Circuit-Forge/kiwi
3
0
Fork 0
Code Issues 12 Pull requests Projects Releases 3 Packages Wiki Activity Actions

fix(recipe_scan): revert to cf-docuvision path (GGUF backend now works)

Browse source 
Route recipe_scan back through task_allocate -> cf-docuvision -> DocuvisionClient
now that docuvision supports GGUF models via Qwen25VLChatHandler.

Two-step pipeline: docuvision OCRs image(s), LLMRouter structures OCR text to JSON.
Removes the non-functional cf-text image_url path (cf-text rejects content arrays).
This commit is contained in:
pyr0ball 2026-05-16 19:25:01 -07:00
parent c72b4415db
commit 21a9b85067
1 changed files with 19 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

24
app/services/recipe/recipe_scanner.py
View file

@ -266,18 +266,32 @@ def _call_vision_backend(
errors: list[str] = [] errors: list[str] = []
# 1. Try cf-orch task allocation → Qwen2-VL GGUF on cf-text (direct multimodal extraction). # 1. Try cf-orch task allocation → cf-docuvision (Qwen2-VL GGUF via llama.cpp).
# One-step: the VLM receives the image(s) directly and returns structured recipe JSON. # Two-step: docuvision OCRs the image(s), then LLMRouter structures the text into JSON.
cf_orch_url = os.environ.get("CF_ORCH_URL") cf_orch_url = os.environ.get("CF_ORCH_URL")
if cf_orch_url: if cf_orch_url:
try: try:
from app.services.task_inference import TaskNotRegistered, task_allocate from app.services.task_inference import TaskNotRegistered, task_allocate
from app.services.ocr.docuvision_client import DocuvisionClient
from circuitforge_core.llm.router import LLMRouter
try: try:
_progress("allocating", "Starting vision service...") _progress("allocating", "Starting vision service...")
with task_allocate("kiwi", "recipe_scan", service_hint="cf-text", ttl_s=120.0) as alloc: with task_allocate("kiwi", "recipe_scan", service_hint="cf-docuvision", ttl_s=120.0) as alloc:
_progress("scanning", "Extracting recipe from photo...") _progress("scanning", "Extracting recipe text from photo...")
text = _call_via_cf_text_vlm(alloc.url, image_paths, prompt) doc_client = DocuvisionClient(alloc.url)
ocr_parts: list[str] = []
for i, path in enumerate(image_paths):
result = doc_client.extract_text(path, hint="text")
prefix = f"(Page {i + 1} of the same recipe)\n" if len(image_paths) > 1 else ""
ocr_parts.append(f"{prefix}{result.text}")
combined_ocr = "\n\n".join(ocr_parts)
if not combined_ocr.strip():
raise ValueError("Docuvision returned no text — image may not be a recipe")
_progress("structuring", "Parsing recipe structure...")
text = LLMRouter().complete(_build_ocr_extraction_prompt(combined_ocr))
if text: if text:
return text return text