feat(bot): add DeepSeek crisis fallback engine for Jev failure and remove difflib heuristics

- Add app/fallback_matcher.py for intelligent crisis entity matching with token and latency observability
- Route Jev failure/timeouts in app/matching.py to DeepSeek fallback engine
- Handle MatchingFailureError defensively in handlers/__init__.py with honest error message
- Enforce strict 'new' entity status in handlers/quote.py and implement defensive get_or_create on commit
- Clean up app/resolve.py to eliminate difflib and character ratio heuristics
- Pass all smoke tests (59/59), fallback matching tests (15/15), and matching unit tests (13/13)
This commit is contained in:
2026-10-16 02:05:01 +03:30
parent 76b297f823
commit 9e848f535f
9 changed files with 870 additions and 124 deletions
+20 -6
View File
@@ -229,16 +229,30 @@ async def handle_message(ctx: BotContext, chat_id: int, message: dict) -> None:
await obs.event(trace_id, "entities.extracted",
payload={"patch": patch, "intent": intent, "entity_mentions": mentions})
# Phase 3: Jev per-field matching — only when it can be applied (flag-gated;
# a failure leaves jev_map empty and the difflib path runs unchanged).
# Entity matching (Jev primary, DeepSeek crisis fallback)
jev_map: dict = {}
if settings.JEV_ENABLED and mentions and intent in ("register_quote", "update_draft", "add_material"):
if mentions and intent in ("register_quote", "update_draft", "add_material"):
try:
from app import matching
jev_map = await matching.resolve_mentions(ctx, patch, mentions, trace_id)
jev_map = await matching.resolve_mentions(
ctx, patch, mentions, trace_id=trace_id, raw_input=input_payload
)
except matching.MatchingFailureError:
logger.error("entity matching failed completely (crisis fallback failed)")
await ctx.send(
chat_id,
"⚠️ سامانه در پردازش هوشمند موجودیت‌ها با اختلال مواجه شد. لطفاً دوباره پیام را ارسال کنید.",
reply_to=message_id,
)
return
except Exception:
logger.exception("jev matching failed")
jev_map = {}
logger.exception("unexpected matching failure")
await ctx.send(
chat_id,
"⚠️ سامانه در پردازش هوشمند موجودیت‌ها با اختلال مواجه شد. لطفاً دوباره پیام را ارسال کنید.",
reply_to=message_id,
)
return
if source == "voice":
await ctx._log(chat_id, "in", "voice", consensus_text)
+12 -9
View File
@@ -9,7 +9,7 @@ from app.drafts import (
KIND_MATERIAL, merge_material_patch, material_missing, new_material_payload,
)
from app.keyboards import pick_buttons, simple_review_keyboard
from app.resolve import safe_resolve
from app.nlu import normalize_text
from app.texts import ERR_GENERIC, esc, material_committed_report, material_review_text
logger = logging.getLogger("bot.material")
@@ -45,16 +45,19 @@ async def resolve_payload(ctx, payload: dict, trace_id: str | None = None,
elif hit["status"] == "ambiguity" and hit.get("candidates"):
options["unit"] = [{"id": c["id"], "name": c["name"]}
for c in hit["candidates"][:6]]
elif hit["status"] == "new":
options.pop("unit", None)
else:
res = await safe_resolve(ctx.resolver, "unit", payload["unit_name"])
await obs.log_resolution(trace_id, "unit", payload["unit_name"], res.status,
[{"id": c.id, "name": c.name} for c in (res.candidates or [])[:4]])
if res.resolved:
payload["canonical_unit_id"] = res.candidates[0].id
payload["unit_name"] = res.candidates[0].name
try:
units = await ctx.resolver.units()
except Exception:
units = []
norm_u = normalize_text(payload["unit_name"])
exact = [u for u in units if normalize_text(u.name) == norm_u or (u.extra or {}).get("code") == norm_u]
if len(exact) == 1:
payload["canonical_unit_id"] = exact[0].id
payload["unit_name"] = exact[0].name
options.pop("unit", None)
elif res.candidates:
options["unit"] = [{"id": c.id, "name": c.name} for c in res.candidates]
else:
options.pop("unit", None)
+154 -70
View File
@@ -10,6 +10,7 @@ from app.drafts import (
new_quote_payload, quote_ready,
)
from app.keyboards import quote_review_keyboard
from app.nlu import normalize_text
from app.resolve import safe_resolve
from app.texts import ERR_GENERIC, quote_committed_report, quote_review_text, esc
@@ -67,15 +68,9 @@ async def handle_entities(
# ---------------------------------------------------------------------------
# resolution
# resolution (pure intelligence; zero local difflib heuristics)
# ---------------------------------------------------------------------------
async def _log_res(trace_id, field: str, mention, res) -> None:
"""Observability: record one difflib resolution attempt (baseline for phase 6)."""
