Spaces:
Running
Running
Devdan Schretlen
Claude Opus 5 (1M context)
Expand the calendar corpus to every faculty, and rewrite the UI copy
fcdba1d | """ | |
| Router — the "adaptive" in adaptive RAG (Phase 4). | |
| One small function decides how much machinery a question deserves: | |
| simple a single fact or page lookup -> the fast Phase-2 retriever | |
| complex needs several pages, a -> the agentic loop (decompose -> | |
| comparison, or multi-step reasoning multi-hop retrieve -> synthesize) | |
| The classification is a cheap zero-shot LLM call reusing the generation backend, so there | |
| is no extra model to configure. Anything the model returns other than a clear "complex" | |
| falls back to "simple" — the cheaper path is the safe default, so a misparse never sends a | |
| simple question down the expensive agent. | |
| """ | |
| from __future__ import annotations | |
| from . import generate, trace | |
| CLASSIFY_SYSTEM = ( | |
| "You are a query router for a UBC academic-calendar Q&A system. Classify the user's " | |
| "question as either 'simple' or 'complex'.\n" | |
| " simple: answerable from one calendar page (a single course's prerequisites or " | |
| "description, one program's requirements, one policy or date).\n" | |
| " complex: needs multiple courses or programs, a comparison (between programs, " | |
| "student cohorts, or calendar editions), or a prerequisite/requirement chain " | |
| "(e.g. 'compare X and Y', or 'what do I need to take before Z').\n" | |
| "Reply with exactly one word: simple or complex." | |
| ) | |
| def classify(question: str) -> str: | |
| """Return "simple" or "complex" for `question`. Defaults to "simple" on any surprise.""" | |
| trace.step("ROUTE - classify (simple vs complex)") | |
| reply = generate.complete(CLASSIFY_SYSTEM, question, purpose="route").strip().lower() | |
| route = "complex" if "complex" in reply else "simple" | |
| trace.event("route", decision=route, raw_reply=reply) | |
| trace.detail("decision", route) | |
| return route | |