Initial commit with LFS-tracked PDF

.gitignore

@@ -0,0 +1,5 @@
+__pycache__/
+*.pyc
+.env
+/data/
+chroma_db/

Dockerfile

@@ -0,0 +1,29 @@
+FROM python:3.10-slim
+
+# install OS deps needed by PyMuPDF / some packages
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    build-essential libgl1 git curl \
+    && rm -rf /var/lib/apt/lists/*
+
+# create a non-root user
+RUN useradd -m -u 1000 appuser
+USER appuser
+ENV PATH="/home/appuser/.local/bin:$PATH"
+WORKDIR /home/appuser/app
+
+# copy requirements and install
+COPY --chown=appuser:appuser requirements.txt /home/appuser/app/requirements.txt
+RUN python -m pip install --upgrade pip
+RUN pip install --no-cache-dir -r /home/appuser/app/requirements.txt
+
+# copy project files
+COPY --chown=appuser:appuser . /home/appuser/app
+
+# ensure HF cache + persistence go to /data
+ENV HF_HOME=/data/.huggingface
+ENV PERSIST_DIR=/data/chroma_db
+ENV RESUME_PATH=/home/appuser/app/media/resume/resume-ayush.pdf
+
+EXPOSE 7860
+# use uvicorn to serve FastAPI (HF will call this)
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", "--workers", "1"]

agent_startup.py

@@ -0,0 +1,52 @@
+import os
+from dotenv import load_dotenv
+load_dotenv()
+
+# Optionally write Google SA JSON (secret) to /data/sa.json for Gemini auth
+sa_json = os.getenv("GOOGLE_SERVICE_ACCOUNT_JSON")
+if sa_json:
+    sa_path = "/data/sa.json"
+    if not os.path.exists(sa_path):
+        with open(sa_path, "w", encoding="utf-8") as f:
+            f.write(sa_json)
+    os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = sa_path
+
+# --- embeddings + vectordb + llm (same as yours) ---
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_chroma import Chroma
+from langchain_google_genai import ChatGoogleGenerativeAI
+from langchain_community.document_loaders import PyMuPDFLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+
+print("loading embedding")
+embedding_model = HuggingFaceEmbeddings(
+    model_name="sentence-transformers/paraphrase-MiniLM-L3-v2",
+    model_kwargs={'device': "cpu"}
+)
+
+persist_dir = os.getenv("PERSIST_DIR", "/data/chroma_db")
+os.makedirs(persist_dir, exist_ok=True)
+
+print("loading vectordb")
+vectordb = Chroma(
+    persist_directory=persist_dir,
+    embedding_function=embedding_model,
+    collection_name="general_embeddings"
+)
+
+# Load resume into DB if empty
+try:
+    ids = vectordb.get().get("ids", [])
+except Exception:
+    ids = []
+if len(ids) == 0:
+    resume_path = os.getenv("RESUME_PATH", "media/resume/resume-ayush.pdf")
+    if os.path.exists(resume_path):
+        loader = PyMuPDFLoader(resume_path)
+        docs = loader.load()
+        splitter = RecursiveCharacterTextSplitter(chunk_size=580, chunk_overlap=60)
+        split_docs = splitter.split_documents(docs)
+        vectordb.add_documents(split_docs)
+
+print("loading llm")
+llm = ChatGoogleGenerativeAI(temperature=0.2, model=os.getenv("GEMINI_MODEL", "gemini-2.0-flash"))

app.py

@@ -0,0 +1,31 @@
+import os
+from fastapi import FastAPI, Request
+from fastapi.middleware.cors import CORSMiddleware
+from dotenv import load_dotenv
+from portfolio_agent import portfolio_agent
+
+load_dotenv()
+
+app = FastAPI()
+
+# Allow frontend access
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"], 
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.get("/")
+def root():
+    return {"status": "Portfolio Agent API is running"}
+
+@app.post("/chat")
+async def chat(request: Request):
+    data = await request.json()
+    message = data.get("message", "")
+    extra_context = os.getenv("extra_context")
+
+    answer = portfolio_agent(message, extra_context)
+    return {"reply": answer}

media/resume/resume-ayush.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74822c224e0dbbb1271533582f7738a4046ec42393e3660afaf25dc0a0bf45de
+size 108292

portfolio_agent.py

@@ -0,0 +1,33 @@
+from langchain_core.prompts import PromptTemplate
+from langchain.chains.retrieval import create_retrieval_chain
+from langchain.chains.combine_documents import create_stuff_documents_chain
+from agent_startup import vectordb, llm
+
+def portfolio_agent(question: str, extra_context: str):
+    prompt = PromptTemplate(
+        template="""You are my intelligent assistant, representing me to recruiters and HR professionals visiting my portfolio. 
+Use the following resume information: {context} and extra details: {extra_context} to answer questions as if you are me.
+
+Your goal is to provide clear, confident, and engaging responses that highlight my strengths, achievements, and suitability for exciting opportunities. 
+Be professional, personable, and persuasive. Where relevant, emphasize my unique skills, experience, and passion for growth.
+
+Question from recruiter/HR: {input}
+
+Your answer (as me):
+""",
+        input_variables=["context", "extra_context", "input"]
+    )
+
+    combine_docs_chain = create_stuff_documents_chain(llm, prompt)
+    rag_chain = create_retrieval_chain(
+        retriever=vectordb.as_retriever(search_kwargs={"k": 5}),
+        combine_docs_chain=combine_docs_chain,
+    )
+
+    result = rag_chain.invoke({
+        "context": "",
+        "extra_context": extra_context,
+        "input": question
+    })
+
+    return result["answer"]

requirements.txt

@@ -0,0 +1,10 @@
+fastapi
+uvicorn
+python-dotenv
+langchain
+langchain-core
+langchain-chroma
+langchain-community
+langchain-huggingface
+langchain-google-genai
+PyMuPDF

runtime.txt

@@ -0,0 +1 @@
+python-3.10