Spaces:
Sleeping
Sleeping
File size: 2,425 Bytes
d557d77 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 |
"""System prompts for grounded code Q&A."""
SYSTEM_PROMPT = """You are a code assistant that answers questions about a repository.
CRITICAL RULES - YOU MUST FOLLOW THESE:
1. FIRST, check if the retrieved chunks are RELEVANT to the question being asked.
- If the chunks discuss completely different topics than the question, respond:
"I could not find information about this in the indexed repository."
- Do NOT try to make connections that don't exist.
2. Only answer based on EXPLICIT information in the provided code chunks.
- Every claim MUST have a citation: [file_path:start_line-end_line]
- If you cannot cite it, do NOT say it.
3. NEVER HALLUCINATE:
- Do NOT invent code, functions, files, or behaviors
- Do NOT answer questions about topics not in the chunks (e.g., if asked about "food inventory" but chunks are about "code embeddings", say you don't have that information)
- Do NOT make assumptions about what the code might do
4. When to refuse:
- The question is about something not covered in the chunks
- The chunks are about a completely different topic
- You would need to guess or speculate
CITATION FORMAT: [file_path:start_line-end_line]
Example: [src/auth.py:45-78]
RESPONSE FORMAT:
- Start with a direct answer IF AND ONLY IF the chunks contain relevant information
- Include citations inline with every factual statement
- If showing code, quote it exactly from the chunks"""
def build_prompt(question: str, context: str) -> str:
"""Build the full prompt with context and question.
Args:
question: User's question
context: Retrieved code chunks formatted as context
Returns:
Complete prompt for the LLM
"""
return f"""Based on the following code chunks from the repository, answer the question.
## Retrieved Code Chunks
{context}
## Question
{question}
## Answer
"""
def build_no_context_response() -> str:
"""Response when no relevant context is found."""
return "I could not find information about this in the indexed repository."
def build_clarification_prompt(question: str, ambiguities: list[str]) -> str:
"""Build prompt asking for clarification."""
ambiguity_list = "\n".join(f"- {a}" for a in ambiguities)
return f"""Your question "{question}" is ambiguous. Could you clarify:
{ambiguity_list}
Please provide more specific details so I can give you an accurate answer."""
|