| """ |
| Chatbot logic — context-stuffs the extracted study material into the |
| system prompt and streams the response back. |
| """ |
|
|
| from prompts import CHAT_SYSTEM_PROMPT |
| from llm import groq, CHAT_MODEL |
| from rag import retrieve_relevant_chunks, find_direct_reference, clean_math_output |
|
|
| def extract_text_content(content) -> str: |
| """ |
| Gradio's Chatbot can hand back message content as either a plain string |
| or a list of content-part dicts (e.g. [{"type": "text", "text": "..."}]) |
| depending on version/message type. Normalize to a plain string either way. |
| """ |
| if isinstance(content, str): |
| return content |
| if isinstance(content, list): |
| return " ".join( |
| part.get("text", "") for part in content |
| if isinstance(part, dict) and part.get("type") == "text" |
| ) |
| return str(content) |
|
|
| def chat(history, study_context, vector_store): |
| """ |
| Streams a response from the LLM based on the chat history and study context. |
| """ |
| latest_message = extract_text_content(history[-1]["content"]) |
|
|
| relevant_chunks = retrieve_relevant_chunks(vector_store, latest_message, k=4) |
| direct_match = find_direct_reference(study_context, latest_message) |
|
|
| pieces = relevant_chunks + ([direct_match] if direct_match else []) |
| retrieved_context = "\n\n".join(pieces) |
|
|
| system_message = CHAT_SYSTEM_PROMPT.format(study_context=retrieved_context) |
|
|
| history_for_api = [{"role": h["role"], "content": h["content"]} for h in history] |
| messages = [{"role": "system", "content": system_message}] + history_for_api |
|
|
| stream = groq.chat.completions.create(model=CHAT_MODEL, messages=messages, stream=True) |
| response = "" |
| for chunk in stream: |
| response += chunk.choices[0].delta.content or "" |
| yield history + [{"role": "assistant", "content": clean_math_output(response)}] |
|
|
