feat: Prep Dockerfile for AWS deployments (#26)

.github/workflows/push-image.yml

@@ -62,7 +62,8 @@ jobs:
       if: inputs.build_image
       run: |
         cd ${{ inputs.dockerfile_folder }}
-        docker build -t "${{ inputs.image_name }}:${{ inputs.image_tag }}" .
+        echo "${{secrets.DOT_ENV_FILE_CONTENTS}}" > .env
+        docker build -t "${{ inputs.image_name }}:${{ inputs.image_tag }}" --build-arg GURU_CARDS_URL="https://docs.google.com/uc?export=download&id=${{ secrets.GURU_CARDS_URL_ID }}" .
 
     - name: "Publish image to AWS ECR'"
       if: inputs.build_image

02-household-queries/.chainlit/config.toml

@@ -1,7 +1,6 @@
 [project]
 # Whether to enable telemetry (default: true). No personal data is collected.
-enable_telemetry = true
-
+enable_telemetry = false
 
 # List of environment variables to be provided by each user to use the app.
 user_env = []
@@ -23,7 +22,7 @@ allow_origins = ["*"]
 prompt_playground = true
 
 # Process and display HTML in messages. This can be a security risk (see https://stackoverflow.com/questions/19603097/why-is-it-dangerous-to-render-user-generated-html-or-javascript)
-unsafe_allow_html = false
+unsafe_allow_html = true
 
 # Process and display mathematical expressions. This can clash with "$" characters in messages.
 latex = false
@@ -46,22 +45,22 @@ latex = false
 name = "Household Chatbot"
 
 # Show the readme while the thread is empty.
-show_readme_as_default = true
+show_readme_as_default = false
 
 # Description of the app and chatbot. This is used for HTML tags.
 # description = ""
 
 # Large size content are by default collapsed for a cleaner ui
-default_collapse_content = true
+default_collapse_content = false
 
 # The default value for the expand messages settings.
-default_expand_messages = false
+default_expand_messages = true
 
 # Hide the chain of thought details from the user in the UI.
 hide_cot = false
 
 # Link to your github repo. This will add a github button in the UI's header.
-# github = ""
+github = "https://github.com/navapbc/labs-gen-ai-experiments/tree/main/02-household-queries"
 
 # Specify a CSS file that can be used to customize the user interface.
 # The CSS file can be served from the public directory or via an external link.

02-household-queries/.dockerignore

@@ -0,0 +1,9 @@
+__pycache__
+chroma_db/
+*cache/
+*.log
+log/
+compiled_module-*.json
+
+*.DS_STORE
+

02-household-queries/Dockerfile

@@ -2,23 +2,30 @@ FROM python:3.11-slim
 
 WORKDIR /app
 
-# RUN apt-get update && apt-get install -y \
-#     build-essential \
-#     curl \
-#     software-properties-common \
-#     git \
-#     && rm -rf /var/lib/apt/lists/*
-
-COPY . .
-RUN pip3 install -r requirements.txt
+COPY requirements.txt .
+RUN pip3 install --no-cache-dir -r requirements.txt
 
 # Associate GHCR package with GitHub repo; otherwise, it's associated with the navapbc GitHub organization
 # See https://github.com/orgs/navapbc/packages?repo_name=labs-gen-ai-experiments
 # LABEL org.opencontainers.image.source https://github.com/navapbc/labs-gen-ai-experiments
 
+RUN apt-get update && apt-get install -y \
+    curl unzip \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY . .
+ARG GURU_CARDS_URL
+RUN echo "Downloading from ${GURU_CARDS_URL}" \
+    && curl -L "${GURU_CARDS_URL}" > download.zip \
+    && unzip -o download.zip \
+    && rm download.zip \
+    && mv guru_cards_for_nava--Multi-benefit.json guru_cards_for_nava.json
+
+RUN ./decompose-questions.py 0
+
 EXPOSE 8000
 HEALTHCHECK CMD curl http://localhost:8000 || exit 1
-ENTRYPOINT ["chainlit", "run", "--port", "8000", "-h", "chainlit-household-bot.py"]
+ENTRYPOINT ["chainlit", "run", "--port", "8000", "-h", "chainlit-summaries.py"]
 
