From eec04383aa9e672481d97bc64a687244f0c08870 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Thu, 16 May 2024 16:32:44 -0400
Subject: [PATCH 01/13] replace 3.5 with 4o and use gemini flash 1.5

---
 04-call-summaries/llm.py |  10 ++-
 04-call-summaries/run.py | 179 ++++++++++++++++++++++-----------------
 2 files changed, 109 insertions(+), 80 deletions(-)

diff --git a/04-call-summaries/llm.py b/04-call-summaries/llm.py
index 1975b1c..44cc6f1 100644
--- a/04-call-summaries/llm.py
+++ b/04-call-summaries/llm.py
@@ -36,7 +36,7 @@ def ollama_client(
 
 
 def google_gemini_client(
-    model_name="gemini-pro",
+    model_name="gemini-1.5-flash-latest",
     prompt=None,
     settings=None,
 ):
@@ -55,8 +55,8 @@ def google_gemini_client(
 
 def gpt3_5(prompt, model="gpt-3.5-turbo"):
     # Get API key from https://platform.openai.com/api-keys
-    OPEN_AI_API_KEY = os.environ.get("OPEN_AI_API_KEY")
-    openai_client = OpenAI(api_key=OPEN_AI_API_KEY)  # Uses OPENAI_API_KEY
+    OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
+    openai_client = OpenAI(api_key=OPENAI_API_KEY)  # Uses OPENAI_API_KEY
     return (
         openai_client.chat.completions.create(
             model=model, messages=[{"role": "user", "content": prompt}]
@@ -70,6 +70,10 @@ def gpt_4_turbo(prompt):
     return gpt3_5(prompt, model="gpt-4-turbo")
 
 
+def gpt_4o(prompt):
+    return gpt3_5(prompt, model="gpt-4o")
+
+
 def claude(prompt, model="claude-3-opus-20240229", max_tokens=1024):
     # Get API key from https://console.anthropic.com/settings/keys
     ANTHROPIC_API_KEY = os.environ.get("ANTHROPIC_API_KEY")
diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index 43c8eea..9fc07f8 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -1,10 +1,11 @@
 from langchain_core.prompts import PromptTemplate
 
-from llm import claude, google_gemini_client, gpt3_5, gpt_4_turbo, ollama_client
+from llm import claude, google_gemini_client, gpt3_5, gpt_4_turbo, gpt_4o, ollama_client
+from chunking import chunking_ingest
 
 
 # download transcripts from https://drive.google.com/drive/folders/19r6x3Zep4N9Rl_x4n4H6RpWkXviwbxyw?usp=sharing
-def get_transcript(file_path="./transcript.txt"):
+def get_transcript(file_path="./multi_benefit_transcript.txt"):
     file = open(file_path, encoding="utf-8")
     content = file.read()
     file.close()
@@ -32,88 +33,112 @@ def get_transcript(file_path="./transcript.txt"):
 8. Next Steps for Agent
 """
 
+CHUNKING_PROMPT = """
+You are a helpful AI assistant tasked with summarizing transcripts, however we can only process the transcripts in pieces.
+Fill out the fields with the text given {text_given}. If the template already has the field filled out, do not overwrite this information. 
+If the template is blank, fill out the following template the best you can using the text:
+1. Caller Information:
+- Name
+- Contact Information
+- Availability
+- Household Information
+2. Reason/Type of Call: e.g., Applying for benefits, Follow-ups
+3. Previous Benefits History:
+- Applied for
+- Receives
+- Denied
+4. Benefits Discussion: Prefix the discussed benefit with a hashtag (e.g., #SNAP, #LIHEAP)
+5. Discussion Points:
+- Key information points
+6. Documents Needed: e.g., Income verification, Housing documentation
+7. Next Steps for Client
+8. Next Steps for Agent
+"""
+
 print("""
       Select an llm
       1. openhermes (default)
       2. dolphin
       3. gemini
-      4. gpt 3.5
-      5. gpt 4
+      4. gpt 4
+      5. gpt 4o
       6. claude 3
       7. all
       """)
 
-
 transcript = get_transcript()
-llm = input() or "1"
-prompt_template = PromptTemplate.from_template(PROMPT)
-formatted_prompt = prompt_template.format(transcript=transcript)
-
-if llm == "2":
-    test = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
-    print("""----------
-        Dolphin
-        """)
-elif llm == "3":
-    test = google_gemini_client(prompt=formatted_prompt).text
-    print("""----------
-        Gemini
-        """)
-elif llm == "4":
-    print("""----------
-      GPT 3.5
-      """)
-    test = gpt3_5(prompt=formatted_prompt)
-elif llm == "5":
-    print("""----------
-        GPT 4
-        """)
-    test = gpt_4_turbo(prompt=formatted_prompt)
-elif llm == "6":
-    print("""----------
-        Claude 3
-        """)
-    test = claude(prompt=formatted_prompt)
-elif llm == "7":
-    test_open_hermes = ollama_client(model_name="openhermes", prompt=formatted_prompt)
-    print("""
-        Openhermes
-        """)
-    print(test_open_hermes)
-
-    test_dolphin = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
-    print("""----------
-        Dolphin
-        """)
-    print(test_dolphin)
-
-    test_gemini = google_gemini_client(prompt=formatted_prompt).text
-    print("""----------
-        Gemini
-        """)
-    print(test_gemini)
-
-    print("""----------
-      GPT 3.5
-      """)
-    test_gpt3_5 = gpt3_5(prompt=formatted_prompt)
-    print(test_gpt3_5)
-
-    print("""----------
-        GPT 4
-        """)
-    test_gpt4 = gpt_4_turbo(prompt=formatted_prompt)
-    print(test_gpt4)
-
-    print("""----------
-        Claude 3
-        """)
-    test_claude = claude(prompt=formatted_prompt)
-    print(test_claude)
-else:
-    test = ollama_client(model_name="openhermes", prompt=formatted_prompt)
-    print("""
-        Openhermes
-        """)
-if test:
-    print(test)
+# llm = input() or "1"
+# prompt_template = PromptTemplate.from_template(PROMPT)
+# formatted_prompt = prompt_template.format(transcript=transcript)
+
+# if llm == "2":
+#     test = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
+#     print("""----------
+#         Dolphin
+#         """)
+# elif llm == "3":
+#     test = google_gemini_client(prompt=formatted_prompt).text
+#     print("""----------
+#         Gemini Flash 1.5
+#         """)
+# elif llm == "4":
+#     print("""----------
+#       GPT 4
+#       """)
+#     test = gpt_4_turbo(prompt=formatted_prompt)
+# elif llm == "5":
+#     print("""----------
+#       GPT 4o
+#       """)
+#     test = gpt_4o(prompt=formatted_prompt)
+# elif llm == "6":
+#     print("""----------
+#         Claude 3
+#         """)
+#     test = claude(prompt=formatted_prompt)
+# elif llm == "7":
+#     test_open_hermes = ollama_client(model_name="openhermes", prompt=formatted_prompt)
+#     print("""
+#         Openhermes
+#         """)
+#     print(test_open_hermes)
+
+#     test_dolphin = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
+#     print("""----------
+#         Dolphin
+#         """)
+#     print(test_dolphin)
+
+#     test_gemini = google_gemini_client(prompt=formatted_prompt).text
+#     print("""----------
+#         Gemini
+#         """)
+#     print(test_gemini)
+
+#     print("""----------
+#         GPT 4
+#         """)
+#     test_gpt4 = gpt_4_turbo(prompt=formatted_prompt)
+#     print(test_gpt4)
+
+#     print("""----------
+#         GPT 4o
+#         """)
+#     test_gpt4o = gpt_4o(prompt=formatted_prompt)
+#     print(test_gpt4o)
+
+#     print("""----------
+#         Claude 3
+#         """)
+#     test_claude = claude(prompt=formatted_prompt)
+#     print(test_claude)
+# else:
+#     test = ollama_client(model_name="openhermes", prompt=formatted_prompt)
+#     print("""
+#         Openhermes
+#         """)
+# if test:
+#     print(test)
+
+
+chunking_ingest(transcript=transcript, prompt=CHUNKING_PROMPT)

From fa8961b7499fe7c821731b3bd33b42ff35e676e0 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Thu, 16 May 2024 16:33:53 -0400
Subject: [PATCH 02/13] format

