utilizing llama index can net an extra $1000 in winnings

README.md

@@ -4,7 +4,7 @@ emoji: 💬
 colorFrom: yellow
 colorTo: purple
 sdk: gradio
-sdk_version: 5.32.1
+sdk_version: 5.33.1
 app_file: app.py
 pinned: false
 license: apache-2.0

docs/plans/timeline.md

@@ -0,0 +1,11 @@
+Here are the key dates for the Gradio Agents & MCP Hackathon you’ll want to keep on your radar:
+
+* **May 20 – 26, 2025**: Pre-Hackathon announcements period ([huggingface.co][1])
+* **June 2 – 10, 2025**: Official hackathon window (also the period during which sign-ups remain open) ([huggingface.co][1])
+* **June 3, 2025 — 9 AM PST / 4 PM UTC**: Live kickoff YouTube event ([huggingface.co][1])
+* **June 4 – 5, 2025**: Gradio Office Hours with MCP Support (Abubakar), MistralAI, LlamaIndex, Custom Components team, and Sambanova ([huggingface.co][1])
+* **June 10, 2025 — 11:59 PM UTC**: Final submission deadline (last chance to push your Space and to join the org for sponsor credits) ([huggingface.co][1])
+* **June 11 – 16, 2025**: Judging period ([huggingface.co][1])
+* **June 17, 2025**: Winners announced ([huggingface.co][1])
+
+[1]: https://huggingface.co/Agents-MCP-Hackathon "Agents-MCP-Hackathon (Agents-MCP-Hackathon)"

docs/plans/work_plan.md

@@ -0,0 +1,27 @@
+Focus on Track 1. Trying to build two full projects in four days will leave you with nothing polished. Here’s your kill-plan for Track 1:
+
+1. **Today (June 6):**
+
+   * Finalize your symptom→ICD mapping module and JSON output.
+   * Stand up a minimal Gradio/MCP endpoint locally.
+
+2. **June 7:**
+
+   * Wire up the front-end UI (Gradio space).
+   * Hook it into Sonova (or your local HuggingFace‐hosted model) and verify the API calls work.
+
+3. **June 8:**
+
+   * Write README + usage docs.
+   * Deploy your Space to HF.
+
+4. **June 9:**
+
+   * End-to-end testing with 5–10 real symptom examples.
+   * Polish prompts, confidence scores, and UI copy.
+
+5. **June 10 (morning):**
+
+   * Final sanity check, package, and submit before end of day.
+
+Only if you finish these steps by end of Day 2 should you consider a tiny Track 2 add-on (e.g. a single custom slider). Otherwise nail Track 1.

requirements.txt

@@ -1 +1,3 @@
-gradio[mcp]
+gradio==3.53.1
+llama-index==0.6.9
+openai==0.27.0

utils/llama_index_utils.py

@@ -0,0 +1,16 @@
+# llama_index_utils.py
+from llama_index import SimpleDirectoryReader, GPTVectorStoreIndex
+
+_index = None
+
+def build_index(data_path="data/icd10cm_tabular_2025"):
+    global _index
+    if _index is None:
+        docs = SimpleDirectoryReader(data_path).load_data()
+        _index = GPTVectorStoreIndex.from_documents(docs)
+    return _index
+
+def query_symptoms(prompt: str, top_k: int = 5):
+    idx = build_index()
+    qe = idx.as_query_engine(similarity_top_k=top_k)
+    return qe.query(prompt)