await obs.log_resolution(trace_id, field, mention, res.status,
[{"id": c.id, "name": c.name} for c in (res.candidates or [])[:4]])
async def resolve_payload(ctx, payload: dict, trace_id: str | None = None,
jev_map: dict | None = None) -> None:
options = payload.setdefault("options", {})
@@ -86,9 +81,14 @@ async def resolve_payload(ctx, payload: dict, trace_id: str | None = None,
if hit["status"] == "exact":
matching.apply_exact(payload, hit)
options.pop("sup", None)
elif hit["status"] == "ambiguity" and hit.get("candidates"):
options["sup"] = [{"id": c["id"], "name": c["name"]} for c in hit["candidates"][:6]]
elif hit["status"] == "new":
options.pop("sup", None)
if hit.get("inferred_category_slug"):
payload["supplier_inferred_category_slug"] = hit["inferred_category_slug"]
else:
res = await safe_resolve(ctx.resolver, "supplier", payload["supplier_name"])
await _log_res(trace_id, "supplier", payload["supplier_name"], res)
if res.resolved:
c = res.candidates[0]
payload["supplier_id"] = c.id
@@ -111,18 +111,25 @@ async def resolve_payload(ctx, payload: dict, trace_id: str | None = None,
if hit["status"] == "exact":
matching.apply_exact(payload, hit)
payload["project_name"] = hit["name"]
resolved_projects[hit["name"]] = hit["id"]
resolved_projects[payload["project_name"]] = hit["id"]
if hit["id"] not in payload.get("project_ids", []):
payload.setdefault("project_ids", []).append(hit["id"])
options.pop("prj", None)
elif hit["status"] == "ambiguity" and hit.get("candidates"):
options["prj"] = [{"id": c["id"], "name": c["name"]} for c in hit["candidates"][:6]]
elif hit["status"] == "new":
options.pop("prj", None)
else:
res = await safe_resolve(ctx.resolver, "project", payload["project_name"])
await _log_res(trace_id, "project", payload["project_name"], res)
if res.resolved:
c = res.candidates[0]
payload["project_id"] = c.id
payload["project_name"] = c.name
resolved_projects[c.name] = c.id
resolved_projects[payload["project_name"]] = c.id
if payload["project_id"] not in payload.get("project_ids", []):
payload.setdefault("project_ids", []).append(payload["project_id"])
if c.id not in payload.get("project_ids", []):
payload.setdefault("project_ids", []).append(c.id)
options.pop("prj", None)
elif res.candidates:
options["prj"] = [{"id": c.id, "name": c.name} for c in res.candidates]
@@ -131,21 +138,23 @@ async def resolve_payload(ctx, payload: dict, trace_id: str | None = None,
for p_name in payload.get("project_names") or []:
hit = matching.lookup(jev_map, p_name)
if hit is not None:
if hit["status"] == "exact":
matching.apply_exact(payload, hit)
continue
res = await safe_resolve(ctx.resolver, "project", p_name)
await _log_res(trace_id, "project", p_name, res)
if res.resolved:
c = res.candidates[0]
p_id = c.id
resolved_projects[c.name] = p_id
resolved_projects[p_name] = p_id
if p_id not in payload.get("project_ids", []):
payload.setdefault("project_ids", []).append(p_id)
if hit is not None and hit.get("status") == "exact":
matching.apply_exact(payload, hit)
resolved_projects[hit["name"]] = hit["id"]
resolved_projects[p_name] = hit["id"]
if hit["id"] not in payload.get("project_ids", []):
payload.setdefault("project_ids", []).append(hit["id"])
if not payload.get("project_id"):
payload["project_id"] = p_id
payload["project_id"] = hit["id"]
elif hit is None:
res = await safe_resolve(ctx.resolver, "project", p_name)
if res.resolved:
resolved_projects[p_name] = res.candidates[0].id
resolved_projects[res.candidates[0].name] = res.candidates[0].id
if res.candidates[0].id not in payload.get("project_ids", []):
payload.setdefault("project_ids", []).append(res.candidates[0].id)
if not payload.get("project_id"):
payload["project_id"] = res.candidates[0].id
for i, item in enumerate(payload.get("items") or [], start=1):
if not item.get("material_id") and item.get("material_name"):
@@ -154,11 +163,15 @@ async def resolve_payload(ctx, payload: dict, trace_id: str | None = None,
if hit["status"] == "exact":
matching.apply_exact(payload, hit)
options.pop(f"mat{i}", None)
elif hit["status"] == "ambiguity" and hit.get("candidates"):
options[f"mat{i}"] = [{"id": c["id"], "name": c["name"]} for c in hit["candidates"][:6]]
elif hit["status"] == "new":
options.pop(f"mat{i}", None)
else:
res = await safe_resolve(ctx.resolver, "material", item["material_name"])
await _log_res(trace_id, "material", item["material_name"], res)
if res.resolved:
_set_material(item, res.candidates[0].id, res.candidates[0].name, res.candidates[0].extra)
c = res.candidates[0]
_set_material(item, c.id, c.name, c.extra)
options.pop(f"mat{i}", None)
elif res.candidates:
options[f"mat{i}"] = [{"id": c.id, "name": c.name} for c in res.candidates]
@@ -179,15 +192,13 @@ async def _resolve_unit(ctx, item: dict, trace_id: str | None = None,
return
if item.get("unit_name"):
hit = matching.lookup(jev_map, item["unit_name"])
if hit is not None:
if hit["status"] == "exact":
item["unit_id"] = hit["id"]
item["unit_explicit"] = True
item["unit_name"] = hit["name"]
return
else:
if hit is not None and hit.get("status") == "exact":
item["unit_id"] = hit["id"]
item["unit_explicit"] = True
item["unit_name"] = hit["name"]
return
if hit is None:
res = await safe_resolve(ctx.resolver, "unit", item["unit_name"])
await _log_res(trace_id, "unit", item["unit_name"], res)
if res.resolved:
item["unit_id"] = res.candidates[0].id
item["unit_explicit"] = True
@@ -369,46 +380,106 @@ async def commit(ctx, chat_id, user, draft, only_resolved: bool = False,
newly_created: dict = {}
# 1. Sequential pre-creation for new supplier if user wants to commit
# 1. Defensive pre-creation for new supplier (get_or_create)
if not payload.get("supplier_id") and payload.get("supplier_name"):
cat_id = payload.get("supplier_category_id")
if not cat_id:
cats = await ctx.resolver.categories()
if cats:
cat_id = cats[0].id
if cat_id:
try:
sup_body = {
"name": payload["supplier_name"],
"category_id": cat_id,
}
if payload.get("supplier_phone"):
sup_body["phone"] = payload["supplier_phone"]
created_sup = await ctx.api.create_supplier(sup_body, actor_id=user.actor_id)
payload["supplier_id"] = created_sup["id"]
newly_created["supplier"] = payload["supplier_name"]
ctx.resolver.invalidate("suppliers")
except Exception as e:
logger.warning("Could not auto-create supplier: %s", e)
sup_name = payload["supplier_name"].strip()
# 1.a) Check if already in DB
try:
existing_sups = await ctx.api.list_suppliers(q=sup_name, page_size=10)
for s in existing_sups:
if normalize_text(s.get("name") or "") == normalize_text(sup_name):
payload["supplier_id"] = str(s["id"])
payload["supplier_name"] = s.get("name")
break
except Exception as e:
logger.warning("Could not check existing suppliers: %s", e)
# 2. Sequential pre-creation for new materials
# 1.b) If still not in DB, create it
if not payload.get("supplier_id"):
cat_id = payload.get("supplier_category_id")
if not cat_id and payload.get("supplier_inferred_category_slug"):
inferred_slug = payload["supplier_inferred_category_slug"]
try:
cats = await ctx.resolver.categories()
for c in cats:
if (c.extra or {}).get("slug") == inferred_slug or c.name == inferred_slug:
cat_id = c.id
break
except Exception:
pass
if not cat_id:
try:
cats = await ctx.resolver.categories()
if cats:
cat_id = cats[0].id
except Exception:
pass
if cat_id:
try:
sup_body = {
"name": sup_name,
"category_id": cat_id,
}
if payload.get("supplier_phone"):
sup_body["phone"] = payload["supplier_phone"]
created_sup = await ctx.api.create_supplier(sup_body, actor_id=user.actor_id)
payload["supplier_id"] = created_sup["id"]
newly_created["supplier"] = sup_name
ctx.resolver.invalidate("suppliers")
except Exception as e:
logger.warning("Could not auto-create supplier: %s — trying fallback fetch", e)
try:
existing_sups = await ctx.api.list_suppliers(q=sup_name, page_size=10)
for s in existing_sups:
if normalize_text(s.get("name") or "") == normalize_text(sup_name):
payload["supplier_id"] = str(s["id"])
payload["supplier_name"] = s.get("name")
break
except Exception:
pass
# 2. Defensive pre-creation for new materials (get_or_create)
created_mats = []
for i, item in enumerate(payload.get("items") or [], start=1):
if not item.get("material_id") and item.get("material_name") and item.get("unit_id"):
mat_name = item["material_name"].strip()
# 2.a) Check if already in DB
try:
created_mat = await ctx.api.create_material({
"name": item["material_name"],
"canonical_unit_id": int(item["unit_id"]),
}, actor_id=user.actor_id)
_set_material(item, created_mat["id"], created_mat.get("name") or item["material_name"])
created_mats.append(item["material_name"])
ctx.resolver.invalidate("materials")
existing_mats = await ctx.api.list_materials(q=mat_name, page_size=10)
for m in existing_mats:
if normalize_text(m.get("name") or "") == normalize_text(mat_name):
_set_material(item, str(m["id"]), m.get("name") or mat_name,
{"canonical_unit_id": m.get("canonical_unit_id")})
break
except Exception as e:
logger.warning("Could not auto-create material: %s", e)
logger.warning("Could not check existing materials: %s", e)
# 2.b) If still not in DB, create it
if not item.get("material_id"):
try:
created_mat = await ctx.api.create_material({
"name": mat_name,
"canonical_unit_id": int(item["unit_id"]),
}, actor_id=user.actor_id)
_set_material(item, created_mat["id"], created_mat.get("name") or mat_name)
created_mats.append(mat_name)
ctx.resolver.invalidate("materials")
except Exception as e:
logger.warning("Could not auto-create material %s: %s — trying fallback fetch", mat_name, e)
try:
existing_mats = await ctx.api.list_materials(q=mat_name, page_size=10)
for m in existing_mats:
if normalize_text(m.get("name") or "") == normalize_text(mat_name):
_set_material(item, str(m["id"]), m.get("name") or mat_name,
{"canonical_unit_id": m.get("canonical_unit_id")})
break
except Exception:
pass
if created_mats:
newly_created["materials"] = created_mats