---
 04-call-summaries/run.py | 147 +++++++++++++++++++--------------------
 1 file changed, 72 insertions(+), 75 deletions(-)

diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index 9fc07f8..f2a5e68 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -67,78 +67,75 @@ def get_transcript(file_path="./multi_benefit_transcript.txt"):
       """)
 
 transcript = get_transcript()
-# llm = input() or "1"
-# prompt_template = PromptTemplate.from_template(PROMPT)
-# formatted_prompt = prompt_template.format(transcript=transcript)
-
-# if llm == "2":
-#     test = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
-#     print("""----------
-#         Dolphin
-#         """)
-# elif llm == "3":
-#     test = google_gemini_client(prompt=formatted_prompt).text
-#     print("""----------
-#         Gemini Flash 1.5
-#         """)
-# elif llm == "4":
-#     print("""----------
-#       GPT 4
-#       """)
-#     test = gpt_4_turbo(prompt=formatted_prompt)
-# elif llm == "5":
-#     print("""----------
-#       GPT 4o
-#       """)
-#     test = gpt_4o(prompt=formatted_prompt)
-# elif llm == "6":
-#     print("""----------
-#         Claude 3
-#         """)
-#     test = claude(prompt=formatted_prompt)
-# elif llm == "7":
-#     test_open_hermes = ollama_client(model_name="openhermes", prompt=formatted_prompt)
-#     print("""
-#         Openhermes
-#         """)
-#     print(test_open_hermes)
-
-#     test_dolphin = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
-#     print("""----------
-#         Dolphin
-#         """)
-#     print(test_dolphin)
-
-#     test_gemini = google_gemini_client(prompt=formatted_prompt).text
-#     print("""----------
-#         Gemini
-#         """)
-#     print(test_gemini)
-
-#     print("""----------
-#         GPT 4
-#         """)
-#     test_gpt4 = gpt_4_turbo(prompt=formatted_prompt)
-#     print(test_gpt4)
-
-#     print("""----------
-#         GPT 4o
-#         """)
-#     test_gpt4o = gpt_4o(prompt=formatted_prompt)
-#     print(test_gpt4o)
-
-#     print("""----------
-#         Claude 3
-#         """)
-#     test_claude = claude(prompt=formatted_prompt)
-#     print(test_claude)
-# else:
-#     test = ollama_client(model_name="openhermes", prompt=formatted_prompt)
-#     print("""
-#         Openhermes
-#         """)
-# if test:
-#     print(test)
-
-
-chunking_ingest(transcript=transcript, prompt=CHUNKING_PROMPT)
+llm = input() or "1"
+prompt_template = PromptTemplate.from_template(PROMPT)
+formatted_prompt = prompt_template.format(transcript=transcript)
+
+if llm == "2":
+    test = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
+    print("""----------
+        Dolphin
+        """)
+elif llm == "3":
+    test = google_gemini_client(prompt=formatted_prompt).text
+    print("""----------
+        Gemini Flash 1.5
+        """)
+elif llm == "4":
+    print("""----------
+      GPT 4
+      """)
+    test = gpt_4_turbo(prompt=formatted_prompt)
+elif llm == "5":
+    print("""----------
+      GPT 4o
+      """)
+    test = gpt_4o(prompt=formatted_prompt)
+elif llm == "6":
+    print("""----------
+        Claude 3
+        """)
+    test = claude(prompt=formatted_prompt)
+elif llm == "7":
+    test_open_hermes = ollama_client(model_name="openhermes", prompt=formatted_prompt)
+    print("""
+        Openhermes
+        """)
+    print(test_open_hermes)
+
+    test_dolphin = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
+    print("""----------
+        Dolphin
+        """)
+    print(test_dolphin)
+
+    test_gemini = google_gemini_client(prompt=formatted_prompt).text
+    print("""----------
+        Gemini
+        """)
+    print(test_gemini)
+
+    print("""----------
+        GPT 4
+        """)
+    test_gpt4 = gpt_4_turbo(prompt=formatted_prompt)
+    print(test_gpt4)
+
+    print("""----------
+        GPT 4o
+        """)
+    test_gpt4o = gpt_4o(prompt=formatted_prompt)
+    print(test_gpt4o)
+
+    print("""----------
+        Claude 3
+        """)
+    test_claude = claude(prompt=formatted_prompt)
+    print(test_claude)
+else:
+    test = ollama_client(model_name="openhermes", prompt=formatted_prompt)
+    print("""
+        Openhermes
+        """)
+if test:
+    print(test)

From 61b4fd08da5278de244c301122a92e20dde19d71 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Thu, 16 May 2024 16:36:20 -0400
Subject: [PATCH 03/13] set up chunking function

---
 04-call-summaries/chunking.py      | 86 ++++++++++++++++++++++++++++++
 04-call-summaries/requirements.txt |  7 ++-
 04-call-summaries/run.py           |  2 +
 3 files changed, 94 insertions(+), 1 deletion(-)
 create mode 100644 04-call-summaries/chunking.py

diff --git a/04-call-summaries/chunking.py b/04-call-summaries/chunking.py
new file mode 100644
index 0000000..2c619ad
--- /dev/null
+++ b/04-call-summaries/chunking.py
@@ -0,0 +1,86 @@
+from langchain_text_splitters import (
+    RecursiveCharacterTextSplitter,
+    NLTKTextSplitter,
+    SpacyTextSplitter,
+)
+from llm import ollama_client
+
+# from langchain_openai import OpenAIEmbeddings
+from langchain_community.embeddings.sentence_transformer import (
+    SentenceTransformerEmbeddings,
+)
+from langchain.chains.conversational_retrieval.base import ConversationalRetrievalChain
+from langchain.chains.llm import LLMChain
+from langchain_core.prompts import PromptTemplate
+
+from langchain_community.vectorstores.faiss import FAISS
+from langchain.docstore.document import Document
+
+
+# split text into chunks
+def get_text_chunks(text, chunk_size, chunk_overlap, text_splitter_choice):
+    if text_splitter_choice == "2":
+        text_splitter = NLTKTextSplitter()
+    elif text_splitter_choice == "3":
+        text_splitter = SpacyTextSplitter()
+    else:
+        text_splitter = RecursiveCharacterTextSplitter(
+            chunk_size=chunk_size, chunk_overlap=chunk_overlap
+        )
+
+    texts = text_splitter.split_text(text)
+
+    docs = [
+        Document(
+            page_content=t,
+        )
+        for t in texts
+    ]
+
+    return docs
+
+
+CHUNKING_PROMPT = """
+You are a helpful AI assistant tasked with summarizing transcripts, however we can only process the transcripts in pieces.
+Fill out the fields with the text given {text}. If the template {template} already has the field filled out, do not overwrite this information.
+Please fill out the data with the following template:
+1. Caller Information:
+- Name
+- Contact Information
+- Availability
+- Household Information
+2. Reason/Type of Call: e.g., Applying for benefits, Follow-ups
+3. Previous Benefits History:
+- Applied for
+- Receives
+- Denied
+4. Benefits Discussion: Prefix the discussed benefit with a hashtag (e.g., #SNAP, #LIHEAP)
+5. Discussion Points:
+- Key information points
+6. Documents Needed: e.g., Income verification, Housing documentation
+7. Next Steps for Client
+8. Next Steps for Agent
+"""
+
+
+def chunking_ingest(transcript, prompt):
+    # text_splitter_choice= input() or "2"
+    text_chunks = get_text_chunks(
+        transcript, chunk_size=750, chunk_overlap=300, text_splitter_choice="2"
+    )
+
+    finalized_template = ""
+    prompt_template = PromptTemplate.from_template(prompt)
+
+    print("TEXT CHUNK")
+    for text in text_chunks:
+        formatted_prompt = prompt_template.format(
+            text=text.page_content, template=finalized_template
+        )
+        print("formatted_prompt", formatted_prompt)
+        answer = ollama_client(model_name="openhermes", prompt=formatted_prompt)
+        print("answer______________________")
+        print(answer)
+        finalized_template += "\n" + answer
+    print(finalized_template)
+
diff --git a/04-call-summaries/requirements.txt b/04-call-summaries/requirements.txt
index 5b8da85..c986355 100644
--- a/04-call-summaries/requirements.txt
+++ b/04-call-summaries/requirements.txt
@@ -3,4 +3,9 @@ google-generativeai
 tokenizers
 langchain
 langchain_community
-openai
\ No newline at end of file
+openai
+sentence-transformers
+nltk
+spacy==3.7.4
+langchain-text-splitters
+langchain_openai
\ No newline at end of file
diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index f2a5e68..684192d 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -139,3 +139,5 @@ def get_transcript(file_path="./multi_benefit_transcript.txt"):
         """)
 if test:
     print(test)
+
+chunking_ingest(transcript=transcript, prompt=CHUNKING_PROMPT)

From 6b34a5beccfd14a84ee6822620b1db611a99b07c Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Fri, 17 May 2024 12:15:47 -0400
Subject: [PATCH 04/13] refactor to not instantiate with every message call

---
 04-call-summaries/llm.py |  92 ++++++++++++++++---------------
 04-call-summaries/run.py | 113 +++++++++++++++++++--------------------
 2 files changed, 104 insertions(+), 101 deletions(-)

diff --git a/04-call-summaries/llm.py b/04-call-summaries/llm.py
index 44cc6f1..b186f54 100644
--- a/04-call-summaries/llm.py
+++ b/04-call-summaries/llm.py
@@ -16,10 +16,7 @@ def get_transcript(file_path="./transcript.txt"):
 
 
 def ollama_client(
-    model_name=None,
-    prompt=None,
-    callbacks=None,
-    settings=None,
+    model_name=None, prompt=None, callbacks=None, settings=None, client=None
 ):
     if not settings:
         settings = {
@@ -32,13 +29,14 @@ def ollama_client(
 
     print("LLM settings:", model_name, settings)
     # To connect via another URL: Ollama(base_url='http://localhost:11434', ...)
-    return Ollama(model=model_name, callbacks=callbacks, **settings).invoke(prompt)
+
+    if client:
+        return client.invoke(prompt)
+    return Ollama(model=model_name, callbacks=callbacks, **settings)
 
 
 def google_gemini_client(
-    model_name="gemini-1.5-flash-latest",
-    prompt=None,
-    settings=None,
+    model_name="gemini-1.5-flash-latest", prompt=None, settings=None, client=None
 ):
     # Get a Google API key by following the steps after clicking on Get an API key button
     # at https://ai.google.dev/tutorials/setup
@@ -49,43 +47,49 @@ def google_gemini_client(
     genai.configure(api_key=GOOGLE_API_KEY)
     if settings:
         genai.GenerationConfig(**settings)
-    model = genai.GenerativeModel(model_name)
-    return model.generate_content(prompt)
-
-
-def gpt3_5(prompt, model="gpt-3.5-turbo"):
-    # Get API key from https://platform.openai.com/api-keys
-    OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
-    openai_client = OpenAI(api_key=OPENAI_API_KEY)  # Uses OPENAI_API_KEY
-    return (
-        openai_client.chat.completions.create(
-            model=model, messages=[{"role": "user", "content": prompt}]
+    if client:
+        return client.generate_content(prompt).text
+    return genai.GenerativeModel(model_name)
+
+
+def gpt_client(prompt=None, model_choice="gpt-4o", client=None):
+    if client:
+        if model_choice == "gpt3":
+            model = "gpt-3.5-turbo"
+        elif model_choice == "gpt4":
+            model = "gpt-4-turbo"
+        else:
+            model = model_choice
+        return (
+            client.chat.completions.create(
+                model=model, messages=[{"role": "user", "content": prompt}]
+            )
+            .choices[0]
+            .message.content
         )
-        .choices[0]
-        .message.content
-    )
-
-
-def gpt_4_turbo(prompt):
-    return gpt3_5(prompt, model="gpt-4-turbo")
-
+    else:
+        # Get API key from https://platform.openai.com/api-keys
+        OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
+        return OpenAI(api_key=OPENAI_API_KEY)  # Uses OPENAI_API_KEY
 
-def gpt_4o(prompt):
-    return gpt3_5(prompt, model="gpt-4o")
 
+def claude_client(
+    prompt=None, model="claude-3-opus-20240229", max_tokens=1024, client=None
+):
+    if client:
+        generated_response = client.messages.create(
+            model=model,
+            max_tokens=max_tokens,
+            messages=[{"role": "user", "content": prompt}],
+        ).content
+        text_response = "\n".join(
+            [text_block.text for text_block in generated_response]
+        )
 
-def claude(prompt, model="claude-3-opus-20240229", max_tokens=1024):
-    # Get API key from https://console.anthropic.com/settings/keys
-    ANTHROPIC_API_KEY = os.environ.get("ANTHROPIC_API_KEY")
-
-    client = anthropic.Anthropic(
-        api_key=ANTHROPIC_API_KEY,
-    )
-    generated_response = client.messages.create(
-        model=model,
-        max_tokens=max_tokens,
-        messages=[{"role": "user", "content": prompt}],
-    ).content
-    text_response = "\n".join([text_block.text for text_block in generated_response])
-
-    return text_response
+        return text_response
+    else:
+        # Get API key from https://console.anthropic.com/settings/keys
+        ANTHROPIC_API_KEY = os.environ.get("ANTHROPIC_API_KEY")
+        return anthropic.Anthropic(
+            api_key=ANTHROPIC_API_KEY,
+        )
diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index 684192d..525cd24 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -1,11 +1,11 @@
 from langchain_core.prompts import PromptTemplate
 
-from llm import claude, google_gemini_client, gpt3_5, gpt_4_turbo, gpt_4o, ollama_client
+from llm import claude_client, google_gemini_client, gpt_client, ollama_client
 from chunking import chunking_ingest
 
 
 # download transcripts from https://drive.google.com/drive/folders/19r6x3Zep4N9Rl_x4n4H6RpWkXviwbxyw?usp=sharing
-def get_transcript(file_path="./multi_benefit_transcript.txt"):
+def get_transcript(file_path="./transcript.txt"):
     file = open(file_path, encoding="utf-8")
     content = file.read()
     file.close()
@@ -35,24 +35,8 @@ def get_transcript(file_path="./multi_benefit_transcript.txt"):
 
 CHUNKING_PROMPT = """
 You are a helpful AI assistant tasked with summarizing transcripts, however we can only process the transcripts in pieces.
-Fill out the fields with the text given {text_given}. If the template already has the field filled out, do not overwrite this information. 
-If the template is blank, fill out the following template the best you can using the text:
-1. Caller Information:
-- Name
-- Contact Information
-- Availability
-- Household Information
-2. Reason/Type of Call: e.g., Applying for benefits, Follow-ups
-3. Previous Benefits History:
-- Applied for
-- Receives
-- Denied
-4. Benefits Discussion: Prefix the discussed benefit with a hashtag (e.g., #SNAP, #LIHEAP)
-5. Discussion Points:
-- Key information points
-6. Documents Needed: e.g., Income verification, Housing documentation
-7. Next Steps for Client
-8. Next Steps for Agent
+Fill out the fields with the text given {text}. If the following template already has the field filled out, do not overwrite this information.
+Please fill out the data with the following template: {template}
 """
 
 print("""
@@ -72,12 +56,14 @@ def get_transcript(file_path="./multi_benefit_transcript.txt"):
 formatted_prompt = prompt_template.format(transcript=transcript)
 
 if llm == "2":
-    test = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
+    ollama = ollama_client(model_name="dolphin-mistral")
+    response = ollama_client(client=ollama, prompt=formatted_prompt)
     print("""----------
         Dolphin
         """)
 elif llm == "3":
-    test = google_gemini_client(prompt=formatted_prompt).text
+    gemini = google_gemini_client()
+    response = google_gemini_client(client=gemini, prompt=formatted_prompt)
     print("""----------
         Gemini Flash 1.5
         """)
@@ -85,59 +71,72 @@ def get_transcript(file_path="./multi_benefit_transcript.txt"):
     print("""----------
       GPT 4
       """)
-    test = gpt_4_turbo(prompt=formatted_prompt)
+    gpt = gpt_client()
+    response = gpt_client(client=gpt, model_choice="gpt4", prompt=formatted_prompt)
 elif llm == "5":
     print("""----------
       GPT 4o
       """)
-    test = gpt_4o(prompt=formatted_prompt)
+    gpt = gpt_client()
+    response = gpt_client(client=gpt, model_choice="gpt4o", prompt=formatted_prompt)
 elif llm == "6":
     print("""----------
         Claude 3
         """)
-    test = claude(prompt=formatted_prompt)
+    claude = claude_client()
+    response = claude_client(client=claude, prompt=formatted_prompt)
 elif llm == "7":
-    test_open_hermes = ollama_client(model_name="openhermes", prompt=formatted_prompt)
-    print("""
-        Openhermes
-        """)
-    print(test_open_hermes)
+    # print("""
+    #     Openhermes
+    #     """)
+    # ollama_openhermes = ollama_client(model_name="openhermes")
+    # ollama_openhermes_response = ollama_client(
+    #     client=ollama_openhermes, prompt=formatted_prompt
+    # )
+    # print(ollama_openhermes_response)
 
-    test_dolphin = ollama_client(model_name="dolphin-mistral", prompt=formatted_prompt)
-    print("""----------
-        Dolphin
-        """)
-    print(test_dolphin)
+    # print("""----------
+    #     Dolphin
+    #     """)
+    # ollama_dolphin = ollama_client(model_name="dolphin-mistral")
+    # dolphin_response = ollama_client(client=ollama_dolphin, prompt=formatted_prompt)
+    # print(dolphin_response)
 
-    test_gemini = google_gemini_client(prompt=formatted_prompt).text
-    print("""----------
-        Gemini
-        """)
-    print(test_gemini)
+    # print("""----------
+    #     Gemini Flash 1.5
+    #     """)
+    # gemini = google_gemini_client()
+    # gemini_response = google_gemini_client(client=gemini, prompt=formatted_prompt)
+    # print(gemini_response)
 
