Remove OpenAI dependencies and references (#178)

* Remove OpenAI dependencies and references

- Remove openai package from backend/pyproject.toml dependencies
- Delete backend/scripts/create_vector_store.py (OpenAI-only script)
- Update frontend privacy policy to reference Google policies instead of OpenAI
- Remove OpenAI package installation from GitHub workflow
- Update documentation to reflect migration from OpenAI to Google Gemini
- Update .gitignore patterns from GPT-4 to Gemini
- Fix misleading comments in conversation generation script

Fixes #174

Co-authored-by: yangm2 <[email protected]>

* Remove OpenAI dependencies and references

- Remove openai package from backend/pyproject.toml dependencies
- Delete backend/scripts/create_vector_store.py (OpenAI-only script)
- Update frontend privacy policy to reference Google policies instead of OpenAI
- Update documentation to reflect migration from OpenAI to Google Gemini
- Update .gitignore patterns from GPT-4 to Gemini
- Fix misleading comments in conversation generation script

Note: Workflow file .github/workflows/generate_conversations.yml still needs manual update due to permissions (remove openai from pip install and OPENAI_API_KEY env var)

Fixes #174

Co-authored-by: yangm2 <[email protected]>

---------

Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>

Author: yangm2

.gitignore

@@ -55,10 +55,10 @@ build/
 chatlog.jsonl
 /backend/chatlog.jsonl
 /backend/combined_training.jsonl
-/backend/combined_training_gpt-4.1.jsonl
+/backend/combined_training_gemini-2.0.jsonl
 /backend/feedback.jsonl
 /backend/data
 /backend/scripts/eval_results.json
 /backend/scripts/generate_conversation/*.csv
 !/backend/scripts/generate_conversation/tenant_questions_facts_full.csv
-combined_training_gpt-4_1.jsonl
+combined_training_gemini-2.0.jsonl

README.md

@@ -13,11 +13,10 @@ Live at https://tenantfirstaid.com/
  - [uv](https://docs.astral.sh/uv/getting-started/installation/)
  - [docker](https://www.docker.com/)
 
-1. copy `backend/.env.example` to a new file named `.env` in the same directory and populate it with your `OPENAI_API_KEY`. You can set an invalid key, in which case the bot will return error messages. This may still be useful for developing other features.
+1. copy `backend/.env.example` to a new file named `.env` in the same directory. The chatbot now uses Google Gemini instead of OpenAI.
 1. `cd backend`
 1. `docker-compose up` (use `-d` if you want to run this in the background, otherwise open a new terminal)
 1. `uv sync`
-1. If you have not uploaded the Oregon Housing Law documents to a vector store in OpenAI, run `uv run scripts/create_vector_store.py` and follow the instructions to add the vector store ID to your `.env`.
 1. `uv run python -m tenantfirstaid.app`
 1. Open a new terminal / tab
 1. `cd ../frontend`

backend/.env.example

@@ -1,7 +1,7 @@
 # Specify a different model
 MODEL_NAME=gpt-2.5-flash
 
-# Vector store ID for OpenAI (use the create_vector_store script to create one)
+# Vector store ID for Gemini (deprecated - no longer needed)
 VECTOR_STORE_ID=my_vector_store_id
 
 # DB Info

backend/pyproject.toml

@@ -15,7 +15,6 @@ dependencies = [
   "google-auth>=2.40.3",
   "google-genai>=1.28.0",
   "google-cloud-aiplatform>=1.106.0",
-  "openai==1.89",
   "jsonlines",
   "simplejson",
   "python-dotenv",
@@ -48,5 +47,4 @@ dev = [
 
 gen_convo = [
       "pandas",
-      "openai",
 ]

backend/scripts/create_vector_store.py

@@ -1,7 +1,6 @@
 # /// script
 # requires-python = "~=3.11"
 # dependencies = [
-#     "openai",
 #     "pandas",
 #     "python-dotenv",
 # ]
@@ -68,7 +67,7 @@ def _reverse_message_roles(self, messages):
         return reversed_messages
 
     def bot_response(self):
-        """Generates a response from the bot using the OpenAI API."""
+        """Generates a response from the bot using the Gemini API."""
         tries = 0
         while tries < 3:
             # Use the BOT_INSTRUCTIONS for bot responses
@@ -90,7 +89,7 @@ def bot_response(self):
         return failure_message, None
 
     def user_response(self):
-        """Generates a response from the user using the OpenAI API."""
+        """Generates a response from the user using the Gemini API."""
         tries = 0
         while tries < 3:
             try:
@@ -217,4 +216,4 @@ def process_csv(input_file=None, output_file=None, num_turns=5, num_rows=None):
     process_csv(
         num_turns=args.num_turns, num_rows=args.num_rows, output_file=args.output_file
     )
-# This script generates conversations between a user and a bot using the OpenAI API.
+# This script generates conversations between a user and a bot using the Gemini API.

backend/scripts/generate_conversation/chat.py

@@ -1,6 +1,6 @@
 # Remote server setup
 
-Ubuntu LTS 24.04 hosted on Digital Ocean with 2 CPUs and 2GB RAM. Stack includes Certbot, Porkbun, OpenAI, and Nginx.
+Ubuntu LTS 24.04 hosted on Digital Ocean with 2 CPUs and 2GB RAM. Stack includes Certbot, Porkbun, Google Gemini, and Nginx.
 
 This folder is just a copy of the things on the remote server. There is no Github action or similar to sync these.
 

config/README.md

@@ -77,19 +77,19 @@ export default function PrivacyPolicy() {
           <li>
             Other Venders: Our Services connect with other technology platforms
             and anonymously shared/stored data is governed by their policies
-            (OpenAI:{" "}
+            (Google:{" "}
             <Link
-              to="https://openai.com/policies/privacy-policy"
+              to="https://policies.google.com/privacy"
               className="underline text-blue-600"
             >
               Privacy Policy
             </Link>{" "}
             and{" "}
             <Link
-              to="https://openai.com/policies/terms-of-use"
+              to="https://policies.google.com/terms"
               className="underline text-blue-600"
             >
-              Terms of Use
+              Terms of Service
             </Link>
             )
           </li>