 # To test:
 # docker build -t chainlit1 .

02-household-queries/chainlit-summaries.py

@@ -17,6 +17,13 @@
 @cl.on_chat_start
 async def init_chat():
     # settings = das.init()
+
+    # elements = [
+    #     cl.Text(name="side-text", display="side", content="Side Text"),
+    #     cl.Text(name="page-text", display="page", content="Page Text"),
+    # ]
+    # await cl.Message("hello side-text and page-text", elements=elements).send()
+
     pass
 
 
@@ -25,19 +32,37 @@ async def message_submitted(message: cl.Message):
     generated_results = on_question(message.content)
     print(json.dumps(dataclasses.asdict(generated_results), indent=2))
 
-    await cl.Message(format_as_markdown(generated_results)).send()
+    message_args = format_as_markdown(generated_results)
+    # await cl.Message(**message_args).send()
+    await cl.Message(content=message_args["content"], elements=message_args["elements"]).send()
+    # await cl.Message(content= message_args["content"], elements=[cl.Text(content="SIDE", display="inline")]).send()
 
 
 def format_as_markdown(gen_results):
-    resp = ["", f"## Q: {gen_results.question}", "### Derived Questions"]
+    resp = ["", f"## Q: {gen_results.question}"]
 
+    dq_resp = ["<details><summary>Derived Questions</summary>", ""]
     for dq in gen_results.derived_questions:
-        resp.append(f"1. {dq.derived_question}")
+        dq_resp.append(f"- {dq.derived_question}")
+    dq_resp += ["</details>", ""]
 
-    resp.append("### Guru cards")
-    for card in gen_results.cards:
+    cards_resp = []
+    for i, card in enumerate(gen_results.cards, 1):
         if card.summary:
-            resp += [f"#### {card.card_title}", f"Summary: {card.summary}", ""]
-            resp += [f"\n.\n> {q}" for q in card.quotes]
+            cards_resp += [
+                f"<details><summary>{i}. <a href='https://link/to/guru_card'>{card.card_title}</a></summary>",
+                "",
+                f"   Summary: {card.summary}",
+                "",
+            ]
+            indented_quotes = [q.strip().replace("\n", "\n   ") for q in card.quotes]
+            cards_resp += [f"\n   Quote:\n   ```\n   {q}\n   ```" for q in indented_quotes]
+            cards_resp += ["</details>", ""]
 
-    return "\n".join(resp)
+    return {
+        "content": "\n".join(resp + dq_resp + cards_resp),
+        "elements": [
+            # cl.Text(name="Derived Questions", content="\n".join(dq_resp), display="side"),
+            # cl.Text(name="Guru Cards", content="\n".join(cards_resp), display="inline")
+        ],
+    }

02-household-queries/decompose_and_summarize.py

@@ -180,6 +180,7 @@ def collect_retrieved_cards(derived_qs, gen_results):
     gen_results.cards = collate_by_card_score_sum(gen_results.derived_questions)
 
 
+@debugging.timer
 def collate_by_card_score_sum(derived_question_entries):
     all_retrieved_cards = dict()
     for dq_entry in derived_question_entries:
@@ -230,6 +231,7 @@ def retrieve_cards(derived_qs, vectordb, retrieve_k=5):
     return results
 
 
+@debugging.timer
 def create_summaries(gen_results, summarizer, guru_card_texts):
     print(f"Summarizing {len(gen_results.cards)} retrieved cards...")
     for i, card_entry in enumerate(gen_results.cards):
@@ -246,6 +248,7 @@ def create_summaries(gen_results, summarizer, guru_card_texts):
         card_entry.summary = prediction.answer
 
 
+@debugging.timer
 def format_response(gen_results):
     resp = ["==="]
     resp.append("Q: {gen_results.question}")