-    print("""----------
-        GPT 4
-        """)
-    test_gpt4 = gpt_4_turbo(prompt=formatted_prompt)
-    print(test_gpt4)
+    # gpt = gpt_client()
+    # print("""----------
+    #   GPT 4
+    #   """)
+    # gpt_4_response = gpt_client(
+    #     client=gpt, model_choice="gpt4", prompt=formatted_prompt
+    # )
+    # print(gpt_4_response)
 
-    print("""----------
-        GPT 4o
-        """)
-    test_gpt4o = gpt_4o(prompt=formatted_prompt)
-    print(test_gpt4o)
+    # print("""----------
+    #   GPT 4o
+    #   """)
+    # gpt_4o_response = gpt_client(
+    #     client=gpt, model_choice="gpt-4o", prompt=formatted_prompt
+    # )
+    # print(gpt_4o_response)
 
     print("""----------
         Claude 3
         """)
-    test_claude = claude(prompt=formatted_prompt)
-    print(test_claude)
+    claude = claude_client()
+    claude_response = claude_client(client=claude, prompt=formatted_prompt)
+    print(claude_response)
 else:
-    test = ollama_client(model_name="openhermes", prompt=formatted_prompt)
+    ollama = ollama_client(model_name="openhermes")
+    response = ollama_client(client=ollama, prompt=formatted_prompt)
     print("""
         Openhermes
         """)
-if test:
-    print(test)
-
-chunking_ingest(transcript=transcript, prompt=CHUNKING_PROMPT)
+if response:
+    print(response)

From 693052020649cb5fdec2b8dfd5ad0237215d1acd Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Fri, 17 May 2024 14:08:46 -0400
Subject: [PATCH 05/13] create function for stuffing code

---
 04-call-summaries/run.py | 182 +++++++++++++++++++--------------------
 1 file changed, 91 insertions(+), 91 deletions(-)

diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index 525cd24..bbd04c2 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -39,104 +39,104 @@ def get_transcript(file_path="./transcript.txt"):
 Please fill out the data with the following template: {template}
 """
 
-print("""
-      Select an llm
-      1. openhermes (default)
-      2. dolphin
-      3. gemini
-      4. gpt 4
-      5. gpt 4o
-      6. claude 3
-      7. all
-      """)
 
 transcript = get_transcript()
-llm = input() or "1"
 prompt_template = PromptTemplate.from_template(PROMPT)
 formatted_prompt = prompt_template.format(transcript=transcript)
 
-if llm == "2":
-    ollama = ollama_client(model_name="dolphin-mistral")
-    response = ollama_client(client=ollama, prompt=formatted_prompt)
-    print("""----------
-        Dolphin
+
+def stuffing_summary(prompt=None):
+    print("""
+        Select an llm
+        1. openhermes (default)
+        2. dolphin
+        3. gemini
+        4. gpt 4
+        5. gpt 4o
+        6. claude 3
+        7. all
         """)
-elif llm == "3":
-    gemini = google_gemini_client()
-    response = google_gemini_client(client=gemini, prompt=formatted_prompt)
-    print("""----------
-        Gemini Flash 1.5
+
+    llm = input() or "1"
+
+    if llm == "2":
+        ollama = ollama_client(model_name="dolphin-mistral")
+        response = ollama_client(client=ollama, prompt=prompt)
+        print("""----------
+            Dolphin
+            """)
+    elif llm == "3":
+        gemini = google_gemini_client()
+        response = google_gemini_client(client=gemini, prompt=prompt)
+        print("""----------
+            Gemini Flash 1.5
+            """)
+    elif llm == "4":
+        print("""----------
+        GPT 4
         """)
-elif llm == "4":
-    print("""----------
-      GPT 4
-      """)
-    gpt = gpt_client()
-    response = gpt_client(client=gpt, model_choice="gpt4", prompt=formatted_prompt)
-elif llm == "5":
-    print("""----------
-      GPT 4o
-      """)
-    gpt = gpt_client()
-    response = gpt_client(client=gpt, model_choice="gpt4o", prompt=formatted_prompt)
-elif llm == "6":
-    print("""----------
-        Claude 3
+        gpt = gpt_client()
+        response = gpt_client(client=gpt, model_choice="gpt4", prompt=prompt)
+    elif llm == "5":
+        print("""----------
+        GPT 4o
         """)
-    claude = claude_client()
-    response = claude_client(client=claude, prompt=formatted_prompt)
-elif llm == "7":
-    # print("""
-    #     Openhermes
-    #     """)
-    # ollama_openhermes = ollama_client(model_name="openhermes")
-    # ollama_openhermes_response = ollama_client(
-    #     client=ollama_openhermes, prompt=formatted_prompt
-    # )
-    # print(ollama_openhermes_response)
-
-    # print("""----------
-    #     Dolphin
-    #     """)
-    # ollama_dolphin = ollama_client(model_name="dolphin-mistral")
-    # dolphin_response = ollama_client(client=ollama_dolphin, prompt=formatted_prompt)
-    # print(dolphin_response)
-
-    # print("""----------
-    #     Gemini Flash 1.5
-    #     """)
-    # gemini = google_gemini_client()
-    # gemini_response = google_gemini_client(client=gemini, prompt=formatted_prompt)
-    # print(gemini_response)
-
-    # gpt = gpt_client()
-    # print("""----------
-    #   GPT 4
-    #   """)
-    # gpt_4_response = gpt_client(
-    #     client=gpt, model_choice="gpt4", prompt=formatted_prompt
-    # )
-    # print(gpt_4_response)
-
-    # print("""----------
-    #   GPT 4o
-    #   """)
-    # gpt_4o_response = gpt_client(
-    #     client=gpt, model_choice="gpt-4o", prompt=formatted_prompt
-    # )
-    # print(gpt_4o_response)
-
-    print("""----------
-        Claude 3
+        gpt = gpt_client()
+        response = gpt_client(client=gpt, model_choice="gpt4o", prompt=prompt)
+    elif llm == "6":
+        print("""----------
+            Claude 3
+            """)
+        claude = claude_client()
+        response = claude_client(client=claude, prompt=prompt)
+    elif llm == "7":
+        print("""
+            Openhermes
+            """)
+        ollama_openhermes = ollama_client(model_name="openhermes")
+        ollama_openhermes_response = ollama_client(
+            client=ollama_openhermes, prompt=prompt
+        )
+        print(ollama_openhermes_response)
+
+        print("""----------
+            Dolphin
+            """)
+        ollama_dolphin = ollama_client(model_name="dolphin-mistral")
+        dolphin_response = ollama_client(client=ollama_dolphin, prompt=prompt)
+        print(dolphin_response)
+
+        print("""----------
+            Gemini Flash 1.5
+            """)
+        gemini = google_gemini_client()
+        gemini_response = google_gemini_client(client=gemini, prompt=prompt)
+        print(gemini_response)
+
+        gpt = gpt_client()
+        print("""----------
+        GPT 4
         """)
-    claude = claude_client()
-    claude_response = claude_client(client=claude, prompt=formatted_prompt)
-    print(claude_response)
-else:
-    ollama = ollama_client(model_name="openhermes")
-    response = ollama_client(client=ollama, prompt=formatted_prompt)
-    print("""
-        Openhermes
+        gpt_4_response = gpt_client(client=gpt, model_choice="gpt4", prompt=prompt)
+        print(gpt_4_response)
+
+        print("""----------
+        GPT 4o
         """)
-if response:
-    print(response)
+        gpt_4o_response = gpt_client(client=gpt, model_choice="gpt-4o", prompt=prompt)
+        print(gpt_4o_response)
+
+        print("""----------
+            Claude 3
+            """)
+        claude = claude_client()
+        claude_response = claude_client(client=claude, prompt=prompt)
+        print(claude_response)
+    else:
+        ollama = ollama_client(model_name="openhermes")
+        response = ollama_client(client=ollama, prompt=prompt)
+        print("""
+            Openhermes
+            """)
+    if response:
+        print(response)

From 93af4ab711688b758a1151480c2cc9638bc3e457 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Fri, 17 May 2024 16:25:23 -0400
Subject: [PATCH 06/13] implement chunking code

