Skip to content

RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.

Update infinity json_extract_string usage (#18139)

qinling0210b73aac38f4bb
diff --git a/api/db/services/dialog_service.py b/api/db/services/dialog_service.py
index 79ef4bc1c9..a7ded8e8f1 100644
--- a/api/db/services/dialog_service.py
+++ b/api/db/services/dialog_service.py
@@ -1141,8 +1141,8 @@ RULES:
- Question mentions "not null" or "excluding null"
- Add NULL check for count specific column
- DO NOT add NULL check for COUNT(*) queries (COUNT(*) counts all rows including nulls)
-7. json_extract_string() returns JSON-quoted strings ("value"), so WHERE comparisons MUST wrap values in double-quotes inside single-quotes (no spaces between quotes): '"value"' (e.g. WHERE json_extract_string(chunk_data, '$.name') = '"Alice"')
-8. For partial text search, use LIKE with wildcards: '"%value%"' (e.g. WHERE json_extract_string(chunk_data, '$.name') LIKE '"%Alice%"')
+7. json_extract_string() returns plain (unquoted) strings, so WHERE comparisons use plain single-quoted values: 'value' (e.g. WHERE json_extract_string(chunk_data, '$.name') = 'Alice')
+8. For partial text search, use LIKE with wildcards: '%value%' (e.g. WHERE json_extract_string(chunk_data, '$.name') LIKE '%Alice%')
9. Output ONLY the SQL, no explanations"""
user_prompt = """Table: {}
Fields (EXACT case): {}
diff --git a/internal/service/chat_pipeline.go b/internal/service/chat_pipeline.go
index 29d46142e9..f029701a8c 100644
--- a/internal/service/chat_pipeline.go
+++ b/internal/service/chat_pipeline.go
@@ -2966,8 +2966,8 @@ RULES:
- Question mentions "not null" or "excluding null"
- Add NULL check for count specific column
- DO NOT add NULL check for COUNT(*) queries (COUNT(*) counts all rows including nulls)
-7. json_extract_string() returns JSON-quoted strings ("value"), so WHERE comparisons MUST wrap values in double-quotes inside single-quotes (no spaces between quotes): '"value"' (e.g. WHERE json_extract_string(chunk_data, '$.name') = '"Alice"')
-8. For partial text search, use LIKE with wildcards: '"%value%"' (e.g. WHERE json_extract_string(chunk_data, '$.name') LIKE '"%Alice%"')
+7. json_extract_string() returns plain (unquoted) strings, so WHERE comparisons use plain single-quoted values: 'value' (e.g. WHERE json_extract_string(chunk_data, '$.name') = 'Alice')
+8. For partial text search, use LIKE with wildcards: '%value%' (e.g. WHERE json_extract_string(chunk_data, '$.name') LIKE '%Alice%')
9. Output ONLY the SQL, no explanations`
// infinitySQLUserPromptTemplate has 4 %s placeholders: