example_questions():        
+    if len(st.session_state.example_messages) > 0 and st.session_state.first_turn:            
+        selected_example = pills("Queries to Try:", st.session_state.example_messages, index=None)
+        if selected_example:
+            st.session_state.ex_prompt = selected_example
+            st.session_state.first_turn = False
+            return True
+    return False
+
 @st.dialog(title="Agent logs", width='large')
 def show_modal():
     for log_msg in st.session_state.log_messages:
@@ -132,8 +139,16 @@ 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 = escape_dollars_outside_latex(res)
+            response = st.session_state.agent.chat(st.session_state.prompt)
+            res = escape_dollars_outside_latex(response.response)
+            
+            #response = await st.session_state.agent.achat(st.session_state.prompt)
+            #res = escape_dollars_outside_latex(response.response)
+            
+            #res = await st.session_state.agent.astream_chat(st.session_state.prompt)
+            #response = ''.join([token async for token in res.async_response_gen()])
+            #res = escape_dollars_outside_latex(response)
+            
             message = {"role": "assistant", "content": res, "avatar": '🤖'}
             st.session_state.messages.append(message)
             st.markdown(res)