Patch/fix import bleed (#1527)

* Feature/tweak actions (#1507) * up * tweak actions * Sync JS SDK, Harmonize Python SDK KG Methods (#1511) * Feature/move logging (#1492) * move logging provider out * move logging provider to own directory, remove singleton * cleanup * fix refactoring tweak (#1496) * Fix JSON serialization and Prompt ID Bugs for Prompts (#1491) * Bug in get prompts * Add tests * Prevent verbose logging on standup * Remove kg as required key in config, await get_all_prompts * Remove reference to fragment id * comment out ingestion * complete logging port (#1499) * Feature/dev rebased (#1500) * Feature/move logging (#1493) * move logging provider out * move logging provider to own directory, remove singleton * cleanup * Update js package (#1498) * fix refactoring tweak (#1496) * Fix JSON serialization and Prompt ID Bugs for Prompts (#1491) * Bug in get prompts * Add tests * Prevent verbose logging on standup * Remove kg as required key in config, await get_all_prompts * Remove reference to fragment id * comment out ingestion * complete logging port (#1499) --------- Co-authored-by: Nolan Tremelling <[email protected]> * Fix handling for R2R exceptions (#1501) * fix doc test (#1502) * Harmonize python SDK KG methods for optional params, add missing JS methods --------- Co-authored-by: emrgnt-cmplxty <[email protected]> Co-authored-by: emrgnt-cmplxty <[email protected]> * Clean up pagination and offset around KG (#1519) * Move to R2R light for integration testing (#1521) * fix ollama pdf parser --------- Co-authored-by: Nolan Tremelling <[email protected]>
SciPhi-AI · Oct 30, 2024 · 680c327 · 680c327
1 parent 96e2367
commit 680c327
Show file tree

Hide file tree

Showing 3 changed files with 18 additions and 14 deletions.
diff --git a/py/core/configs/local_llm.toml b/py/core/configs/local_llm.toml
@@ -27,3 +27,10 @@ concurrent_request_limit = 2
 
 [orchestration]
 provider = "simple"
+
+
+[ingestion]
+vision_img_model = "ollama/llama3.2-vision"
+vision_pdf_model = "ollama/llama3.2-vision"
+[ingestion.extra_parsers]
+    pdf = "basic"
diff --git a/py/core/providers/ingestion/r2r/base.py b/py/core/providers/ingestion/r2r/base.py
@@ -202,23 +202,21 @@ async def parse(  # type: ignore
         else:
             t0 = time.time()
             contents = ""
-            parser_overrides = ingestion_config_override.get(
-                "parser_overrides", {}
+
+            def check_vlm(model_name: str) -> bool:
+                return "gpt-4o" in model_name
+
+            is_not_vlm = not check_vlm(
+                ingestion_config_override.get("vision_pdf_model")
+                or self.config.vision_pdf_model
             )
-            if document.document_type.value in parser_overrides:
+
+            if document.document_type == DocumentType.PDF and is_not_vlm:
                 logger.info(
-                    f"Using parser_override for {document.document_type} with input value {parser_overrides[document.document_type.value]}"
+                    f"Reverting to basic PDF parser as the provided is not a proper VLM model."
                 )
-                # TODO - Cleanup this approach to be less hardcoded
-                if (
-                    document.document_type != DocumentType.PDF
-                    or parser_overrides[DocumentType.PDF.value] != "zerox"
-                ):
-                    raise ValueError(
-                        "Only Zerox PDF parser override is available."
-                    )
                 async for text in self.parsers[
-                    f"zerox_{DocumentType.PDF.value}"
+                    f"basic_{DocumentType.PDF.value}"
                 ].ingest(file_content, **ingestion_config_override):
                     contents += text + "\n"
             else:

diff --git a/py/r2r.toml b/py/r2r.toml
@@ -15,7 +15,6 @@ require_email_verification = true
 default_admin_email = "[email protected]"
 default_admin_password = "change_me_immediately"
 
-
 [completion]
 provider = "litellm"
 concurrent_request_limit = 256