pyr0ball
6fc8e7faa6
Citation dataclass gains bm25_score field populated from the retrieved chunk. chat.py serializes it. api.ts interface updated to include it. ChatView passes :bm25-score to CitationPanel so the Nat20 threshold check in onMounted actually has data to evaluate.
127 lines
3.1 KiB
Python
127 lines
3.1 KiB
Python
# app/api/chat.py
|
|
"""
|
|
RAG chat endpoint — retrieves relevant page chunks and synthesizes an answer.
|
|
|
|
BSL 1.1 — BYOK gate: requires PAGEPIPER_OLLAMA_URL or a Paid tier license.
|
|
Returns 402 with clear upgrade message if neither is configured.
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import logging
|
|
import os
|
|
|
|
from fastapi import APIRouter, HTTPException
|
|
from pydantic import BaseModel
|
|
|
|
from app.services.retriever import Retriever
|
|
from app.services.synthesizer import Synthesizer
|
|
|
|
logger = logging.getLogger(__name__)
|
|
router = APIRouter(prefix="/api/chat", tags=["chat"])
|
|
|
|
|
|
class ChatTurn(BaseModel):
|
|
role: str # "user" | "assistant"
|
|
content: str
|
|
|
|
|
|
class ChatRequest(BaseModel):
|
|
message: str
|
|
history: list[ChatTurn] = []
|
|
doc_ids: list[str] | None = None
|
|
top_k: int = 5
|
|
|
|
|
|
class ChatResponse(BaseModel):
|
|
answer: str
|
|
citations: list[dict]
|
|
|
|
|
|
def _get_llm_router():
|
|
"""Return LLMRouter if Ollama configured, else None."""
|
|
from app.config import get_llm_config
|
|
|
|
cfg = get_llm_config()
|
|
if cfg is None:
|
|
return None
|
|
from circuitforge_core.llm import LLMRouter
|
|
|
|
return LLMRouter(cfg)
|
|
|
|
|
|
def _get_db_path() -> str:
|
|
"""Read lazily so test fixtures take effect."""
|
|
import pathlib
|
|
|
|
data_dir = pathlib.Path(os.environ.get("PAGEPIPER_DATA_DIR", "data"))
|
|
return str(data_dir / "pagepiper.db")
|
|
|
|
|
|
def _get_vec_db_path() -> str:
|
|
import pathlib
|
|
|
|
data_dir = pathlib.Path(os.environ.get("PAGEPIPER_DATA_DIR", "data"))
|
|
return str(data_dir / "pagepiper_vecs.db")
|
|
|
|
|
|
def _require_llm():
|
|
"""Return LLMRouter or raise 402."""
|
|
llm = _get_llm_router()
|
|
if llm is None:
|
|
raise HTTPException(
|
|
status_code=402,
|
|
detail={
|
|
"error": "ollama_required",
|
|
"message": (
|
|
"RAG chat requires Ollama. Set PAGEPIPER_OLLAMA_URL in your .env file, "
|
|
"then restart. Run: ollama pull nomic-embed-text && ollama pull mistral:7b"
|
|
),
|
|
},
|
|
)
|
|
return llm
|
|
|
|
|
|
@router.post("")
|
|
def chat(req: ChatRequest) -> ChatResponse:
|
|
llm = _require_llm()
|
|
|
|
from app.main import _bm25
|
|
|
|
retriever = Retriever(_bm25)
|
|
chunks = retriever.hybrid_search(
|
|
query=req.message,
|
|
top_k=req.top_k,
|
|
doc_ids=req.doc_ids,
|
|
db_path=_get_db_path(),
|
|
vec_db_path=_get_vec_db_path(),
|
|
llm=llm,
|
|
)
|
|
|
|
if not chunks:
|
|
return ChatResponse(
|
|
answer=(
|
|
"I couldn't find any relevant passages. "
|
|
"Try a different query or check which documents are indexed."
|
|
),
|
|
citations=[],
|
|
)
|
|
|
|
synth = Synthesizer(llm)
|
|
result = synth.synthesize(
|
|
message=req.message,
|
|
history=[t.model_dump() for t in req.history],
|
|
chunks=chunks,
|
|
)
|
|
|
|
return ChatResponse(
|
|
answer=result.answer,
|
|
citations=[
|
|
{
|
|
"doc_id": c.doc_id,
|
|
"page_number": c.page_number,
|
|
"snippet": c.snippet,
|
|
"bm25_score": c.bm25_score,
|
|
}
|
|
for c in result.citations
|
|
],
|
|
)
|