---
 04-call-summaries/chunking.py | 123 +++++++++++++++++++++++++++-------
 04-call-summaries/run.py      |   7 --
 2 files changed, 97 insertions(+), 33 deletions(-)

diff --git a/04-call-summaries/chunking.py b/04-call-summaries/chunking.py
index 2c619ad..8076e1a 100644
--- a/04-call-summaries/chunking.py
+++ b/04-call-summaries/chunking.py
@@ -3,18 +3,10 @@
     NLTKTextSplitter,
     SpacyTextSplitter,
 )
-from llm import ollama_client
-
-# from langchain_openai import OpenAIEmbeddings
-from langchain_community.embeddings.sentence_transformer import (
-    SentenceTransformerEmbeddings,
-)
-from langchain.chains.conversational_retrieval.base import ConversationalRetrievalChain
-from langchain.chains.llm import LLMChain
 from langchain_core.prompts import PromptTemplate
-
-from langchain_community.vectorstores.faiss import FAISS
 from langchain.docstore.document import Document
+from llm import google_gemini_client, claude_client, gpt_client, ollama_client
+from run import get_transcript
 
 
 # split text into chunks
@@ -42,8 +34,11 @@ def get_text_chunks(text, chunk_size, chunk_overlap, text_splitter_choice):
 
 CHUNKING_PROMPT = """
 You are a helpful AI assistant tasked with summarizing transcripts, however we can only process the transcripts in pieces.
-Fill out the fields with the text given {text}. If the template {template} already has the field filled out, do not overwrite this information.
-Please fill out the data with the following template:
+Fill out the fields with the text given {text}. If the following template already has the field filled out, do not overwrite this information.
+Please fill out the data with the following template: {template}
+"""
+
+initial_temp = """
 1. Caller Information:
 - Name
 - Contact Information
@@ -64,23 +59,99 @@ def get_text_chunks(text, chunk_size, chunk_overlap, text_splitter_choice):
 
 
 def chunking_ingest(transcript, prompt):
-    # text_splitter_choice= input() or "2"
     text_chunks = get_text_chunks(
         transcript, chunk_size=750, chunk_overlap=300, text_splitter_choice="2"
     )
-
-    finalized_template = ""
     prompt_template = PromptTemplate.from_template(prompt)
+    template = initial_temp
+
+    print("""
+        Select an llm
+        1. openhermes (default)
+        2. dolphin
+        3. gemini
+        4. gpt 4
+        5. gpt 4o
+        6. claude 3
+        """)
+
+    llm = input() or "1"
+
+    if llm == "2":
+        client = ollama_client(model_name="dolphin-mistral")
+        print("""----------
+            Dolphin
+            """)
+        for text in text_chunks:
+            formatted_prompt = prompt_template.format(
+                text=text.page_content, template=template
+            )
+            print("----------------------")
+            template = ollama_client(client=client, prompt=formatted_prompt)
+        return template
+    elif llm == "3":
+        gemini = google_gemini_client()
+        print("""----------
+            Gemini Flash 1.5
+            """)
+        for text in text_chunks:
+            formatted_prompt = prompt_template.format(
+                text=text.page_content, template=template
+            )
+            print("----------------------")
+            template = google_gemini_client(client=gemini, prompt=formatted_prompt)
+        return template
+    elif llm == "4":
+        print("""----------
+        GPT 4
+        """)
+        gpt = gpt_client()
+        for text in text_chunks:
+            formatted_prompt = prompt_template.format(
+                text=text.page_content, template=template
+            )
+            print("----------------------")
+            template = gpt_client(
+                client=gpt, model_choice="gpt4", prompt=formatted_prompt
+            )
+        return template
+    elif llm == "5":
+        print("""----------
+        GPT 4o
+        """)
+        for text in text_chunks:
+            formatted_prompt = prompt_template.format(
+                text=text.page_content, template=template
+            )
+            print("----------------------")
+            template = gpt_client(
+                client=gpt, model_choice="gpt-4o", prompt=formatted_prompt
+            )
+        return template
+    elif llm == "6":
+        print("""----------
+            Claude 3
+            """)
+        claude = claude_client()
+        for text in text_chunks:
+            formatted_prompt = prompt_template.format(
+                text=text.page_content, template=template
+            )
+            print("----------------------")
+            template = claude_client(client=claude, prompt=formatted_prompt)
+        return template
+    else:
+        print("""
+            Openhermes
+            """)
+        ollama = ollama_client(model_name="openhermes")
+        for text in text_chunks:
+            formatted_prompt = prompt_template.format(
+                text=text.page_content, template=template
+            )
+            print("----------------------")
+            template = ollama_client(client=ollama, prompt=formatted_prompt)
+        return template
 
-    print("TEXT CHUNK")
-    for text in text_chunks:
-        formatted_prompt = prompt_template.format(
-            text=text.page_content, template=finalized_template
-        )
-        print("formatted_prompt", formatted_prompt)
-        answer = ollama_client(model_name="openhermes", prompt=formatted_prompt)
-        print("answer______________________")
-        print(answer)
-        finalized_template += "\n" + answer
-    print(finalized_template)
 
+print(chunking_ingest(transcript=get_transcript(), prompt=CHUNKING_PROMPT))
diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index bbd04c2..0d23d4d 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -1,7 +1,6 @@
 from langchain_core.prompts import PromptTemplate
 
 from llm import claude_client, google_gemini_client, gpt_client, ollama_client
-from chunking import chunking_ingest
 
 
 # download transcripts from https://drive.google.com/drive/folders/19r6x3Zep4N9Rl_x4n4H6RpWkXviwbxyw?usp=sharing
@@ -33,12 +32,6 @@ def get_transcript(file_path="./transcript.txt"):
 8. Next Steps for Agent
 """
 
-CHUNKING_PROMPT = """
-You are a helpful AI assistant tasked with summarizing transcripts, however we can only process the transcripts in pieces.
-Fill out the fields with the text given {text}. If the following template already has the field filled out, do not overwrite this information.
-Please fill out the data with the following template: {template}
-"""
-
 
 transcript = get_transcript()
 prompt_template = PromptTemplate.from_template(PROMPT)

From 11e9db10a1bb0f9ad4a8ca298bf7ec4be6e13768 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Fri, 17 May 2024 16:33:37 -0400
Subject: [PATCH 07/13] fix instantiation err

---
 04-call-summaries/chunking.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/04-call-summaries/chunking.py b/04-call-summaries/chunking.py
index 8076e1a..03e0812 100644
--- a/04-call-summaries/chunking.py
+++ b/04-call-summaries/chunking.py
@@ -119,6 +119,7 @@ def chunking_ingest(transcript, prompt):
         print("""----------
         GPT 4o
         """)
+        gpt = gpt_client()
         for text in text_chunks:
             formatted_prompt = prompt_template.format(
                 text=text.page_content, template=template

From 26a079a39ce4cbede253c2b3cba5a0df8afca039 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Tue, 21 May 2024 10:52:48 -0400
Subject: [PATCH 08/13] refactor llm choices into class

---
 04-call-summaries/llm.py | 173 +++++++++++++++++++++------------------
 1 file changed, 92 insertions(+), 81 deletions(-)

diff --git a/04-call-summaries/llm.py b/04-call-summaries/llm.py
index b186f54..31990f5 100644
--- a/04-call-summaries/llm.py
+++ b/04-call-summaries/llm.py
@@ -9,87 +9,98 @@
 dotenv.load_dotenv()
 
