zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-08-04 14:50:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

Feat: tenant llm provider (#14595)

Browse Source 
### What problem does this PR solve?

Python implementation of the Go-based model_provider API suite.

### Type of change

- [x] New Feature (non-breaking change which adds functionality)

---------

Co-authored-by: bill <yibie_jingnian@163.com>
This commit is contained in:
Lynn
2026-05-29 17:39:41 +08:00
committed by GitHub
parent b79f79d9b9
commit dc4b82523b
148 changed files with 6059 additions and 3075 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
rag/flow/parser/parser.py
View File

@@ -27,7 +27,7 @@ from PIL import Image
from api.db.services.file2document_service import File2DocumentService
from api.db.services.file_service import FileService
from api.db.services.llm_service import LLMBundle
from api.db.joint_services.tenant_model_service import get_model_config_by_type_and_name, get_tenant_default_model_by_type
from api.db.joint_services.tenant_model_service import get_tenant_default_model_by_type, get_model_config_from_provider_instance
from common import settings
from common.constants import LLMType
from common.misc_utils import get_uuid, thread_pool_exec
@@ -381,7 +381,7 @@ class Parser(ProcessBase):
raise RuntimeError("MinerU model not configured. Please add MinerU in Model Providers or set MINERU_* env.")
tenant_id = self._canvas._tenant_id
ocr_model_config = get_model_config_by_type_and_name(tenant_id, LLMType.OCR, parser_model_name)
ocr_model_config = get_model_config_from_provider_instance(tenant_id, LLMType.OCR, parser_model_name)
ocr_model = LLMBundle(tenant_id, ocr_model_config, lang=conf.get("lang", "Chinese"))
pdf_parser = ocr_model.mdl
@@ -458,7 +458,7 @@ class Parser(ProcessBase):
raise RuntimeError("OpenDataLoader model not configured. Please add OpenDataLoader in Model Providers.")
tenant_id = self._canvas._tenant_id
ocr_model_config = get_model_config_by_type_and_name(tenant_id, LLMType.OCR, parser_model_name)
ocr_model_config = get_model_config_from_provider_instance(tenant_id, LLMType.OCR, parser_model_name)
ocr_model = LLMBundle(tenant_id, ocr_model_config)
pdf_parser = ocr_model.mdl
@@ -563,7 +563,7 @@ class Parser(ProcessBase):
raise RuntimeError("PaddleOCR model not configured. Please add PaddleOCR in Model Providers or set PADDLEOCR_* env.")
tenant_id = self._canvas._tenant_id
ocr_model_config = get_model_config_by_type_and_name(tenant_id, LLMType.OCR, parser_model_name)
ocr_model_config = get_model_config_from_provider_instance(tenant_id, LLMType.OCR, parser_model_name)
ocr_model = LLMBundle(tenant_id, ocr_model_config)
pdf_parser = ocr_model.mdl
@@ -593,7 +593,7 @@ class Parser(ProcessBase):
# Vision parser treats each page as a large image block.
else:
if conf.get("parse_method"):
vision_model_config = get_model_config_by_type_and_name(self._canvas._tenant_id, LLMType.IMAGE2TEXT, conf["parse_method"])
vision_model_config = get_model_config_from_provider_instance(self._canvas._tenant_id, LLMType.IMAGE2TEXT, conf["parse_method"])
else:
vision_model_config = get_tenant_default_model_by_type(self._canvas._tenant_id, LLMType.IMAGE2TEXT)
vision_model = LLMBundle(self._canvas._tenant_id, vision_model_config, lang=self._param.setups["pdf"].get("lang"))
@@ -1067,7 +1067,7 @@ class Parser(ProcessBase):
else:
lang = conf["lang"]
# use VLM to describe the picture
cv_model_config = get_model_config_by_type_and_name(self._canvas.get_tenant_id(), LLMType.IMAGE2TEXT, conf["parse_method"])
cv_model_config = get_model_config_from_provider_instance(self._canvas.get_tenant_id(), LLMType.IMAGE2TEXT, conf["parse_method"])
cv_model = LLMBundle(self._canvas.get_tenant_id(), cv_model_config, lang=lang)
img_binary = io.BytesIO()
img.save(img_binary, format="JPEG")
@@ -1103,7 +1103,7 @@ class Parser(ProcessBase):
tmpf.write(blob)
tmpf.flush()
tmp_path = os.path.abspath(tmpf.name)
seq2txt_model_config = get_model_config_by_type_and_name(self._canvas.get_tenant_id(), LLMType.SPEECH2TEXT, vlm["llm_id"])
seq2txt_model_config = get_model_config_from_provider_instance(self._canvas.get_tenant_id(), LLMType.SPEECH2TEXT, vlm["llm_id"])
seq2txt_mdl = LLMBundle(self._canvas.get_tenant_id(), seq2txt_model_config)
txt = seq2txt_mdl.transcription(tmp_path)
@@ -1116,7 +1116,7 @@ class Parser(ProcessBase):
conf = self._param.setups["video"]
vlm = conf.get("vlm")
self.set_output("output_format", conf["output_format"])
cv_model_config = get_model_config_by_type_and_name(self._canvas.get_tenant_id(), LLMType.IMAGE2TEXT, vlm["llm_id"])
cv_model_config = get_model_config_from_provider_instance(self._canvas.get_tenant_id(), LLMType.IMAGE2TEXT, vlm["llm_id"])
cv_mdl = LLMBundle(self._canvas.get_tenant_id(), cv_model_config)
video_prompt = str(conf.get("prompt", "") or "")
txt = asyncio.run(cv_mdl.async_chat(system="", history=[], gen_conf={}, video_bytes=blob, filename=name, video_prompt=video_prompt))