first commit

.gitignore

@@ -0,0 +1,10 @@
+# Python-generated files
+__pycache__/
+*.py[oc]
+build/
+dist/
+wheels/
+*.egg-info
+
+# Virtual environments
+.venv

.sesskey

@@ -0,0 +1 @@
+3ae1db93-9b8c-44d1-a84f-7fdc45d0dc4d

main.py

@@ -0,0 +1,150 @@
+import os
+
+import httpx
+import streamlit as st
+from google import genai
+from pydantic import BaseModel
+
+
+class Article(BaseModel):
+    id: str
+    title: str
+    summary: str
+    abstract: str
+    url: str
+
+
+class Citation(BaseModel):
+    source_id: str
+
+
+class Statement(BaseModel):
+    text: str
+    citation: Citation | None
+
+
+class Answer(BaseModel):
+    sentences: list[Statement]
+
+
+def improve_prompt(client: genai.Client, prompt: str) -> str:
+    response = client.models.generate_content(
+        model="gemini-2.0-flash-lite",
+        contents=f"Rewrite the following medical question to maximize clarity and specificity for optimal search results.\n\n<query>{prompt}</query>\n\nReturn only the improved query.",
+    )
+    return response.text or ""
+
+
+def format_sources(articles: list[Article]) -> str:
+    sources = []
+    for article in articles:
+        source = f"<source id={article.id}><title>{article.title}</title>"
+        if article.abstract:
+            source += f"<abstract>{article.abstract}</abstract>"
+        if article.summary:
+            source += f"<summary>{article.summary}</summary>"
+        source += "</source>"
+        sources.append(source)
+    return "\n".join(sources)
+
+
+def generate_answer(
+    client: genai.Client, query: str, articles: list[Article]
+) -> list[Statement]:
+    response = client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=f"Answer the query based solely on the provided sources. The answer should be less than 100 words. Justify the answer by citing from the sources. Refuse to answer non-medical related query.\n\n<query>{query}</query>\n\n<sources>{format_sources(articles)}</sources>",
+        config={
+            "response_mime_type": "application/json",
+            "response_schema": Answer,
+        },
+    )
+    return response.parsed.sentences
+
+
+def semantic_scholar(query: str, top_k: int = 10) -> list[Article]:
+    resp = httpx.get(
+        "https://api.semanticscholar.org/graph/v1/paper/search?query",
+        params={
+            "query": query,
+            "limit": top_k,
+            "fields": "title,tldr,abstract,externalIds,url",
+            "fieldsOfStudy": "Medicine,Biology",
+            "minCitationCount": 20,
+        },
+    )
+    results = resp.json()
+    if results["total"]:
+        articles = [
+            Article(
+                id=str(i),
+                title=article["title"],
+                summary=article["tldr"]["text"] if article["tldr"] else "",
+                abstract=article["abstract"] or "",
+                url=article["url"],
+            )
+            for i, article in enumerate(results["data"], 1)
+            if article
+        ]
+    else:
+        articles = []
+    return articles
+
+
+def pubmed(query: str, top_k: int = 10, db: str = "pubmed"):
+    resp = httpx.get(
+        "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi",
+        params={
+            "term": query,
+            "db": db,
+            "retmax": top_k,
+            "retmode": "json",
+        },
+    )
+    id_list = resp.json()["esearchresult"]["idlist"]
+
+    resp = httpx.get(
+        "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi",
+        params={
+            "db": db,
+            "id": ",".join(id_list),
+            "retmode": "json",
+        },
+    )
+    return resp.json()
+
+
+def main():
+    client = genai.Client(api_key=os.environ["GOOGLE_API_KEY"])
+
+    st.title("Ask ~~Jeeves~~ Elna")
+    with st.form("search", border=False):
+        query = st.text_input("Your medical question")
+        submit = st.form_submit_button("Ask")
+
+    if submit:
+        papers = semantic_scholar(query, top_k=2)
+        if papers:
+            paper_titles = {paper.id: paper for paper in papers}
+            statements = generate_answer(client, query, papers)
+            answer_str = ""
+            citations = {}
+            for statement in statements:
+                answer_str += statement.text
+                if statement.citation:
+                    if citation_id := citations.get(statement.citation.source_id):
+                        answer_str += f"[{citation_id}]"
+                    else:
+                        citations[statement.citation.source_id] = len(citations)
+                        answer_str += f"[{citations[statement.citation.source_id]}]"
+                    answer_str += " "
+
+            st.write(answer_str)
+            if citations:
+                st.subheader("Citations")
+                for k, v in citations.items():
+                    st.write(f"[{v}] [{paper_titles[k].title}]({paper_titles[k].url})")
+
+
+if __name__ == "__main__":
+    main()

pyproject.toml

@@ -0,0 +1,36 @@
+[project]
+name = "elna"
+version = "0.1.0"
+description = "Add your description here"
+readme = "README.md"
+requires-python = ">=3.12"
+dependencies = [
+    "google-genai>=1.14.0",
+    "httpx>=0.28.1",
+    "pydantic>=2.11.4",
+    "python-fasthtml>=0.12.15",
+    "streamlit>=1.45.0",
+]
+
+[tool.ruff.lint]
+preview = true
+select = [
+    "E",    # pycodestyle
+    "W",    # pycodestyle warnings
+    "F",    # Pyflakes
+    "I",    # isort
+    "B",    # flake8-bugbear
+    "UP",   # pyupgrade
+    "SIM",  # flake8-simplify
+    "C4",   # flake8-comprehensions
+    "FURB", # refurb
+    "RUF",  # ruff
+]
+ignore = [
+    "F401", # Module imported but unused
+    "E501", # Line too long
+]
+
+[tool.pyright]
+venv = ".venv"
+venvPath = "."