Update the deployment with correct model names, requirements and base…

… image (#39)

app/app.py

@@ -106,11 +106,12 @@ def OPENAI_API_KEY() -> str:
     model_id = st.selectbox(
         label="Model",
         options=[
-            "ibm/granite-20b-code-instruct-v1",
+            "ibm/granite-20b-code-instruct",
             "codellama/codellama-34b-instruct",
-            "meta-llama/llama-2-13b",
-            "ibm/granite-3b-code-plus-v1",
+            "meta-llama/llama-2-13b-chat",
+            "ibm/granite-3b-code-instruct",
             "meta-llama/llama-2-70b",
+            "mistralai/mistral-7b-v0-1",
             "OpenAI/gpt3.5",
             "bigcode/starcoder",
             "tiiuae/falcon-180b",
@@ -251,9 +252,21 @@ def main(prompt_success: bool, prompt_diff: int, actual_doc: str):
     with col3:
         st.subheader("Evaluation Metrics")
 
+        # ROUGE scores
+        st.markdown(
+            "**Automatic evaluation using ROUGE metrics:**",
+            help="Using predefined metrics to evaluate the responses generated by the LLM"
+        )
+
+        scorer = rouge_scorer.RougeScorer(['rouge1', 'rouge2', 'rougeL'], use_stemmer=True)
+        rouge_scores = scorer.score(actual_doc, result)
+        st.markdown(f"ROUGE-1 Score:{rouge_scores['rouge1'].fmeasure:.2f}", help="ROUGE-1 refers to the overlap of unigrams (each word) between the system and reference summaries")
+        st.markdown(f"ROUGE-2 Score: {rouge_scores['rouge2'].fmeasure:.2f}", help="ROUGE-2 refers to the overlap of bigrams between the system and reference summaries")
+        st.markdown(f"ROUGE-L Score: {rouge_scores['rougeL'].fmeasure:.2f}", help="Longest common subsequence problem takes into account sentence-level structure similarity naturally and identifies longest co-occurring in sequence n-grams automatically")
+
         st.markdown(
             "**LLM based evaluation on logic, correctness and helpfulness:**",
-            help="Use Langchain Criteria based Eval to evaluate on cutsom criteria (this list can be updated based on what we are looking to see from the generated docs). Note this is language mo0del based evaluation and not always a true indication of the quality of the output that is generatged."
+            help="Use Langchain Criteria based Eval to evaluate on cutsom criteria (this list can be updated based on what we are looking to see from the generated docs). Note this is language model based evaluation and not always a true indication of the quality of the output that is generated."
         )
 
         lc_score = eval_using_langchain(result, prompt, actual_doc)

app/requirements.txt

@@ -1,7 +1,7 @@
 html2text
 # new ibm version breaks existing code,
 # use this for the time being
-ibm-generative-ai>=2.0.0
+ibm-generative-ai==2.0.0
 python-dotenv
 streamlit-toggle-switch
 streamlit
@@ -11,4 +11,5 @@ openai
 textstat
 scikit-learn
 streamlit-feedback
-langchain
+langchain
+langchain-community

app/resources/deployment.yaml

@@ -20,7 +20,7 @@ spec:
             claimName: user-feedback-db
       containers:
       - name: api-doc-container
-        image: docker.io/ochatterjee/api-docs-gen:v0.1.0
+        image: docker.io/hemaveeradhi/api-docs-gen:v0.2.0
         imagePullPolicy: Always
         env:
           - name: GENAI_KEY