 
-def get_transcript(file_path="./transcript.txt"):
-    file = open(file_path, encoding="utf-8")
-    content = file.read()
-    return content
-
-
-def ollama_client(
-    model_name=None, prompt=None, callbacks=None, settings=None, client=None
-):
-    if not settings:
-        settings = {
-            # "temperature": 0.1,
-            # "system": "",
-            # "template": "",
-            # See https://github.com/langchain-ai/langchain/blob/master/libs/community/langchain_community/llms/ollama.py
-            "stop": None
-        }
-
-    print("LLM settings:", model_name, settings)
-    # To connect via another URL: Ollama(base_url='http://localhost:11434', ...)
-
-    if client:
-        return client.invoke(prompt)
-    return Ollama(model=model_name, callbacks=callbacks, **settings)
-
-
-def google_gemini_client(
-    model_name="gemini-1.5-flash-latest", prompt=None, settings=None, client=None
-):
-    # Get a Google API key by following the steps after clicking on Get an API key button
-    # at https://ai.google.dev/tutorials/setup
-    GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
-
-    print("LLM settings:", model_name, settings)
-
-    genai.configure(api_key=GOOGLE_API_KEY)
-    if settings:
-        genai.GenerationConfig(**settings)
-    if client:
-        return client.generate_content(prompt).text
-    return genai.GenerativeModel(model_name)
-
-
-def gpt_client(prompt=None, model_choice="gpt-4o", client=None):
-    if client:
-        if model_choice == "gpt3":
-            model = "gpt-3.5-turbo"
-        elif model_choice == "gpt4":
-            model = "gpt-4-turbo"
-        else:
-            model = model_choice
-        return (
-            client.chat.completions.create(
-                model=model, messages=[{"role": "user", "content": prompt}]
+class LLM:
+    def __init__(
+        self,
+        client_name=None,
+        model_name=None,
+        max_tokens=1024,
+        settings=None,
+    ):
+        self.client_name = client_name
+        """Name of llm selection"""
+        self.model_name = model_name
+        """User friendly model name"""
+        self.model_version = model_name
+        """Exact model name being passed into the initializer"""
+        self.max_tokens = max_tokens
+        self.client = None
+        self.settings = settings
+
+    def get_client(self):
+        """Retrieves the llm client"""
+        if self.client_name == "ollama":
+            if not settings:
+                settings = {
+                    # "temperature": 0.1,
+                    # "system": "",
+                    # "template": "",
+                    # See https://github.com/langchain-ai/langchain/blob/master/libs/community/langchain_community/llms/ollama.py
+                    "stop": None
+                }
+            if self.model_name is None:
+                self.model_name = "openhermes"
+            # To connect via another URL: Ollama(base_url='http://localhost:11434', ...)
+            self.client = Ollama(model=self.model_version, **self.settings)
+
+        elif self.client_name == "gemini":
+            # Get a Google API key by following the steps after clicking on Get an API key button
+            # at https://ai.google.dev/tutorials/setup
+            GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
+            print("LLM settings:", self.model_name, settings)
+
+            if self.model_version is None:
+                self.model_version = "gemini-1.5-flash-latest"
+            elif self.model_name == "gemini-pro":
+                self.model_version = "gemini-1.5-pro-latest"
+
+            genai.configure(api_key=GOOGLE_API_KEY)
+            if settings:
+                genai.GenerationConfig(**self.settings)
+            self.client = genai.GenerativeModel(self.model_name)
+
+        elif self.client_name == "gpt":
+            if self.model_name is None:
+                self.model_version = "gpt-4o"
+            elif self.model_name == "gpt3":
+                self.model_version = "gpt-3.5-turbo"
+            elif self.model_name == "gpt4":
+                self.model_version = "gpt-4-turbo"
+
+            # Get API key from https://platform.openai.com/api-keys
+            OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
+            self.client = OpenAI(api_key=OPENAI_API_KEY)  # Uses OPENAI_API_KEY
+
+        elif self.client_name == "claude":
+            self.model_version = "claude-3-opus-20240229"
+            ANTHROPIC_API_KEY = os.environ.get("ANTHROPIC_API_KEY")
+            self.client = anthropic.Anthropic(
+                api_key=ANTHROPIC_API_KEY,
             )
-            .choices[0]
-            .message.content
-        )
-    else:
-        # Get API key from https://platform.openai.com/api-keys
-        OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
-        return OpenAI(api_key=OPENAI_API_KEY)  # Uses OPENAI_API_KEY
 
+    def generate_text(self, prompt=None):
+        """Generates response given prompt"""
+        if self.client_name == "ollama":
+            return self.client.invoke(prompt)
+        elif self.client_name == "gemini":
+            return self.client.generate(prompt).text
+        elif self.client_name == "gpt":
+            return (
+                self.client.chat.completions.create(
+                    model=self.model_version,
+                    messages=[{"role": "user", "content": prompt}],
+                )
+                .choices[0]
+                .message.content
+            )
+        elif self.client_name == "claude":
+            generated_response = self.client.messages.create(
+                model=self.model_version,
+                max_tokens=self.max_tokens,
+                messages=[{"role": "user", "content": prompt}],
+            ).content
+            text_response = "\n".join(
+                [text_block.text for text_block in generated_response]
+            )
 
-def claude_client(
-    prompt=None, model="claude-3-opus-20240229", max_tokens=1024, client=None
-):
-    if client:
-        generated_response = client.messages.create(
-            model=model,
-            max_tokens=max_tokens,
-            messages=[{"role": "user", "content": prompt}],
-        ).content
-        text_response = "\n".join(
-            [text_block.text for text_block in generated_response]
-        )
-
-        return text_response
-    else:
-        # Get API key from https://console.anthropic.com/settings/keys
-        ANTHROPIC_API_KEY = os.environ.get("ANTHROPIC_API_KEY")
-        return anthropic.Anthropic(
-            api_key=ANTHROPIC_API_KEY,
-        )
+            return text_response

From 94fbc1b848d8332780f3908eba329a5a09c92a62 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Tue, 21 May 2024 11:03:03 -0400
Subject: [PATCH 09/13] update stuffing code

---
 04-call-summaries/run.py | 51 ++++++++++++++++++++--------------------
 1 file changed, 25 insertions(+), 26 deletions(-)

diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index 0d23d4d..ef790ec 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -1,6 +1,6 @@
 from langchain_core.prompts import PromptTemplate
 
-from llm import claude_client, google_gemini_client, gpt_client, ollama_client
+from llm import LLM
 
 
 # download transcripts from https://drive.google.com/drive/folders/19r6x3Zep4N9Rl_x4n4H6RpWkXviwbxyw?usp=sharing
@@ -53,14 +53,14 @@ def stuffing_summary(prompt=None):
     llm = input() or "1"
 
     if llm == "2":
-        ollama = ollama_client(model_name="dolphin-mistral")
-        response = ollama_client(client=ollama, prompt=prompt)
+        ollama = LLM(client_name="ollama", model_name="dolphin-mistral")
+        response = ollama.generate_text(prompt=prompt)
         print("""----------
             Dolphin
             """)
     elif llm == "3":
-        gemini = google_gemini_client()
-        response = google_gemini_client(client=gemini, prompt=prompt)
+        gemini = LLM(client_name="gemini")
+        response = gemini.generate_text(prompt=prompt)
         print("""----------
             Gemini Flash 1.5
             """)
@@ -68,66 +68,65 @@ def stuffing_summary(prompt=None):
         print("""----------
         GPT 4
         """)
-        gpt = gpt_client()
-        response = gpt_client(client=gpt, model_choice="gpt4", prompt=prompt)
+        gpt = LLM(client_name="gpt", model_name="gpt4")
+        response = gpt.generate_text(prompt=prompt)
     elif llm == "5":
         print("""----------
         GPT 4o
         """)
-        gpt = gpt_client()
-        response = gpt_client(client=gpt, model_choice="gpt4o", prompt=prompt)
+        gpt = LLM(client_name="gpt", model_name="gpt4o")
+        response = gpt.generate_text(prompt=prompt)
     elif llm == "6":
         print("""----------
             Claude 3
             """)
-        claude = claude_client()
-        response = claude_client(client=claude, prompt=prompt)
+        claude = LLM(client_name="claude")
+        response = claude.generate_text(prompt=prompt)
     elif llm == "7":
         print("""
             Openhermes
             """)
-        ollama_openhermes = ollama_client(model_name="openhermes")
-        ollama_openhermes_response = ollama_client(
-            client=ollama_openhermes, prompt=prompt
-        )
+        ollama_openhermes = LLM(client_name="ollama", model_name="openhermes")
+        ollama_openhermes_response = ollama_openhermes.generate_text(prompt=prompt)
         print(ollama_openhermes_response)
 
         print("""----------
             Dolphin
             """)
-        ollama_dolphin = ollama_client(model_name="dolphin-mistral")
-        dolphin_response = ollama_client(client=ollama_dolphin, prompt=prompt)
+        ollama_dolphin = LLM(client_name="ollama", model_name="dolphin-mistral")
+        dolphin_response = ollama_dolphin.generate_text(prompt=prompt)
         print(dolphin_response)
 
         print("""----------
             Gemini Flash 1.5
             """)
-        gemini = google_gemini_client()
-        gemini_response = google_gemini_client(client=gemini, prompt=prompt)
+        gemini = LLM(client_name="gemini")
+        gemini_response = gemini.generate_text(prompt=prompt)
         print(gemini_response)
 
-        gpt = gpt_client()
         print("""----------
         GPT 4
         """)
-        gpt_4_response = gpt_client(client=gpt, model_choice="gpt4", prompt=prompt)
+        gpt_4 = LLM(client_name="gpt", model_name="gpt4")
+        gpt_4_response = gpt_4.generate_text(prompt=prompt)
         print(gpt_4_response)
 
         print("""----------
         GPT 4o
         """)
-        gpt_4o_response = gpt_client(client=gpt, model_choice="gpt-4o", prompt=prompt)
+        gpt_4o = LLM(client_name="_4o", model_name="gpt4o")
+        gpt_4o_response = gpt_4o.generate_text(prompt=prompt)
         print(gpt_4o_response)
 
         print("""----------
             Claude 3
             """)
-        claude = claude_client()
-        claude_response = claude_client(client=claude, prompt=prompt)
+        claude = LLM(client_name="claude")
+        claude_response = claude.generate_text(prompt=prompt)
         print(claude_response)
     else:
-        ollama = ollama_client(model_name="openhermes")
-        response = ollama_client(client=ollama, prompt=prompt)
+        ollama = LLM(client_name="ollama", model_name="openhermes")
+        response = ollama.generate_text(prompt=prompt)
         print("""
             Openhermes
             """)