# 3. Sequential pre-creation for new projects
# 3. Defensive pre-creation for new projects (get_or_create)
created_prjs = []
all_p_names = []
if payload.get("project_names"):
@@ -429,14 +500,26 @@ async def commit(ctx, chat_id, user, draft, only_resolved: bool = False,
all_p_names.append(it_p_str)
resolved_map = payload.setdefault("resolved_projects", {})
all_known_prjs = None
for p_name in all_p_names:
p_id = resolved_map.get(p_name)
if not p_id:
res = await safe_resolve(ctx.resolver, "project", p_name)
if res.resolved:
p_id = res.candidates[0].id
resolved_map[p_name] = p_id
else:
# 3.a) Check if project already exists in catalog
if all_known_prjs is None:
try:
all_known_prjs = await ctx.resolver.projects()
except Exception:
all_known_prjs = []
norm_pn = normalize_text(p_name)
for prj in all_known_prjs:
if normalize_text(prj.name) == norm_pn:
p_id = prj.id
resolved_map[p_name] = p_id
resolved_map[prj.name] = p_id
break
# 3.b) If not in catalog, create it
if not p_id:
try:
created_prj = await ctx.api.create_project({"name": p_name}, actor_id=user.actor_id)
p_id = created_prj.get("id")
@@ -444,6 +527,7 @@ async def commit(ctx, chat_id, user, draft, only_resolved: bool = False,
resolved_map[p_name] = p_id
created_prjs.append(p_name)
ctx.resolver.invalidate("projects")
all_known_prjs = None
except Exception as e:
logger.warning("Could not auto-create project %s: %s", p_name, e)
+19 -10
View File
@@ -8,7 +8,7 @@ from app.drafts import (
KIND_SUPPLIER, merge_supplier_patch, new_supplier_payload, supplier_missing,
)
from app.keyboards import pick_buttons, simple_review_keyboard
from app.resolve import safe_resolve
from app.nlu import normalize_text
from app.texts import ERR_GENERIC, esc, supplier_committed_report, supplier_review_text
logger = logging.getLogger("bot.supplier")
@@ -33,17 +33,26 @@ async def handle_entities(ctx, chat_id, user, patch: dict, source: str | None,
async def resolve_payload(ctx, payload: dict, trace_id: str | None = None) -> None:
options = payload.setdefault("options", {})
if not payload.get("category_id") and payload.get("category_name"):
res = await safe_resolve(ctx.resolver, "category", payload["category_name"])
await obs.log_resolution(trace_id, "category", payload["category_name"], res.status,
[{"id": c.id, "name": c.name} for c in (res.candidates or [])[:4]])
if res.resolved:
payload["category_id"] = res.candidates[0].id
payload["category_name"] = res.candidates[0].name
norm_cat = normalize_text(payload["category_name"])
try:
cats = await ctx.resolver.categories()
except Exception:
cats = []
exact = [c for c in cats if normalize_text(c.name) == norm_cat or (c.extra or {}).get("slug") == norm_cat]
if len(exact) == 1:
payload["category_id"] = exact[0].id
payload["category_name"] = exact[0].name
options.pop("cat", None)
elif res.candidates:
options["cat"] = [{"id": c.id, "name": c.name} for c in res.candidates]
else:
options.pop("cat", None)
partial = [c for c in cats if norm_cat in normalize_text(c.name) and len(norm_cat) >= 3]
if len(partial) == 1:
payload["category_id"] = partial[0].id
payload["category_name"] = partial[0].name
options.pop("cat", None)
elif partial:
options["cat"] = [{"id": c.id, "name": c.name} for c in partial]
else:
options.pop("cat", None)
async def apply_pick(ctx, chat_id, user, draft, field: str, choice: dict) -> None: