Light1YT
fceaaf9a2c
All checks were successful
CI / changes (push) Successful in 6s
CI / frontend-tests (push) Has been skipped
CI / changes (pull_request) Successful in 5s
CI / frontend-tests (pull_request) Has been skipped
Deploy / changes (push) Successful in 5s
Deploy / build-frontend (push) Has been skipped
Deploy / build-backend (push) Successful in 2m41s
CI / backend-tests (push) Successful in 6m24s
CI / backend-tests (pull_request) Successful in 6m23s
Deploy / build-worker (push) Successful in 3m26s
Deploy / deploy (push) Successful in 1m12s
Add the LLM prose-composition path for the parcel-forecast chat, layered over the deterministic Step-1 fallback which stays the safety net. - chat/tools.py: 5 read-only section tools (exec_summary, product_recommendation, forecast, risks, scenarios) — pure slices of the loaded report dict, no DB/ recompute, graceful on missing sections. market_now (raw analyze blob) and meta are deliberately NOT exposed -> highest-PII data cannot reach the LLM. - chat/safe_payload.py: the §19 gate — single place that builds the outbound SafePayload from a section-aggregate allowlist; honors is_confidential hard-block. - chat/orchestrator.py: manual tool-call loop with call-cap/termination, real grounded_in provenance; any LLMResult.ok=False (disabled/timeout/rate_limited/ redaction_refused/call_cap/provider_error/empty) degrades to the deterministic answer. - llm/prompts.py: versioned chat_system@v1 — answer only from sections, never fabricate numbers, advisory tone, decline out-of-scope. - api/v1/chat.py: branch on settings.llm_enabled; sync complete bridged via run_in_threadpool. Default-off -> deterministic path, no provider built. - Tests: fake provider only (no network), planted-secret redaction-boundary + per-reason fallback + call-cap + numbers-from-report coverage. Refs #957
25 lines
1.2 KiB
Python
25 lines
1.2 KiB
Python
"""Чат по §22-форсайту участка (#957). Step 1 детерминированный + Step 2 LLM tool-loop.
|
||
|
||
Под-модули:
|
||
• retrieval — read-only достаёт персистентный SiteFinderReport участка.
|
||
• intents — роутинг RU-вопроса в intent + шаблонный RU-ответ (Step-1, FALLBACK).
|
||
• tools — read-only секционные tool'ы (срезы загруженного report_dict).
|
||
• safe_payload — §19-гейт: единственная сборка исходящего SafePayload (аллоулист).
|
||
• orchestrator — LLM tool-loop + RU-проза поверх отчёта; при сбое LLM → Step-1 fallback.
|
||
|
||
LLM-слой ОПЦИОНАЛЬНЫЙ: при llm_enabled=False или любом сбое возвращается детерм. Step-1.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
from app.services.chat.intents import render_answer, route_intent
|
||
from app.services.chat.orchestrator import ChatResult, orchestrate_chat
|
||
from app.services.chat.retrieval import get_report_for_chat
|
||
|
||
__all__ = [
|
||
"ChatResult",
|
||
"get_report_for_chat",
|
||
"orchestrate_chat",
|
||
"render_answer",
|
||
"route_intent",
|
||
]
|