From 460fc0933e6c644e7fb931d30f7162233eb67a86 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Tue, 21 May 2024 12:30:12 -0400
Subject: [PATCH 10/13] get client call for responses

---
 04-call-summaries/chunking.py | 73 +++++++++--------------------------
 04-call-summaries/llm.py      | 14 +++----
 04-call-summaries/run.py      | 29 ++++++++------
 3 files changed, 42 insertions(+), 74 deletions(-)

diff --git a/04-call-summaries/chunking.py b/04-call-summaries/chunking.py
index 03e0812..d875d32 100644
--- a/04-call-summaries/chunking.py
+++ b/04-call-summaries/chunking.py
@@ -5,7 +5,7 @@
 )
 from langchain_core.prompts import PromptTemplate
 from langchain.docstore.document import Document
-from llm import google_gemini_client, claude_client, gpt_client, ollama_client
+from llm import LLM
 from run import get_transcript
 
 
@@ -78,81 +78,46 @@ def chunking_ingest(transcript, prompt):
     llm = input() or "1"
 
     if llm == "2":
-        client = ollama_client(model_name="dolphin-mistral")
+        client = LLM(client_name="ollama", model_name="dolphin-mistral")
         print("""----------
             Dolphin
             """)
-        for text in text_chunks:
-            formatted_prompt = prompt_template.format(
-                text=text.page_content, template=template
-            )
-            print("----------------------")
-            template = ollama_client(client=client, prompt=formatted_prompt)
-        return template
+
     elif llm == "3":
-        gemini = google_gemini_client()
+        client = LLM(client_name="gemini")
         print("""----------
             Gemini Flash 1.5
             """)
-        for text in text_chunks:
-            formatted_prompt = prompt_template.format(
-                text=text.page_content, template=template
-            )
-            print("----------------------")
-            template = google_gemini_client(client=gemini, prompt=formatted_prompt)
-        return template
     elif llm == "4":
         print("""----------
         GPT 4
         """)
-        gpt = gpt_client()
-        for text in text_chunks:
-            formatted_prompt = prompt_template.format(
-                text=text.page_content, template=template
-            )
-            print("----------------------")
-            template = gpt_client(
-                client=gpt, model_choice="gpt4", prompt=formatted_prompt
-            )
-        return template
+        client = LLM(client_name="gpt", model_name="gpt4")
     elif llm == "5":
         print("""----------
         GPT 4o
         """)
-        gpt = gpt_client()
-        for text in text_chunks:
-            formatted_prompt = prompt_template.format(
-                text=text.page_content, template=template
-            )
-            print("----------------------")
-            template = gpt_client(
-                client=gpt, model_choice="gpt-4o", prompt=formatted_prompt
-            )
-        return template
+        client = LLM(client_name="gpt", model_name="gpt-4o")
     elif llm == "6":
         print("""----------
             Claude 3
             """)
-        claude = claude_client()
-        for text in text_chunks:
-            formatted_prompt = prompt_template.format(
-                text=text.page_content, template=template
-            )
-            print("----------------------")
-            template = claude_client(client=claude, prompt=formatted_prompt)
-        return template
+        client = LLM(client_name="claude")
     else:
         print("""
             Openhermes
             """)
-        ollama = ollama_client(model_name="openhermes")
-        for text in text_chunks:
-            formatted_prompt = prompt_template.format(
-                text=text.page_content, template=template
-            )
-            print("----------------------")
-            template = ollama_client(client=ollama, prompt=formatted_prompt)
-        return template
+        client = LLM(client_name="ollama", model_name="openhermes")
 
+    client.get_client()
+    for text in text_chunks:
+        formatted_prompt = prompt_template.format(
+            text=text.page_content, template=template
+        )
+        print("Processing Text Chunk")
+        template = client.generate_text(prompt=formatted_prompt)
+    print("Complete")
+    return template
 
-print(chunking_ingest(transcript=get_transcript(), prompt=CHUNKING_PROMPT))
+if __name__ == "__main__":
+    print(chunking_ingest(transcript=get_transcript(), prompt=CHUNKING_PROMPT))
diff --git a/04-call-summaries/llm.py b/04-call-summaries/llm.py
index 31990f5..9904adb 100644
--- a/04-call-summaries/llm.py
+++ b/04-call-summaries/llm.py
@@ -30,8 +30,8 @@ def __init__(
     def get_client(self):
         """Retrieves the llm client"""
         if self.client_name == "ollama":
-            if not settings:
-                settings = {
+            if self.settings is None:
+                self.settings = {
                     # "temperature": 0.1,
                     # "system": "",
                     # "template": "",
@@ -47,25 +47,23 @@ def get_client(self):
             # Get a Google API key by following the steps after clicking on Get an API key button
             # at https://ai.google.dev/tutorials/setup
             GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
-            print("LLM settings:", self.model_name, settings)
-
             if self.model_version is None:
                 self.model_version = "gemini-1.5-flash-latest"
             elif self.model_name == "gemini-pro":
                 self.model_version = "gemini-1.5-pro-latest"
 
             genai.configure(api_key=GOOGLE_API_KEY)
-            if settings:
+            if self.settings is not None:
                 genai.GenerationConfig(**self.settings)
             self.client = genai.GenerativeModel(self.model_name)
 
         elif self.client_name == "gpt":
-            if self.model_name is None:
-                self.model_version = "gpt-4o"
-            elif self.model_name == "gpt3":
+            if self.model_name == "gpt3":
                 self.model_version = "gpt-3.5-turbo"
             elif self.model_name == "gpt4":
                 self.model_version = "gpt-4-turbo"
+            else:
+                self.model_version = "gpt-4o"
 
             # Get API key from https://platform.openai.com/api-keys
             OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index ef790ec..65af70d 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -53,14 +53,12 @@ def stuffing_summary(prompt=None):
     llm = input() or "1"
 
     if llm == "2":
-        ollama = LLM(client_name="ollama", model_name="dolphin-mistral")
-        response = ollama.generate_text(prompt=prompt)
+        client = LLM(client_name="ollama", model_name="dolphin-mistral")
         print("""----------
             Dolphin
             """)
     elif llm == "3":
-        gemini = LLM(client_name="gemini")
-        response = gemini.generate_text(prompt=prompt)
+        client = LLM(client_name="gemini")
         print("""----------
             Gemini Flash 1.5
             """)
@@ -68,25 +66,23 @@ def stuffing_summary(prompt=None):
         print("""----------
         GPT 4
         """)
-        gpt = LLM(client_name="gpt", model_name="gpt4")
-        response = gpt.generate_text(prompt=prompt)
+        client = LLM(client_name="gpt", model_name="gpt4")
     elif llm == "5":
         print("""----------
         GPT 4o
         """)
-        gpt = LLM(client_name="gpt", model_name="gpt4o")
-        response = gpt.generate_text(prompt=prompt)
+        client = LLM(client_name="gpt", model_name="gpt4o")
     elif llm == "6":
         print("""----------
             Claude 3
             """)
-        claude = LLM(client_name="claude")
-        response = claude.generate_text(prompt=prompt)
+        client = LLM(client_name="claude")
     elif llm == "7":
         print("""
             Openhermes
             """)
         ollama_openhermes = LLM(client_name="ollama", model_name="openhermes")
+        ollama_openhermes.get_client()
         ollama_openhermes_response = ollama_openhermes.generate_text(prompt=prompt)
         print(ollama_openhermes_response)
 
@@ -94,6 +90,7 @@ def stuffing_summary(prompt=None):
             Dolphin
             """)
         ollama_dolphin = LLM(client_name="ollama", model_name="dolphin-mistral")
+        ollama_dolphin.get_client()
         dolphin_response = ollama_dolphin.generate_text(prompt=prompt)
         print(dolphin_response)
 
@@ -101,6 +98,7 @@ def stuffing_summary(prompt=None):
             Gemini Flash 1.5
             """)
         gemini = LLM(client_name="gemini")
+        gemini.get_client()
         gemini_response = gemini.generate_text(prompt=prompt)
         print(gemini_response)
 
@@ -108,6 +106,7 @@ def stuffing_summary(prompt=None):
         GPT 4
         """)
         gpt_4 = LLM(client_name="gpt", model_name="gpt4")
+        gpt_4.get_client()
         gpt_4_response = gpt_4.generate_text(prompt=prompt)
         print(gpt_4_response)
 
@@ -115,6 +114,7 @@ def stuffing_summary(prompt=None):
         GPT 4o
         """)
         gpt_4o = LLM(client_name="_4o", model_name="gpt4o")
+        gpt_4o.get_client()
         gpt_4o_response = gpt_4o.generate_text(prompt=prompt)
         print(gpt_4o_response)
 
@@ -122,13 +122,18 @@ def stuffing_summary(prompt=None):
             Claude 3
             """)
         claude = LLM(client_name="claude")
+        claude.get_client()
         claude_response = claude.generate_text(prompt=prompt)
         print(claude_response)
     else:
-        ollama = LLM(client_name="ollama", model_name="openhermes")
-        response = ollama.generate_text(prompt=prompt)
+        client = LLM(client_name="ollama", model_name="openhermes")
         print("""
             Openhermes
             """)
+    client.get_client()
+    response = client.generate_text(prompt=prompt)
     if response:
         print(response)
+
+if __name__ == "__main__":
+    stuffing_summary(prompt=formatted_prompt)

From 920ab0a55dd3ffcafbc64edb285f8212b7946f13 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Tue, 21 May 2024 12:51:48 -0400
Subject: [PATCH 11/13] formatted

---
 04-call-summaries/chunking.py | 1 +
 04-call-summaries/run.py      | 1 +
 2 files changed, 2 insertions(+)

diff --git a/04-call-summaries/chunking.py b/04-call-summaries/chunking.py
index d875d32..f3c7433 100644
--- a/04-call-summaries/chunking.py
+++ b/04-call-summaries/chunking.py
@@ -119,5 +119,6 @@ def chunking_ingest(transcript, prompt):
     print("Complete")
     return template
 
+
 if __name__ == "__main__":
     print(chunking_ingest(transcript=get_transcript(), prompt=CHUNKING_PROMPT))
diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index 65af70d..7500164 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -135,5 +135,6 @@ def stuffing_summary(prompt=None):
     if response:
         print(response)
 
+
 if __name__ == "__main__":
     stuffing_summary(prompt=formatted_prompt)

From 89e7171e1e53f3dbaa8af4bd2860816318528568 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Wed, 22 May 2024 10:57:24 -0400
Subject: [PATCH 12/13] fixes to gemini model default

---
 04-call-summaries/llm.py | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/04-call-summaries/llm.py b/04-call-summaries/llm.py
index 9904adb..b092fcd 100644
--- a/04-call-summaries/llm.py
+++ b/04-call-summaries/llm.py
@@ -47,15 +47,15 @@ def get_client(self):
             # Get a Google API key by following the steps after clicking on Get an API key button
             # at https://ai.google.dev/tutorials/setup
             GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
-            if self.model_version is None:
-                self.model_version = "gemini-1.5-flash-latest"
-            elif self.model_name == "gemini-pro":
+            if self.model_name == "gemini-pro":
                 self.model_version = "gemini-1.5-pro-latest"
+            else:
+                self.model_version = "gemini-1.5-flash-latest"
 
             genai.configure(api_key=GOOGLE_API_KEY)
             if self.settings is not None:
                 genai.GenerationConfig(**self.settings)
-            self.client = genai.GenerativeModel(self.model_name)
+            self.client = genai.GenerativeModel(self.model_version)
 
         elif self.client_name == "gpt":
             if self.model_name == "gpt3":
@@ -81,7 +81,7 @@ def generate_text(self, prompt=None):
         if self.client_name == "ollama":
             return self.client.invoke(prompt)
         elif self.client_name == "gemini":
-            return self.client.generate(prompt).text
+            return self.client.generate_content(prompt).text
         elif self.client_name == "gpt":
             return (
                 self.client.chat.completions.create(

From f5b604495df23b3bcbfbbb77252ecc10636a1353 Mon Sep 17 00:00:00 2001
From: C Cheng <charliecheng@navapbc.com>
Date: Wed, 22 May 2024 10:59:06 -0400
Subject: [PATCH 13/13] rename function call

---
 04-call-summaries/chunking.py | 18 ++++++++++++++----
 04-call-summaries/llm.py      |  2 +-
 04-call-summaries/run.py      | 14 +++++++-------
 3 files changed, 22 insertions(+), 12 deletions(-)

diff --git a/04-call-summaries/chunking.py b/04-call-summaries/chunking.py
index f3c7433..5aa10bf 100644
--- a/04-call-summaries/chunking.py
+++ b/04-call-summaries/chunking.py
@@ -1,3 +1,4 @@
+import datetime
 from langchain_text_splitters import (
     RecursiveCharacterTextSplitter,
     NLTKTextSplitter,
@@ -34,8 +35,8 @@ def get_text_chunks(text, chunk_size, chunk_overlap, text_splitter_choice):
 
 CHUNKING_PROMPT = """
 You are a helpful AI assistant tasked with summarizing transcripts, however we can only process the transcripts in pieces.
-Fill out the fields with the text given {text}. If the following template already has the field filled out, do not overwrite this information.
-Please fill out the data with the following template: {template}
+Please fill out and return the following template: {template} with data in the text: {text}
+If the following template already has the field filled out, do not overwrite this information.
 """
 
 initial_temp = """
@@ -109,7 +110,9 @@ def chunking_ingest(transcript, prompt):
             """)
         client = LLM(client_name="ollama", model_name="openhermes")
 
-    client.get_client()
+    client.init_client()
+    ct = datetime.datetime.now()
+    print("current time:-", ct)
     for text in text_chunks:
         formatted_prompt = prompt_template.format(
             text=text.page_content, template=template
@@ -121,4 +124,11 @@ def chunking_ingest(transcript, prompt):
 
 
 if __name__ == "__main__":
-    print(chunking_ingest(transcript=get_transcript(), prompt=CHUNKING_PROMPT))
+    print(
+        chunking_ingest(
+            transcript=get_transcript("./son_calling_behalf_mother_transcript.txt"),
+            prompt=CHUNKING_PROMPT,
+        )
+    )
+    ct = datetime.datetime.now()
+    print(ct)
diff --git a/04-call-summaries/llm.py b/04-call-summaries/llm.py
index b092fcd..b9e68c9 100644
--- a/04-call-summaries/llm.py
+++ b/04-call-summaries/llm.py
@@ -27,7 +27,7 @@ def __init__(
         self.client = None
         self.settings = settings
 
-    def get_client(self):
+    def init_client(self):
         """Retrieves the llm client"""
         if self.client_name == "ollama":
             if self.settings is None:
diff --git a/04-call-summaries/run.py b/04-call-summaries/run.py
index 7500164..b271d7b 100644
--- a/04-call-summaries/run.py
+++ b/04-call-summaries/run.py
@@ -82,7 +82,7 @@ def stuffing_summary(prompt=None):
             Openhermes
             """)
         ollama_openhermes = LLM(client_name="ollama", model_name="openhermes")
-        ollama_openhermes.get_client()
+        ollama_openhermes.init_client()
         ollama_openhermes_response = ollama_openhermes.generate_text(prompt=prompt)
         print(ollama_openhermes_response)
 
@@ -90,7 +90,7 @@ def stuffing_summary(prompt=None):
             Dolphin
             """)
         ollama_dolphin = LLM(client_name="ollama", model_name="dolphin-mistral")
-        ollama_dolphin.get_client()
+        ollama_dolphin.init_client()
         dolphin_response = ollama_dolphin.generate_text(prompt=prompt)
         print(dolphin_response)
 
@@ -98,7 +98,7 @@ def stuffing_summary(prompt=None):
             Gemini Flash 1.5
             """)
         gemini = LLM(client_name="gemini")
-        gemini.get_client()
+        gemini.init_client()
         gemini_response = gemini.generate_text(prompt=prompt)
         print(gemini_response)
 
@@ -106,7 +106,7 @@ def stuffing_summary(prompt=None):
         GPT 4
         """)
         gpt_4 = LLM(client_name="gpt", model_name="gpt4")
-        gpt_4.get_client()
+        gpt_4.init_client()
         gpt_4_response = gpt_4.generate_text(prompt=prompt)
         print(gpt_4_response)
 
@@ -114,7 +114,7 @@ def stuffing_summary(prompt=None):
         GPT 4o
         """)
         gpt_4o = LLM(client_name="_4o", model_name="gpt4o")
-        gpt_4o.get_client()
+        gpt_4o.init_client()
         gpt_4o_response = gpt_4o.generate_text(prompt=prompt)
         print(gpt_4o_response)
 
@@ -122,7 +122,7 @@ def stuffing_summary(prompt=None):
             Claude 3
             """)
         claude = LLM(client_name="claude")
-        claude.get_client()
+        claude.init_client()
         claude_response = claude.generate_text(prompt=prompt)
         print(claude_response)
     else:
@@ -130,7 +130,7 @@ def stuffing_summary(prompt=None):
         print("""
             Openhermes
             """)
-    client.get_client()
+    client.init_client()
     response = client.generate_text(prompt=prompt)
     if response:
         print(response)